[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

WO2022016941A1 - 行驶装置的避障路径的规划方法和装置 - Google Patents

行驶装置的避障路径的规划方法和装置 Download PDF

Info

Publication number
WO2022016941A1
WO2022016941A1 PCT/CN2021/090523 CN2021090523W WO2022016941A1 WO 2022016941 A1 WO2022016941 A1 WO 2022016941A1 CN 2021090523 W CN2021090523 W CN 2021090523W WO 2022016941 A1 WO2022016941 A1 WO 2022016941A1
Authority
WO
WIPO (PCT)
Prior art keywords
path
cost
obstacle avoidance
lateral
distance
Prior art date
Application number
PCT/CN2021/090523
Other languages
English (en)
French (fr)
Inventor
程思源
王超
王新宇
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to KR1020237005632A priority Critical patent/KR20230041752A/ko
Priority to EP21847177.9A priority patent/EP4180894A4/en
Priority to JP2023504056A priority patent/JP2023535175A/ja
Priority to BR112023001066A priority patent/BR112023001066A2/pt
Publication of WO2022016941A1 publication Critical patent/WO2022016941A1/zh
Priority to US18/156,703 priority patent/US20230159056A1/en

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W60/00—Drive control systems specially adapted for autonomous road vehicles
    • B60W60/001—Planning or execution of driving tasks
    • B60W60/0011—Planning or execution of driving tasks involving control alternatives for a single driving scenario, e.g. planning several paths to avoid obstacles
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/34—Route searching; Route guidance
    • G01C21/3446—Details of route searching algorithms, e.g. Dijkstra, A*, arc-flags, using precalculated routes
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
    • B60W30/08—Active safety systems predicting or avoiding probable or impending collision or attempting to minimise its consequences
    • B60W30/095—Predicting travel path or likelihood of collision
    • B60W30/0956—Predicting travel path or likelihood of collision the prediction being responsive to traffic or environmental parameters
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
    • B60W50/0098—Details of control systems ensuring comfort, safety or stability not otherwise provided for
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/34—Route searching; Route guidance
    • G01C21/3453—Special cost functions, i.e. other than distance or default speed limit of road segments
    • G01C21/3461—Preferred or disfavoured areas, e.g. dangerous zones, toll or emission zones, intersections, manoeuvre types, segments such as motorways, toll roads, ferries
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
    • B60W2050/0001—Details of the control system
    • B60W2050/0043—Signal treatments, identification of variables or parameters, parameter estimation or state estimation
    • B60W2050/0052—Filtering, filters
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W2554/00—Input parameters relating to objects
    • B60W2554/80—Spatial relation or speed relative to objects
    • B60W2554/802—Longitudinal distance
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W2556/00—Input parameters relating to data
    • B60W2556/10—Historical data
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W2754/00—Output or target parameters relating to objects
    • B60W2754/10—Spatial relation or speed relative to objects
    • B60W2754/20—Lateral distance

Definitions

  • the present application relates to the technical field of automatic driving, and more particularly, to a method and device for planning an obstacle avoidance path of a traveling device.
  • Autonomous driving technology uses maps and positioning modules to obtain navigation information, uses sensing devices to perceive the surrounding environment, and combines decision-making, planning, and control modules to achieve autonomous vehicle navigation.
  • the path planning module is responsible for outputting the correct and effective driving path according to the upper-level navigation information, decision-making instructions and environmental perception results, so as to guide the correct driving of the automatic driving device.
  • the path planning module can further realize the planning of the obstacle avoidance path in combination with the obstacle information, so that the automatic driving device can drive safely in the dynamic environment, which is also the focus of path planning.
  • the existing obstacle avoidance path planning methods mainly include: sampling, optimization, search, potential field and feature points, etc. These obstacle avoidance methods usually have good planning effects in simple scenarios, but for real dynamic traffic scenarios, especially In complex and narrow scenes, the path is prone to shaking, and the safe and smooth driving of the vehicle cannot be guaranteed.
  • the present application provides a method and device for planning an obstacle avoidance path of a traveling device, so that the traveling device can avoid obstacles safely and stably in complex and narrow traffic scenarios.
  • a method for planning an obstacle avoidance path of a traveling device includes: acquiring a distance map, where the distance map includes a plurality of first grids, and each of the plurality of first grids has a first The grid is used to record the distance between the first grid and the first grid occupied by the nearest obstacle; according to the distance map, determine the first lateral background cost potential field in the near field range, the near field range is the same as that of the driving device.
  • the longitudinal distance is less than or equal to the range of the first threshold value
  • the first lateral background cost potential field is the first repulsive force field formed by the obstacle in the lateral position to the traveling device
  • the first lateral background cost potential field includes the target extremely small value point
  • the background cost on both lateral sides of the target minimum point is monotonic
  • the background cost represents the near-field lateral collision cost to the traveling device caused by obstacles in the lateral position within the near-field range; according to the first lateral
  • the background cost potential field determines the current obstacle avoidance path.
  • the above-mentioned driving device may be an autonomous vehicle, a robot, or other autonomous driving devices.
  • the repulsive force field is formed by the obstacles in the lateral position to the traveling device, which can represent the collision tendency of the obstacle to the traveling device, and guide the traveling device to avoid the obstacle, which can also be called a collision field or a risk field.
  • the first horizontal background cost potential field includes a target minimum point, and the background cost on both sides of the target minimum point is monotonic, which means that the background on both sides of the target minimum point in the first repulsion field is monotonic. The cost is monotonic.
  • the background cost on both sides of the target minimum point is monotonic, specifically, the potential field on the left side of the target minimum value point is monotonically decreasing, and the potential field on the right side of the target minimum value point is monotonically increasing.
  • the background costs on the lateral sides of the target minimum point are monotonic, which means that the background costs on both lateral sides of the target minimum point are larger than the background costs at the target minimum point.
  • the first lateral background cost potential field may be V-shaped.
  • determining the current obstacle avoidance path according to the first lateral background cost potential field may be determining the current obstacle avoidance path according to the lateral position of the target minimum point in the first lateral background cost potential field.
  • a distance map is first obtained, and then a first horizontal background cost potential field within the near field is constructed according to the distance map, and the first horizontal background cost potential field includes a target minimum value point, and the target minimum value is The background cost on both lateral sides of the value point is monotonic; then the current obstacle avoidance path is determined according to the first lateral background cost potential field.
  • the information of near-field obstacles is fully considered to avoid the influence of distant obstacles on the planned path;
  • the obstacle avoidance path planned based on the lateral collision trend of the obstacle can make the driving device maintain the best lateral distance from the obstacle in a complex and narrow environment, thereby improving the stability of the obstacle avoidance path.
  • determining a first lateral background cost potential field within the near-field range according to the distance map includes: determining a second lateral background within the near-field range according to the distance map a cost potential field, the second lateral background cost potential field is a second repulsive force field formed by the obstacles in the lateral position to the traveling device, and the second lateral background cost potential field includes at least one minimum value point; from the at least one Determine the target minimum point in the minimum value point; perform monotonicity processing on the background costs on both lateral sides of the target minimum value point to generate the first horizontal background cost potential field, and the slope of the monotonicity processing is greater than or equal to second threshold.
  • the second repulsive force field can represent the collision tendency of the obstacle to the traveling device before the monotonicity treatment. That is, monotonic processing of the second repulsion field results in the first repulsion field.
  • a second lateral background cost potential field within the near field is constructed according to the distance map, and then a target minimum value point is determined in at least one minimum value point in the second lateral background cost potential field , and perform monotonicity processing on the background costs on both lateral sides of the target minimum point to generate the first lateral background cost potential field, and the slope of the monotonic processing is greater than or equal to the second threshold. Therefore, the background cost potential field of the lateral position where the target minimum point is located is significantly lower than the background cost potential field at other lateral positions. The background cost potential field will increase significantly, so as to avoid slowly and smoothly and run to the horizontal position where the background cost potential field is the lowest to ensure a safe distance from the obstacle.
  • the target minimum value point may be the smallest minimum value point among all the minimum value points in which there is no obstacle in the near field range between the target minimum value point and the traveling device.
  • the self-vehicle can run smoothly to the lateral position where the target minimum point is located, avoiding the possibility of the traveling device and the target extreme Obstacles exist between the lateral positions of the minimum point, so that the ego vehicle cannot run to the lateral position of the target minimum point, thereby improving the stability of the path.
  • the method before acquiring the distance map, further includes: acquiring an occupancy map, where the occupancy map includes a plurality of second grids, among the plurality of second grids Each second grid of is used to record the probability of the existence of obstacles in the second grid; obtaining the distance map includes: obtaining the distance map according to the occupation map.
  • the probability of existing obstacles in the second grid is greater than or equal to a certain threshold, it can be considered that there are obstacles in the second grid; when the probability of existing obstacles in the second grid is less than a certain threshold, It can be considered that there are no obstacles in the second grid.
  • each second grid in the plurality of second grids may also be directly used to record whether there is an obstacle in the second grid.
  • obtaining the distance map according to the occupation map includes: filtering out the second grid located in the first area in the occupation map, the first area Located in front of the traveling device, if there is an obstacle in the first area, the traveling device cannot avoid the obstacle in the first area; the distance map is obtained according to the screened occupancy map.
  • the traveling device cannot avoid the obstacle in the first area. It can also be described as, if there is an obstacle in the first area, no matter which one the traveling device chooses None of the obstacle avoidance paths can avoid the obstacles in the first area.
  • the obstacle avoidance path planning is based on the occupancy map information provided by the upstream module, and the obstacle information provided by the upstream module not only includes the actual pose of the dynamic and static obstacles and the predicted trajectory projection of the obstacles, but also may include the self-vehicle.
  • the occupancy situation of obstacles caused by the false detection of the front part of the vehicle, or the obstacle information appears in the area on the occupied map due to the uneven ground. At this time, if the obstacles in this area are also considered, it will affect the path planning, so that the obstacles in this area must be avoided during the driving process, which will cause the path to oscillate greatly.
  • the grids that occupy the area in the map may be screened out.
  • the erroneous avoidance of the traveling device due to the erroneous detection of the obstacle is effectively avoided. It is ensured that the remaining occupied grids only contain the necessary information required by the obstacle avoidance algorithm, which improves the stability of the obstacle avoidance path.
  • the method before determining the second lateral background cost potential field within the near-field range according to the distance map, the method further includes: generating a feature vector, where the feature vector includes Path starting point and sampling target point; generating multiple candidate paths according to the reference path, the historical path and the feature vector; determining the second lateral background cost potential field within the near field range according to the distance map includes: according to the sampling The target point, the reference path is laterally offset to generate a plurality of virtual parallel line paths; the distance of each waypoint within the near field range of each virtual parallel line path in the multiple virtual parallel line paths is calculated according to the distance map.
  • the distance of the nearest obstacle to the waypoint calculate the background cost of the virtual parallel line according to the minimum value of the distance on each virtual parallel line path; calculate the background cost of the virtual parallel line path within the near field range according to the multiple virtual parallel line paths A second lateral background cost potential field within the near-field range is generated.
  • the method before determining the current obstacle avoidance path according to the first lateral background cost potential field, the method further includes: calculating the multiple candidate paths according to the distance map One or more of the collision cost, constraint cost, switching cost, lateral offset cost, or smoothness cost on each alternative path in Whether a collision will occur when driving up, the constraint cost is used to evaluate whether each waypoint on the alternative path is within the obstacle avoidance range, and the switching cost is used to evaluate the deviation between the alternative path and the historical path, The lateral offset cost is used to evaluate the degree of lateral deviation of the candidate path from the road centerline, and the smoothness cost is used to evaluate the smoothness of the candidate path; according to the first lateral background cost potential field, determine the current avoidance
  • the obstacle path includes: according to the background cost in the first lateral background cost potential field, or according to the background cost in the first lateral background cost potential field and the collision cost, constraint cost, switching cost, lateral offset cost or smoothness One or more of the costs, calculate
  • one or more of the collision cost, constraint cost, switching cost, lateral offset cost, or smoothness cost may also be considered, so that all The determined obstacle avoidance path can integrate more factors, so that the stability and safety of the obstacle avoidance path can be further improved.
  • the method further includes: determining a final obstacle avoidance path according to an arbitration result of the historical path and the current obstacle avoidance path.
  • the final obstacle avoidance path is determined according to the difference between the current obstacle avoidance path and the historical path. It avoids the situation that the obstacle avoidance path obtained purely based on the cost function or pure logic will produce inter-frame jumping when running in a complex and narrow environment, effectively reducing the planning method's dependence on the cost function, thereby ensuring the stability of the path.
  • determining the final obstacle avoidance path according to the arbitration result of the historical path and the current obstacle avoidance path includes: if the historical path has a collision, but the current obstacle avoidance path does not If there is a collision, the current obstacle avoidance path is used as the final obstacle avoidance path; or, if there is a collision between the historical path and the current obstacle avoidance path, and the collision distance of the current obstacle avoidance path and the collision distance of the historical path are combined If the absolute value of the difference is greater than or equal to the third threshold, the current obstacle avoidance path and the historical path with the larger collision distance are used as the final obstacle avoidance path; or, if both the historical path and the current obstacle avoidance path have collisions , and the absolute value of the difference between the collision distance of the current obstacle avoidance path and the collision distance of the historical path is less than the third threshold, then the historical path is used as the final obstacle avoidance path; or, if there is no collision on the historical path, and The ratio of the final cost
  • the current obstacle avoidance path is obviously better than the historical path, the current obstacle avoidance path is switched; if the current obstacle avoidance path is not obviously better than the historical path, the historical path is used. It avoids the situation that the obstacle avoidance path obtained purely based on the cost function or pure logic will generate inter-frame jumping when running in a complex and narrow environment, thereby ensuring the stability of the path.
  • the method further includes: performing temporal smoothing according to the historical path and the final obstacle avoidance path.
  • time domain smoothing may also be performed according to the historical path and the final obstacle avoidance path.
  • the final obstacle avoidance path follows the historical path, the obstacle avoidance path after time domain smoothing is still the historical path; if the final obstacle avoidance path is the current obstacle avoidance path, the time domain smoothing is performed according to the current obstacle avoidance path and the historical path. .
  • the method further includes: outputting the final obstacle avoidance path to a speed planner, where the speed planner is used for determining the obstacles on the final obstacle avoidance path according to the final obstacle avoidance path and the obstacles on the final obstacle avoidance path. information for speed planning.
  • the speed planner after outputting the final obstacle avoidance path, the speed planner will perform speed planning according to the obstacle information on the final obstacle avoidance path, and will perform deceleration planning if an obstacle that may collide is found on the path. . Thereby improving the security of the path.
  • a device for planning an obstacle avoidance path of a traveling device includes: an acquisition module and a processing module; the acquisition module is used to acquire a distance map, the distance map includes a plurality of first grids, the plurality of Each of the first grids is used to record the distance between the first grid and the first grid occupied by the nearest obstacle; the processing module is used to determine the distance in the near field according to the distance map.
  • a first lateral background cost potential field, the near-field range is a range in which the longitudinal distance from the traveling device is less than or equal to a first threshold, and the first lateral background cost potential field is the obstacle formed by the lateral position of the traveling device.
  • the first repulsion field, the first horizontal background cost potential field includes the target minimum point, the background cost on both sides of the target minimum point is monotonic, and the background cost represents the pair of obstacles in the horizontal position within the near field.
  • the near-field lateral collision cost formed by the traveling device; and the current obstacle avoidance path is determined according to the first lateral background cost potential field.
  • determining a first lateral background cost potential field within a near field range according to the distance map includes: determining a second lateral background cost within a near field range according to the distance map potential field, the second lateral background cost potential field is a second repulsive force field formed by the obstacle in the lateral position to the traveling device, and the second lateral background cost potential field includes at least one minimum value point; Determine the target minimum value point in the minimum value point; perform monotonicity processing on the background costs on both sides of the target minimum value point to generate the first horizontal background cost potential field, and the slope of the monotonicity processing is greater than or equal to the second threshold.
  • the acquiring module before acquiring the distance map, is further configured to: acquire an occupation map, where the occupation map includes a plurality of second grids, the plurality of second grids Each second grid in the grid is used to record the probability of the existence of an obstacle in the second grid; the processing module is further configured to obtain a distance map according to the occupation map.
  • the obtaining the distance map according to the occupancy map includes: filtering out a second grid located in a first area in the occupancy map, where the first area is located in the driving area In front of the device, if there is an obstacle in the first area, the traveling device cannot avoid the obstacle in the first area; the distance map is obtained according to the filtered occupancy map.
  • the processing module is further configured to: generate a feature vector, the feature vector Including a path starting point and a sampling target point; generating a plurality of candidate paths according to the reference path, the historical path and the feature vector; determining the second lateral background cost potential field within the near field range according to the distance map includes: According to the sampling target point, the reference path is laterally offset to generate a plurality of virtual parallel line paths; according to the distance map, each virtual parallel line path of the plurality of virtual parallel line paths within the near field range is calculated.
  • the distance between the waypoint and the nearest obstacle to the waypoint calculate the background cost of the virtual parallel line according to the minimum value of the distance on the path of each virtual parallel line; according to the multiple virtual parallel line paths within the near field range
  • the background cost of generates a second lateral background cost potential field within the near-field range.
  • the processing module before determining the current obstacle avoidance path according to the first lateral background cost potential field, is further configured to: calculate the distance map according to the distance map.
  • One or more of the collision cost, constraint cost, switching cost, lateral offset cost, or smoothness cost on each of the alternative paths wherein the collision cost is used to evaluate the running Whether a collision will occur when driving on the alternative route, the constraint cost is used to evaluate whether each waypoint on the alternative route is within the obstacle avoidance range, and the switching cost is used to evaluate the difference between the alternative route and the historical route.
  • the lateral deviation cost is used to evaluate the degree of lateral deviation of the candidate path from the road centerline, and the smoothness cost is used to evaluate the smoothness of the candidate path; according to the first lateral background cost potential field, Determining the current obstacle avoidance path includes: according to the background cost in the first lateral background cost potential field, or according to the background cost in the first lateral background cost potential field and the collision cost, constraint cost, switching cost, and lateral offset cost or one or more of smoothness costs, calculate the final cost on each alternative path; determine the current obstacle avoidance path according to the final cost.
  • the processing module is further configured to: determine the final obstacle avoidance path according to the arbitration result of the historical path and the current obstacle avoidance path.
  • determining the final obstacle avoidance path according to the arbitration result of the historical path and the current obstacle avoidance path includes: if the historical path collides, but the current obstacle avoidance path If the path does not collide, the current obstacle avoidance path is used as the final obstacle avoidance path; or, if both the historical path and the current obstacle avoidance path collide, and the collision distance of the current obstacle avoidance path collides with the historical path
  • the absolute value of the difference between the distances is greater than or equal to the third threshold, then the current obstacle avoidance path and the historical path with the larger collision distance are used as the final obstacle avoidance path; or, if both the historical path and the current obstacle avoidance path are If there is a collision, and the absolute value of the difference between the collision distance of the current obstacle avoidance path and the collision distance of the historical path is less than the third threshold, the historical path is used as the final obstacle avoidance path; or, if there is no collision on the historical path , and the ratio of the final cost of the current
  • the processing module is further configured to: perform time domain smoothing according to the historical path and the final obstacle avoidance path.
  • the apparatus may further include an output module for outputting the final obstacle avoidance path to the speed planner, where the speed planner is used for calculating the final obstacle avoidance path and the final avoidance path according to the The obstacle information on the obstacle path is used for speed planning.
  • a planning device comprising an input and output interface, a processor and a memory
  • the processor is used for controlling the input and output interface to send and receive signals or information
  • the memory is used for storing a computer program
  • the processor is used for calling from the memory And run the computer program, so that the planning apparatus executes the method in the above-mentioned first aspect.
  • a vehicle comprising various modules for performing the method described in the first aspect above.
  • a computer-readable medium stores a program code, which, when the computer program code is executed on a computer, causes the computer to execute the method in the above-mentioned first aspect .
  • a chip in a sixth aspect, includes a processor and a data interface, the processor reads an instruction stored in a memory through the data interface, and executes the method in the first aspect.
  • the chip may further include a memory, in which instructions are stored, the processor is configured to execute the instructions stored in the memory, and when the instructions are executed, the The processor is configured to perform the method in the first aspect above.
  • the above chip may specifically be a field programmable gate array FPGA or an application specific integrated circuit ASIC.
  • the method of the first aspect may specifically refer to the method in the first aspect and any one of the various implementation manners of the first aspect.
  • FIG. 1 is an example diagram of a Frenet coordinate system provided by an embodiment of the present application.
  • FIG. 2 is an example diagram of a complex and narrow urban road scene provided by an embodiment of the present application.
  • FIG. 3 is an example diagram of a system architecture for obstacle avoidance path planning provided by an embodiment of the present application.
  • FIG. 4 is an example diagram of a method for planning an obstacle avoidance path of a traveling device provided by an embodiment of the present application
  • FIG. 5 is an exemplary diagram of a core flow of a method for planning an obstacle avoidance path provided by an embodiment of the present application
  • FIG. 6 is an example diagram of a scenario for generating an alternative path provided by an embodiment of the present application.
  • FIG. 7 is an example diagram of another alternative path generation scenario provided by an embodiment of the present application.
  • FIG. 8 is an exemplary diagram of an invalid occupied grid provided by an embodiment of the present application.
  • FIG. 9 is an exemplary diagram of a full collision triangle provided by an embodiment of the present application.
  • FIG. 10 is an example diagram of converting an occupation map into a distance map provided by an embodiment of the present application.
  • FIG. 11 is an exemplary diagram of a method for generating a V-shaped near-field lateral background cost potential field provided by an embodiment of the present application
  • FIG. 12 is an example diagram of generation of a background virtual line provided by an embodiment of the present application.
  • FIG. 13 is an example diagram of a near-field lateral background cost potential field monotonicity processing principle provided by an embodiment of the present application.
  • FIG. 14 is an example diagram of a vehicle body multi-circle collision model provided by an embodiment of the present application.
  • 15 is a schematic diagram of a precise obstacle distance calculation provided by an embodiment of the present application.
  • 16 is an example diagram of a horizontal partition provided by an embodiment of the present application.
  • 17 is an example diagram of an arbitration method for outputting an obstacle avoidance path provided by an embodiment of the present application.
  • FIG. 18 is an example diagram of a scenario in which a historical path is used when an alternative path is fully collided according to an embodiment of the present application
  • FIG. 19 is an example diagram of a scenario of switching to the current optimal path when an alternative path is fully collided, provided by an embodiment of the present application.
  • FIG. 20 is an example diagram of a time-domain smoothing provided by an embodiment of the present application.
  • 21 is an example diagram of an obstacle avoidance path generation result in a complex and narrow scene provided by an embodiment of the present application.
  • 22 is an example diagram of a simple urban road scene provided by an embodiment of the present application.
  • FIG. 23 is an example diagram of another alternative path generation scenario provided by an embodiment of the present application.
  • FIG. 24 is an example diagram of an obstacle avoidance path generation result in a simple urban road scene provided by an embodiment of the present application.
  • 25 is an example diagram of a device for planning an obstacle avoidance path of a traveling device according to an embodiment of the present application.
  • FIG. 26 is an exemplary block diagram of the hardware structure of an apparatus for planning an obstacle avoidance path provided by an embodiment of the present application.
  • Path planning In the process of autonomous driving, the automatic driving device outputs a correct and effective driving path according to the upper-layer navigation information, decision-making instructions and environmental perception results, which is used as the main input of the lower-layer controller to guide the correct driving of the vehicle, and completes lane keeping, lane changing, Obstacle avoidance and other actions.
  • the process of generating this path is called path planning. It should be understood that, in the embodiments of the present application, for the convenience of description, the self-vehicle is used to represent the automatic driving device.
  • Occupation map It is composed of grids, also known as the occupancy grid map.
  • the occupancy map contains obstacle information in a given area near the vehicle.
  • the grid values can be expressed as occupied, unoccupied, unknown and other states.
  • Reference path the reference datum used by the ego vehicle for path planning, such as the centerline of the current road.
  • the path mainly reflects the real map information, which can guide the driving direction of the vehicle, but has no obstacle avoidance function.
  • Obstacle avoidance range The driving range of the self-vehicle around the reference path that can be used for obstacle avoidance, providing spatial constraints for obstacle avoidance path planning.
  • Avoidance path On the basis of the reference path and avoidable obstacle range, the self-vehicle performs path planning according to the obstacle information, and obtains a path that can avoid obstacles.
  • FIG. 1 is an example diagram of a Frenet coordinate system provided by the embodiment of the present application. As shown in FIG. 1 , the midpoint of the rear axle of the vehicle is used as the origin of the SL system, and the direction along the reference path is called The longitudinal direction (S direction), the direction perpendicular to the path is called the lateral direction (L direction). The reference path is the centerline of the road.
  • S direction The longitudinal direction
  • L direction the direction perpendicular to the path
  • the reference path is the centerline of the road.
  • Autonomous driving technology uses maps and positioning modules to obtain navigation information, uses sensing devices to perceive the surrounding environment, and combines decision-making, planning, and control modules to achieve autonomous vehicle navigation.
  • autonomous driving technology has made great progress and has been widely used in military and civilian fields.
  • the planning module is responsible for outputting the correct and effective driving path according to the navigation information, decision-making instructions and environmental perception results, so as to realize the vehicle's lane keeping, lane changing, obstacle avoidance and other actions.
  • the path planning module there is usually a reference path, which can be generated based on the current road centerline, or obtained through manual collection, mainly to provide a baseline for path planning for autonomous vehicles.
  • a reference path which can be generated based on the current road centerline, or obtained through manual collection, mainly to provide a baseline for path planning for autonomous vehicles.
  • the reference path On the basis of the reference path, combined with the obstacle information, it is necessary to further realize the obstacle avoidance path planning of the vehicle and realize the safe driving of the vehicle in the dynamic environment, which is also the focus of the path planning.
  • the commonly used obstacle avoidance path planning methods are: sampling, optimization, search, potential field and feature points, etc.
  • the main characteristics of each method are as follows:
  • the sampling-based obstacle avoidance method has high computational efficiency and is convenient for engineering transformation, but usually cannot obtain the most Optimal solution, jitter may occur in complex scenes;
  • the optimization-based method can fully consider the input information, and can obtain an approximate theoretical optimal solution, but it consumes computing resources and is easily affected by the environment, such as obstacles in the distance from the vehicle The change of the distance will directly affect the near path of the vehicle, thus causing the vehicle to vibrate or oscillate when driving in a dynamic environment;
  • the search-based method can deal with complex scenes, but it will consume a lot of computing resources, and it is easy to produce uneven Path planning is usually only suitable for low-speed scenarios, and the support for dynamic scenarios is not good enough;
  • the method based on potential field can comprehensively consider the obstacle information in the environment, but it will cause local bending and vibration;
  • the existing obstacle avoidance path planning methods usually have good planning effects in simple scenarios, but for real dynamic traffic scenarios, especially in complex and narrow scenarios, path shaking is easy to occur, and the safety and stability of vehicles cannot be guaranteed. drive.
  • the embodiment of the present application constructs a first lateral background cost potential field in the near field range, and the background costs on both lateral sides of the target minimum point of the first lateral background cost potential field are monotonic . It effectively reflects the lateral collision trend between the self-vehicle and the obstacle, and the obstacle avoidance path planned based on this can make the driving device maintain the best lateral distance from the obstacle in the complex and narrow environment, and avoid the planned obstacle avoidance. The path is affected by changes in the distant environment, which can improve the stability of the obstacle avoidance path.
  • FIG. 2 is an example diagram of a complex and narrow urban road scene provided by an embodiment of the present application.
  • FIG. 2 when the self-vehicle is driving on the road, there are a large number of dynamic obstacles (number 1-5) and/or static obstacles (number 6) in the avoidable range, forming a narrow and crowded traffic scene.
  • static obstacles can be vehicles parked on both sides of the road or other obstacles, or other obstacles temporarily placed on the road; dynamic obstacles can be other vehicles running within the obstacle avoidance range of the own vehicle.
  • the space for the vehicle to travel is narrow and the travel path is complicated.
  • the reference path and lateral obstacle avoidance range, as well as the occupancy map of all obstacle sizes, positions, velocities and predicted trajectory projections are known.
  • FIG. 3 is an example diagram of a system architecture for obstacle avoidance path planning provided by an embodiment of the present application.
  • the system architecture 300 includes a reference path planner 310 , an obstacle decision maker 320 , an obstacle avoidance path planner 330 and a speed planner 340 .
  • the reference path planner 310 is used to provide the reference path and the lateral obstacle avoidance range to the obstacle avoidance path planner 330 .
  • the reference path refers to the center line of the road. It should be understood that the reference path may also have other representations, which are not limited herein.
  • the lateral obstacle avoidance range refers to the range within which the vehicle can avoid obstacles in the direction perpendicular to the path during the driving process.
  • the lateral obstacle avoidance range can be the width of the current road.
  • the obstacle decider 320 is used to provide the obstacle avoidance path planner 330 with an occupancy map including the obstacle and the projection of the predicted trajectory of the obstacle. It should be understood that the present application does not limit the manner in which the obstacle decision maker 320 obtains the obstacle and the occupancy map projected by the predicted trajectory of the obstacle.
  • the obstacle avoidance path planner 330 is used to plan the obstacle avoidance path according to the reference path provided by the reference path planner 310 and the obstacle decision maker 320 , the lateral obstacle avoidance range, and the occupancy map of the obstacle and the projected obstacle prediction trajectory.
  • the obstacle avoidance path planner 330 includes an alternative path generation module 331 and an alternative path evaluation module 332 .
  • the alternative path generation module 331 is used to generate a cluster of alternative paths according to the reference path and the lateral obstacle avoidance range. The generation method of the alternative paths will be described in detail below, and will not be repeated here; Based on the evaluation of the alternative path according to the alternative path itself and the occupancy information of obstacles, the optimal path of the current frame is selected as the current obstacle avoidance path.
  • the optimal path of the current frame selected by the alternative path evaluation module 332 may be used as the currently output obstacle avoidance path.
  • the obstacle avoidance path planner 330 may further include an output path arbitration module 433 for determining the final output obstacle avoidance path according to the optimal path of the current frame and the historical (previous frame) path.
  • the obstacle avoidance path planner 330 may further include a path time-domain smoothing module 334 for performing time-domain smoothing on the final output obstacle avoidance path.
  • the obstacle avoidance path planner 330 outputs the planned obstacle avoidance path to the back-end speed planner 340 . If there is a possibility of collision between the vehicle and the obstacle on the obstacle avoidance path, the speed planner will decelerate according to the relative position and speed of the collision obstacle and the vehicle.
  • FIG. 4 is an example diagram of a method for planning an obstacle avoidance path of a traveling device according to an embodiment of the present application. It should be understood that the method 400 can be applied in the above-mentioned scenario 200, and can also be applied in the above-mentioned system architecture 300. As shown in FIG. 4, the method 400 includes steps S410-S430. These steps are described in detail below.
  • the distance map includes a plurality of first grids, and each first grid in the plurality of first grids is used to record the distance of the first grid from the first grid that is most recently occupied by the obstacle.
  • the method 400 further includes: acquiring an occupancy map, where the occupancy map includes a plurality of second grids, and each second grid in the plurality of second grids is used to record the second grid The probability that there is an obstacle in the grid. And according to the occupation map, get the distance map.
  • the probability of existing obstacles in the second grid is greater than or equal to a certain threshold, it can be considered that there are obstacles in the second grid; when the probability of existing obstacles in the second grid is less than a certain threshold, It can be considered that there are no obstacles in the second grid.
  • each second grid in the plurality of second grids may also be directly used to record whether there is an obstacle in the second grid.
  • the value recorded in the second grid is 1, it indicates that there is an obstacle; when the value recorded in the second grid is 0, it indicates that there is no obstacle.
  • the occupancy map is provided by an upstream module, such as the obstacle decider 320 described above.
  • the occupancy map contains information about obstacles in a given area near the ego vehicle.
  • the obstacle information refers to the information of the obstacle or the predicted trajectory of the obstacle
  • the grid occupied by the obstacle refers to the grid occupied by the obstacle or the projection of the predicted trajectory of the obstacle.
  • the occupation map needs to be converted into a distance map.
  • the conversion method will be described in detail in the following specific implementation manner, and will not be repeated here.
  • the obstacle avoidance path planning is based on the occupancy map information provided by the upstream module, and the obstacle information provided by the upstream module not only includes the actual pose of the dynamic and static obstacles and the predicted trajectory projection of the obstacles, but also may include the self-vehicle.
  • the occupancy situation of obstacles caused by the false detection of the front part of the vehicle, or the obstacle information appears in the area on the occupied map due to the uneven ground. At this time, if the obstacles in this area are also considered, it will affect the path planning, so that the obstacles in this area must be avoided during the driving process, which will cause the path to oscillate greatly.
  • the second grid located in the first area in the occupied map may be filtered out.
  • the first area is located in front of the traveling device, and if there is an obstacle in the first area, the traveling device cannot avoid the obstacle in the first area; and then the distance map is obtained according to the filtered occupancy map.
  • the erroneous avoidance of the traveling device due to the erroneous detection of the obstacle is effectively avoided. It is ensured that the remaining occupied grids only contain the necessary information required by the obstacle avoidance algorithm, which improves the stability of the obstacle avoidance path.
  • the grid screening method will be described in detail in the following specific implementation manner, and will not be repeated here.
  • S420 Determine a first lateral background cost potential field within a near field range according to the distance map.
  • the near field range is the range in which the longitudinal distance from the traveling device is less than or equal to the first threshold value
  • the first lateral background cost potential field is the first repulsive force field formed by the obstacles in the lateral position to the traveling device
  • the first lateral The background cost potential field includes the target minimum point.
  • the background cost on the lateral sides of the target minimum point is monotonic, and the background cost represents the near-field lateral collision to the traveling device caused by the obstacles at the lateral position within the near-field range. cost.
  • the above-mentioned driving device may be an autonomous driving vehicle, a robot, or other autonomous driving device, which is not limited in this application.
  • the present application is represented by a self-vehicle.
  • the near field range may be a range extending 5m longitudinally along the path direction from the origin of the Frenet coordinate system.
  • the origin may be the center of the rear axle of the vehicle.
  • the length of the ego vehicle may be 3-6m, so the obstacles in the lateral position within the near field range can be understood as obstacles on the lateral sides of the ego car.
  • the near-field range may be determined according to the length of the vehicle, the driving speed, and the complexity of the scene, which is not limited in this application.
  • the first lateral background cost potential field is the first repulsive force field formed by the obstacle in the lateral position to the ego vehicle.
  • the repulsive force field is formed by the obstacles in the lateral position to the traveling device, which can represent the collision tendency of the obstacle to the traveling device, and guide the traveling device to avoid the obstacle, which can also be called the collision field or the risk field.
  • the first repulsive force field can represent the collision tendency of the traveling device formed by the obstacle after the monotonicity treatment.
  • the first lateral background cost potential field can indicate the collision tendency between the ego vehicle and the obstacle in the lateral position.
  • the larger the background cost the smaller the distance between the ego vehicle and the obstacle, and the easier it is to collide; the smaller the background cost, the greater the distance between the ego car and the obstacle.
  • the background cost on the lateral sides of the target minimum point is monotonic, specifically, the potential field on the left side of the target minimum point decreases monotonically, and the potential field on the right side of the target minimum point increases monotonically. It means that the background cost on both sides of the target minimum point is larger than the background cost at the target minimum point. That is to say, the potential field at the lateral position where the target minimum point is located is the lowest, and the ego vehicle can travel stably along the path where the lateral position is located.
  • the first lateral background cost potential field may be in a "V" shape, a "/" shape, or a " ⁇ " shape.
  • the slope of the monotonic function may be a constant value, or may be a value that changes continuously with the lateral position, which is not limited in this application.
  • a V-shaped first lateral background cost potential field is used.
  • determining the current obstacle avoidance path according to the first lateral background cost potential field may be determining the current obstacle avoidance path according to the lateral position of the target minimum point in the first lateral background cost potential field. That is to say, the current obstacle avoidance path is determined according to the lateral position with the lowest potential field, so that the ego vehicle can drive stably along the position with the lowest potential field.
  • the horizontal background cost potential field in the near field is considered when planning the current obstacle avoidance path.
  • the change of the distant obstacle will cause the path near the vehicle to change, thus making the driving unstable;
  • the first horizontal background cost potential field in the near field which effectively reflects the lateral collision trend between the vehicle and the obstacle.
  • the obstacle avoidance path planned based on this can make the vehicle maintain the best lateral distance from the obstacle in a complex and narrow environment, thereby improving the stability of the obstacle avoidance path. .
  • a second lateral background cost potential field within the near field range may be determined first according to the distance map, where the second lateral background cost potential field is a second repulsive force field formed by the obstacle in the lateral position to the traveling device, the The second horizontal background cost potential field includes at least one minimum value point; the target minimum value point is determined from the at least one minimum value point; the background cost on both sides of the target minimum value point is monotonically processed to generate the first Horizontal background cost potential field, the slope of the monotonicity process is greater than or equal to the second threshold. For example, any value whose slope is greater than or equal to 10/1m can be taken, and the specific value needs to be determined in combination with the actual situation, which is not limited here.
  • the second repulsive force field can represent the collision tendency of the obstacle to the traveling device before the monotonicity treatment. That is, monotonic processing of the second repulsion field results in the first repulsion field.
  • the construction method of the first lateral background cost potential field may be: firstly constructing the second lateral background cost potential field within the near field range according to the distance map, and then constructing at least the second lateral background cost potential field in the second lateral background cost potential field.
  • a target minimum point is determined from a minimum value point, and the background cost on the lateral sides of the target minimum value point is monotonically processed to generate the first horizontal background cost potential field.
  • the slope of the monotonicity process to be greater than or equal to the second threshold, the background cost potential field at the lateral position where the target minimum point is located is significantly lower than the background cost potential field at other lateral positions, so that the self-vehicle can be guaranteed. Drive steadily on the path with the lowest potential field.
  • the target minimum value point may be the smallest minimum value point among all the minimum value points in which there is no obstacle in the near field range between the target minimum value point and the traveling device.
  • the minimum minimum point among these minimum points can be used as the target minimum point; one of them can also be selected as the target minimum point according to the actual situation, for example, the minimum minimum point and the second minimum point
  • the background cost of the minimum point is very close, and the lateral position where the second minimum minimum point is located is closer to the ego vehicle.
  • the second minimum minimum value point can be used as the target minimum value point.
  • the minimum minimum point among all minimum points with no obstacles in the near field range from the lateral position where the traveling device is located is used as the target minimum. value points.
  • the target minimum value point can be the minimum minimum value point among all the minimum value points where there is no obstacle in the near-field range between the target minimum value point and the lateral position where the traveling device is located, the running process of the ego vehicle can be enabled. Keep the best safe distance from obstacles.
  • the method 400 may further include: generating a feature vector, the feature vector including a path start point and a sampling target point; according to the reference path, The historical path and the feature vector generate multiple alternative paths.
  • multiple alternative paths may be generated according to the position of the self-vehicle, and then the multiple alternative paths may be evaluated to select the optimal alternative path as the current avoidance path. obstacle path.
  • the generation method of the alternative path will be described in detail in the following embodiments, and will not be repeated here.
  • the reference path can be laterally offset to generate a plurality of virtual parallel line paths; according to the distance map, the distance of each virtual parallel line path in the near field range of the plurality of virtual parallel line paths is calculated.
  • the distance between each waypoint and the nearest obstacle to the waypoint; calculate the background cost of the virtual parallel line according to the minimum distance on the path of each virtual parallel line; according to the background of multiple virtual parallel line paths in the near field The cost generates a second lateral background cost potential field in the near field.
  • one of the collision cost, the constraint cost, the switching cost, the lateral offset cost, or the smoothness cost may also be combined. More factors can be integrated into the determined obstacle avoidance path, so that the stability and safety of the obstacle avoidance path can be further improved.
  • the collision cost and constraints on each of the multiple candidate paths may also be calculated according to the distance map.
  • the collision cost is used to evaluate whether the driving device will collide when driving on the alternative path
  • the constraint cost is used to evaluate whether each waypoint on the alternative path is within the obstacle avoidance range
  • the switching cost is used to evaluate the alternative path
  • the deviation from the historical path the lateral offset cost is used to evaluate the degree of lateral deviation of the candidate path from the road centerline
  • the smoothness cost is used to evaluate the smoothness of the candidate path.
  • the background cost in the first lateral background cost potential field or the background cost in the first lateral background cost potential field and the collision cost, constraint cost, switching cost, and lateral offset cost can be used.
  • the smoothness costs calculate the final cost on each alternative path; then determine the current obstacle avoidance path according to the final cost.
  • the final cost may be calculated according to the weighted summation of various costs. It should be understood that in order to enable the ego vehicle to maintain a safe distance from the obstacle, the weights of the background cost and the collision cost can be made larger. It should be understood that the acquisition of the above various costs and the specific evaluation methods will be described in detail in the following specific implementation manner, and will not be repeated here.
  • the method 400 further includes: determining a final obstacle avoidance path according to the arbitration result of the historical path and the current obstacle avoidance path.
  • the historical path refers to the path traveled by the vehicle in the previous frame
  • the current obstacle avoidance path refers to the currently planned path.
  • the final obstacle avoidance path is determined, that is, a path is selected from the historical path and the current obstacle avoidance path as the final output obstacle avoidance path.
  • the final obstacle avoidance path is determined according to the difference between the current obstacle avoidance path and the historical path. It avoids the situation that the obstacle avoidance path obtained purely based on the cost function or pure logic will produce inter-frame jumping when running in a complex and narrow environment, effectively reducing the planning method's dependence on the cost function, thereby ensuring the stability of the path.
  • the current obstacle avoidance path is significantly better than the historical path, switch to the current obstacle avoidance path; if the current obstacle avoidance path is not significantly better than the historical path, the historical path is used. It avoids the situation that the obstacle avoidance path obtained purely based on the cost function or pure logic will generate inter-frame jumping when running in a complex and narrow environment, thereby ensuring the stability of the path.
  • the specific arbitration conditions will be described in detail in the following specific implementation manner, and will not be repeated here.
  • the method 400 may further include: performing temporal smoothing according to the historical path and the final obstacle avoidance path.
  • time domain smoothing may also be performed according to the historical path and the final obstacle avoidance path.
  • the obstacle avoidance path after time domain smoothing is still the historical path; if the final obstacle avoidance path is the current obstacle avoidance path, the time domain smoothing is performed according to the current obstacle avoidance path and the historical path. .
  • the method 400 may further include: outputting the final obstacle avoidance path to a speed planner, where the speed planner is configured to perform speed planning according to the final obstacle avoidance path and information on obstacles on the final obstacle avoidance path.
  • the speed planner after outputting the final obstacle avoidance path, the speed planner will perform speed planning according to the obstacle information on the final obstacle avoidance path, and will perform deceleration planning if an obstacle that may collide is found on the path. . Thereby improving the security of the path.
  • FIG. 5 is an example diagram of a core flow of an obstacle avoidance path planning method provided by an embodiment of the present application. It should be understood that the planning method shown in FIG. 5 can be applied to the complex and narrow scene shown in FIG. 2 .
  • the core process 500 includes steps S510 , S520 , S530 and S540 . These steps are described in detail below.
  • FIG. 6 is an example diagram of an alternative path generation scenario provided by an embodiment of the present application, and these steps are described in detail below with reference to FIG. 6 .
  • a feature vector including the path start point (start point) and the sampling target point (goal point, referred to as the sampling point) is generated, as shown in the figure Arrows shown in 6.
  • the start point can be found in the historical path.
  • the path passed by the center of the rear axle of the ego vehicle in the previous frame is used as the historical path, so the waypoint closest to the center point of the rear axle of the ego vehicle can be found in the historical path, then the actual position of the point and the The orientation can form the start point feature vector.
  • each path includes many waypoints, and the distance between the waypoints can be determined according to the actual situation. For example, one waypoint can be set every 0.1m, or one waypoint can be set every 0.1*v ego vehicle , where v ego vehicle is the running speed of the ego vehicle.
  • the goal point is within the avoidable range in front of the vehicle body, and is obtained by sampling based on the reference path. It should also be understood that an alternative path may contain multiple goal points, but only one start point. Both the start point and the goal point are represented by vectors.
  • a multi-segment candidate path is generated by connecting the reference path, the historical path and the feature vector, as shown in FIG. 6 .
  • the first paragraph after the start point, obtained by intercepting the historical path.
  • the second paragraph use a smooth curve to connect the start point and the goal point, so that the curve satisfies the vector constraint to generate a path.
  • the third paragraph translate the reference path to after the goal corresponding to the second paragraph.
  • the three paths are connected in sequence, and the entire alternative path is obtained by splicing.
  • a smooth control point can be generated every several waypoints, and a smooth candidate path can be generated based on the control point sequence and the third-order B-spline.
  • one smooth control point may be generated based on three waypoints, or may be based on four, five, six, etc., which is not limited in this application.
  • multiple alternative paths may be generated in the manner shown in FIG. 7 .
  • the reference path is the centerline of the current lane
  • the width of the lane and the obstacle avoidance range is 3.5m
  • the widths of the left and right sides of the lane centerline are 1.75m each.
  • the closest waypoint to the center point of the rear axle of the vehicle can be found in the historical path, and the feature vector of the start point is formed by the actual position and orientation of this point.
  • max(vehicle speed, 4.17m/s) refers to taking the larger value between the current vehicle speed and 4.17m/s; 1.2*max(vehicle speed, 4.17m/s) means 1.2 times the current vehicle speed and 4.17m/s A larger value gets the s-coordinate of the goal point in the first row. It should be understood that the above data is only used as an example, and cannot be used as a limitation on the present application.
  • History segment After the start point, it is obtained by intercepting the history path.
  • Polynomial segment Generate a path using vector constraints of start points and multiple rows of goal points. The specific method is to use any goal point combined with the vector constraint of the start point, calculate the cubic polynomial equation under the Frenet system, and discretize it into waypoints to obtain the second path.
  • Extension Translate the reference path after the goal point corresponding to the second segment. The three paths are connected in sequence, and the entire alternative path is obtained by splicing. Each path corresponds to a goal point one-to-one, and in this example, a total of 33 alternative paths are generated.
  • the alternative paths need to be evaluated, that is, a suitable alternative path is selected as the current obstacle avoidance path among these alternative paths.
  • the evaluation process of the alternative path is described in detail below.
  • the evaluation of the alternative path mainly includes steps S521, S522 and S523, and these steps will be described in detail below with reference to the accompanying drawings.
  • the obstacle avoidance path planning is based on the occupancy map information provided by the upstream module, and the obstacle information provided by the upstream module not only includes the actual pose of the dynamic and static obstacles and the predicted trajectory projection of the obstacles, but also may include the self-vehicle.
  • the false detection of the front part of the car appears to be occupied by obstacles, or because the ground is uneven, the occupied map has obstacle information in this area. At this time, if the obstacles in this area are also considered, it will affect the path planning, so that the wrongly detected obstacles must be avoided during the driving process, which will greatly change the path planning and cause the path to oscillate.
  • a full collision triangle can be constructed to filter the invalid occupancy grid.
  • FIG. 9 is an example diagram of a full collision triangle provided by an embodiment of the present application.
  • the coverage area of the occupied map can be 50m in front of the vehicle body, 10m behind, and 20m left and right, with a resolution of 0.2m*0.2m (that is, the occupied area of each grid), and the occupied part of the grid in the occupied map is the value of 1, the unoccupied part of the grid value is 0.
  • the full collision triangle area means that in a certain triangle area, no matter which path the ego vehicle chooses, it cannot avoid obstacles in this area.
  • the vertex of the triangle is where the start point of the alternative path is located, its base is equal to the width of the vehicle, and the two vertices of the base of the triangle are located on the outermost alternative path.
  • the occupied grids that fall in the full collision triangle are directly screened out.
  • the occupancy map may be converted into a distance map.
  • FIG. 10 is an example diagram of converting an occupancy map into a distance map provided by an embodiment of the present application.
  • the occupancy map can be converted to a distance map by using openCV's DistanceTransform function.
  • the part with a grid value of 1 indicates that the grid is occupied, that is, there is an obstacle; the part with a grid value of 0 indicates that the grid is not occupied, that is, there is no obstacle .
  • the size of the grid in the distance map converted by the occupancy map is equal to the occupancy map, and the value of each grid in the distance map is the distance from the grid to the nearest occupied grid point.
  • the part of the distance map with a grid value of 0 indicates that the distance from the grid to the obstacle is 0, that is, the grid is occupied; the part of the distance map with a grid value of 0.2 indicates that the grid is the closest
  • the distance of obstacles is 0.2m. It should be understood that the value of each grid in the distance map is the distance from the grid to the nearest occupied grid point, which means that if there are usually multiple obstacles around a grid, but only the grid is recorded in the grid Distance to the nearest obstacle. Alternatively, the distance may be a straight-line distance.
  • the erroneous avoidance of the own vehicle due to the erroneous detection of obstacles is effectively avoided. It is ensured that the remaining occupied grids only contain the necessary information required by the obstacle avoidance algorithm, which improves the stability of the obstacle avoidance path.
  • screening out invalid grids in the planning of the obstacle avoidance path means that the obstacles in this area are not considered in the planning of the obstacle avoidance path. But in the actual scene, if there are real obstacles in this area, the speed planning module will decelerate accordingly to avoid collision.
  • a V-shaped near-field lateral background cost potential field may be constructed based on the distance map information after screening out invalid grids, that is, a first lateral background cost potential field within the aforementioned near-field range is constructed.
  • FIG. 11 is an example diagram of a method for generating a V-shaped near-field lateral background cost potential field provided by an embodiment of the present application.
  • the method, step S522 includes steps S5221, S5222 and S5223, which will be described in detail below.
  • the generated virtual parallel line path is shown in FIG. 12 , that is, the reference path is translated laterally to the lateral position where the sampling target point is located.
  • the virtual parallel line path may also be referred to as a background cost virtual line, and the collision trend of the ego vehicle on each background cost virtual line may be expressed by the background cost of the background cost virtual line. The greater the background cost, the greater the possibility of collision; the smaller the background cost, the less likely the collision.
  • the background cost on each background cost virtual line is calculated.
  • the distances of all the waypoints in the near field range on each background cost virtual line to the nearest obstacle are calculated, and the background cost of the background cost virtual line is calculated based on the minimum distance.
  • the formula for calculating the background cost according to the minimum distance can be as follows:
  • the critical value is It should also be understood that the critical value in the above formula (1) is only an example, and in practical applications, the critical value can also be taken as Any value, as long as it can ensure that the vehicle is at a certain safe distance from the obstacle in the lateral direction.
  • a background cost virtual line has 4 waypoints in the near field, and the distances from the nearest obstacles are 0.6, 0.5, 0.7 and 1.2 respectively, then the minimum value of 0.5 is used to calculate the background cost virtual line. background cost. also because Therefore, the background cost of the background cost virtual line
  • the virtual parallel line path includes a plurality of waypoints, and the distance between the waypoints may be determined according to the actual situation. For example, one waypoint can be set every 0.1m, or one waypoint can be set every 0.1*v ego vehicle , where v ego vehicle is the running speed of the ego vehicle. This application does not limit this.
  • the distance value of the grid where the waypoint is located can be directly obtained, or the distance value of the four grids around it can be calculated.
  • the distance information in the lattice is calculated by the bilinear interpolation method, which is not limited in this application.
  • an initial near-field lateral background cost potential field that is, the second lateral background cost potential field within the above-mentioned near-field range, is generated according to the background costs of the plurality of background cost virtual lines within the near-field range.
  • FIG. 13 is an example diagram of a processing principle of a near-field lateral background cost potential field monotonicity provided by an embodiment of the present application.
  • the initial near-field lateral background cost potential field formed by the background costs of multiple background cost virtual lines usually has multiple minima in the lateral (along the L direction) distribution. It is necessary to first find multiple minimum points in the initial near-field lateral background cost potential field, and record the lateral position and the corresponding background cost value. Exemplarily, as shown in the left graph in FIG. 13 , in the lateral distribution of the background cost, there are 3 minimum points. Then find the lateral position of the current ego vehicle and the position of the three minimum values respectively.
  • the virtual line of the background cost at the lateral position where the maximum background cost is located is in the near-field range.
  • the distance from a certain waypoint to the obstacle is 0, that is, there is an obstacle on the waypoint.
  • a method of customizing the maximum background cost can be adopted, for example, taking the maximum background cost as 10 9 .
  • point 1 has the lowest background cost, which means that if you drive on the virtual line of the background cost, you can maintain the best lateral distance from the obstacle. If you follow the traditional optimization algorithm When selecting a path, the background cost virtual line where the horizontal position of point 1 is located is used as the optimal path.
  • the length of the ego car is usually as high as 3-6m, so in the actual scene, due to the existence of intermediate obstacles, the ego car cannot successfully travel to the path with the lowest background cost. That is, the point with the lowest background cost in the near-field lateral background cost potential field is not necessarily the most suitable point.
  • the embodiment of the present application it is also necessary to determine a target minimum value among the multiple minimum values, and the target minimum value is not within the near field range from the lateral position where the vehicle is located.
  • the minimum value points according to the background cost value from small to large, starting from the smallest minimum value, check in turn whether there is an obstacle between the point and the lateral position of the vehicle, that is, whether there is a maximum background cost (with the obstacle The point at which the object distance is 0). If it exists, the corresponding minimum point is regarded as an illegal point, and the point is discarded; if it does not exist, the background cost on both sides of it is treated as a monotonic function with a fixed step size based on this point, forming a V-shaped near-field lateral collision potential field.
  • each background cost virtual line can be directly calculated from the slope of the corresponding horizontal position.
  • the slope of the function during monotonicity processing should be greater than or equal to a certain threshold, for example, any value with a slope greater than or equal to 10/1m can be taken, and the specific value needs to be determined according to the actual situation, which is not limited here.
  • a certain threshold for example, any value with a slope greater than or equal to 10/1m can be taken, and the specific value needs to be determined according to the actual situation, which is not limited here.
  • the evaluation of the alternative paths can be performed by combining the background cost with the collision cost, the constraint cost, the switching cost, the lateral offset cost, and the smoothness cost.
  • the collision cost is used to evaluate the possibility of the self-vehicle colliding with obstacles when driving on the alternative path. Specifically, the collision cost of each waypoint on each candidate path among all the candidate paths is calculated, and the final cost of the candidate path is obtained by summing the collision costs of all the waypoints on each candidate path. It should be understood that the collision cost of each waypoint is calculated from its distance from the nearest obstacle, and the closer the distance is, the greater the collision cost.
  • formula (2) is a calculation method of collision cost provided by the embodiment of the present application:
  • cost obj collision expense d min.i alternative path for the first point from the i-th road distance to the obstacle; max (d min .i / 0.7,0.2 ) for the collection of d min .i / 0.7 and 0.2
  • the larger value; pt i .s is the ordinate of the i-th waypoint on the alternative path; width is the vehicle width.
  • the critical value is When d min.i is greater than the critical value, it is regarded as no collision, and when d min.i is less than or equal to the critical value, it is regarded as collision. It should also be understood that the critical value in the above formula is only an example, and in practical applications, the critical value can also be taken as Any value, as long as it can ensure that the self-vehicle travels on the path without collision.
  • the collision cost of each waypoint is calculated from its distance from the nearest obstacle, so before calculating the collision cost, it is necessary to calculate the distance of each waypoint from its nearest obstacle.
  • the calculation method of the distance between the waypoint and the nearest obstacle will be described in detail with reference to FIG. 14 and FIG. 15 .
  • FIG. 14 is an example diagram of a vehicle body multi-circle collision model provided by an embodiment of the present application.
  • the vehicle body is equivalent to a multi-circle model, and the center of the circle where the midpoint of the rear axle of the vehicle body multi-circle collision model is located coincides with each waypoint on the alternative path (it should be understood that the driving path of the vehicle is is the path traveled by the center of the rear axle of the vehicle), and the heading of the vehicle is the same as that of the waypoint.
  • the distance between the waypoint on the corresponding candidate path and the obstacle is the minimum distance between the centers of multiple circles in the multi-circle model and the obstacle.
  • the state of the waypoint is regarded as a collision, and the collision cost of this point is defined by the distance.
  • the distance from the waypoint i to the obstacle is described in detail.
  • a certain waypoint i on the alternative path and the four-circle model (C1-C4 ) the center of the rear axle coincides, and the distance d min.i between the waypoint i and the obstacle at this time is the minimum distance between the four circle centers in the four-circle model and the obstacle, that is, the distance between the center of the circle C4 and the obstacle. If the distance is less than a certain threshold, it is considered that the state of the waypoint i is a collision; if it is greater than a certain threshold, it is considered that the state of the waypoint i is no collision.
  • the collision distance of the candidate path is not the same as the distance d min.i between the waypoint and the obstacle.
  • the distance map may be used to calculate the distances between the centers of multiple circles in the multi-circle model and the nearest obstacle.
  • bilinear interpolation can be used to calculate the exact distance between the location of the specific circle center and the nearest obstacle, as shown in Figure 15.
  • the point P may be the center of a certain circle in the multi-circle model, and the points ABCD are respectively the position of the waypoint represented by the grid map in the distance map and the distance value to the nearest obstacle. Therefore, the distance value of point P can be calculated by bilinear interpolation method from the distance information of four points ABCD. After calculating the distances from the centers of the circles to the obstacles through the above methods, the smallest distance is taken as the distance d min.i between the waypoint i and the obstacle.
  • Constraint cost which is used to judge whether each waypoint on the alternative path is within the avoidable obstacle range.
  • the overall constraint cost of each alternative path can be obtained by summing the constraint costs of each waypoint, as shown in formula (3):
  • cost margin constraint expense pt i .s first alternative path for the longitudinal coordinate value i of waypoints; pt i .l i th coordinate values lateral waypoints; the LM is left bound to The lateral coordinate value of the left boundary line of the obstacle avoidance range; RM is the right constraint, that is, the lateral coordinate value of the right boundary line of the obstacle avoidance range. It can be seen that the more waypoints on an alternative path outside the avoidable range of obstacles and the smaller the longitudinal coordinate value of the waypoints outside the avoidable range of obstacles, the greater the constraint cost of the path; When all the waypoints on the selected path are within the avoidable range, the constraint cost is 0. Therefore, driving on an alternative path with a constraint cost of 0 can ensure that the self-vehicle travels within the obstacle avoidance range.
  • the switching cost is used to describe the deviation between the alternative path and the historical path. It is divided into offset switching cost and area switching cost.
  • the offset switching cost describes the change of the lateral path of the alternative path relative to the historical path.
  • the absolute value of the difference between the lateral offset of the goal point of the current candidate path and the lateral offset of the goal point of the historical output path of the previous frame can be taken. As shown in the following formula:
  • cost lchange is the offset switching cost
  • goal old is the lateral offset of the goal point of the historical path, that is, the abscissa of the goal point of the historical path
  • goal now is the lateral offset of the goal point of the current candidate path, that is, the current candidate
  • the abscissa of the goal point of the path is the offset switching cost
  • lateral offset is the degree of offset from the road centerline.
  • the area switching cost describes the degree of variation of the alternative path in a large lateral range. When the path bounces in adjacent areas, the area switching cost is small, and when the path bounces between the left and right areas, the area switching cost is very large. At this time, the area switching cost can be used to evaluate the large change of the path.
  • the obstacle avoidance range may be divided into three areas, as shown in FIG. 16 , the left Z1 , the middle Z2 , and the right Z3 .
  • the three areas can be divided as follows. That is, when the abscissa of the goal point is less than -0.5, the goal point is located in area 1 (Z1); when the abscissa of the goal point is greater than -0.5 and less than 0.5, the goal point is located in area 2 (Z2); When the coordinates are greater than 0.5, the goal point is located in zone 3 (Z3).
  • zone represents the area
  • goal.l is the abscissa value of the goal point.
  • the area switching cost between the candidate path and the historical path can be calculated according to the following formula (6):
  • cost zone is the zone switching cost
  • zone old is the zone where the goal point of the historical path is located
  • zone now is the zone where the goal point of the alternative path is located.
  • each is divided into three regions. It should also be understood that the division of the regions may be of equal length or unequal, which is not limited in this application.
  • the lateral path is divided into more regions, for example, four regions. Either 1 or 2 can be used as the threshold.
  • the area switching cost is 1; when it is less than or equal to 1, the area switching cost is 0.
  • the area switching cost is 1; when it is less than or equal to 2, the area switching cost is 0.
  • the lateral offset cost is used to describe the lateral deviation of the alternative path from the road centerline. Specifically, the cost can be described using the lateral offset of the goal point of the alternative path, as shown in Equation (7).
  • cost offset is the lateral offset cost
  • goal now .l is the lateral offset of the goal point of the current alternative path.
  • the smoothness cost which describes the smoothness of the alternative path, is represented by the curvature of the alternative path. Specifically, the cost can be described by the integral of the curvature square of each waypoint of the alternative path, as shown in formula (8):
  • cost smooth is the cost of smoothness
  • pt i .curve is the curvature at the ith waypoint on the alternative path
  • pt i .s is the longitudinal coordinate value of the ith waypoint on the alternative path. It should be understood that the greater the curvature, the greater the cost.
  • the weighted summation result of the above-mentioned respective costs on each alternative path may be used as the final cost of the alternative path .
  • the candidate path with the smallest final cost is taken as the optimal path of the current frame, and the optimal path of the current frame is the above-mentioned current obstacle avoidance path.
  • cost total w back *cost back +w obj *cost obj +w margin *cost margin + (9)
  • the weight of the background cost and the weight of the collision cost can be made larger.
  • the final cost of the optimal path is much lower than other alternative paths. In the actual driving process of the self-vehicle, if the optimal path is slightly deviated, the final cost will increase significantly, so that the self-vehicle can stably drive on the path with the lowest cost.
  • the path with the lowest cost also has the optimal collision cost and background cost, so that the ego vehicle can slowly evade when the obstacle on the side of the ego vehicle is approaching during the driving process, so as to maintain a safe distance from the obstacle. .
  • the weight of each cost set in this embodiment of the present application may be:
  • the optimal path of the current frame that is, the above-mentioned current obstacle avoidance path
  • the final obstacle avoidance path can be output.
  • the specific arbitration method is shown in Figure 17. It should be understood that the optimal path in the current frame is the above-mentioned current obstacle avoidance path.
  • FIG. 17 is an example diagram of an arbitration method for outputting an obstacle avoidance path provided by an embodiment of the present application.
  • the arbitration method namely step S530, includes steps S531-S536. These steps are described in detail below.
  • step S531 judging whether the historical path currently has a collision. If there is a collision, go to step S532; if there is no collision, go to step S533.
  • step S532 determine whether the current optimal path has a collision. If there is a collision, go to step S534; if there is no collision, go to step S535.
  • step S533 it is judged whether the cost of the current optimal path is obviously better than the historical cost. If it is obviously better than the historical cost, go to step S535; if it is not obviously better than the historical cost, go to step S536.
  • step S534 it is judged whether the collision distance of the current optimal path is obviously better than the historical collision distance. If it is obviously better, go to step S535; if not, go to step S536.
  • the historical path is regenerated in the current frame according to the information of the goal points used to generate the historical path in the previous frame, and the final cost and collision state of the historical path are calculated.
  • the output obstacle avoidance path is determined according to the final cost of the current optimal path and the final cost of the historical path. If the final cost of the current optimal path is significantly better than the final cost of the historical path, the current optimal path is used as the output obstacle avoidance path; otherwise, the historical path is continued. It should be understood that the fact that the final cost of the current optimal path is better than the final cost of the historical path means that the final cost of the current optimal path is smaller than the final cost of the historical path.
  • the optimal path of the current frame is used as the final path of the current output. It should also be understood that if the optimal path of the current frame collides, it means that all paths of the current frame collide. In this case, if the collision distance of the optimal path of the current frame is close to the historical collision distance, as shown in Figure 18, the historical path of the previous frame is used; if the collision distance of the optimal path of the current frame is obviously better than the historical collision distance , as shown in Figure 19, the current optimal path is used as the output final obstacle avoidance path.
  • the collision distance of the optimal path of the current frame is obviously better than the historical collision distance, which means that the difference between the collision distance of the optimal path of the current frame minus the historical collision distance is greater than or equal to a certain threshold, for example, 5m.
  • the final cost of the current optimal path is obviously better than the final cost of the historical path, which means that the ratio of the final cost of the current optimal path to the final cost of the historical path is less than or equal to a certain threshold, such as 60%.
  • the time domain smoothing may also be performed on the obstacle avoidance path.
  • each path corresponds to a unique goal point, so the goal point can be used to smooth the obstacle avoidance path in time domain.
  • the sampling target point after the previous frame smoothing goal old i.e., the sample target points on a history path goal old
  • the final sampling target point goal final current frame is selected (i.e. arbitrated select the final obstacle avoidance path
  • the sampling target point goal final is used to generate a time-domain smoothed sampling target point goal filter using inertial filtering.
  • the final sampling target point selected in the current frame is the sampling target point of the historical path.
  • the results of time-domain smoothing and non-time-domain smoothing are the sampling of the historical path. Target.
  • the final sampling target point selected in the current frame is the current optimal sampling target point, and FIG. 20 shows an example diagram of time domain smoothing in this case. After generating the time-domain smoothed sampling target point goal filter , a three-segment path is regenerated based on the smoothed sampling target point goal filter to obtain the final output path.
  • the specific position of the smoothed target point goal filter can be calculated according to formulas (10) and (11):
  • goal filter .l transverse position of the target sample the smoothed; goal filter .s longitudinal position of the target sample the smoothed; goal old .l an upper lateral position of the target sample the smoothed; goal old .s is the vertical position of the smoothed sampling target point in the previous frame; goal final .l is the horizontal position of the final sampling target point selected in the current frame ; goal final .s is the vertical position of the final sampling target point selected in the current frame position; w l is the weight for smoothing, w s is the smoothing weights in the longitudinal direction in the transverse direction.
  • the weights used for smoothing in the horizontal and vertical directions may be set according to actual conditions.
  • the weight of smoothing in the horizontal and vertical directions can be 1, that is, the position of the smoothed sampling target point goal filter is the same as the position of the smoothed sampling target point goal old in the previous frame, that is, the historical path is used; horizontal and vertical
  • the weight of the smoothing can be set to 0, that is, the position of the smoothed sampling target point goal filter is the same as the position of the final sampling target point goal final selected in the current frame, that is, the sampling target point of the final path selected according to the path arbitration is used as the smoothing target point.
  • the sampling target point goal filter After the sampling target point goal filter .
  • the weight can also select any value from 0 to 1 to perform time-domain smoothing according to the actual road conditions, so that the automatic The car can comprehensively consider the situation of the historical path and the optimal path of the current frame, so as to prevent the frame-to-frame jumping when running directly to the optimal path.
  • a horizontal background cost potential field in the near field is generated.
  • the ego car can slowly avoid and maintain a safe distance from the obstacle.
  • the background cost combined with the path arbitration strategy can ensure the stable obstacle avoidance of the self-vehicle.
  • the weight of the background cost is very high, such as 10 6
  • the background cost will change sharply, so that the final cost will also change significantly, and the optimal path cost obtained in each frame will inevitably be smaller than the historical cost.
  • 60% of the optimal path cost will continuously trigger the path to switch from the historical path to the current optimal path, so that the self-vehicle avoids in time.
  • the collision cost will also be used in the evaluation of the entire alternative path. For example, the existence of the No. 1 obstacle makes the alternative paths near the No. 1 obstacle collide, and the collision cost will be very high. , so according to the weighted summation of various costs such as background cost and collision cost, the final output obstacle avoidance path in this scene will shift to the left as shown in Figure 21, effectively avoiding the obstacles that need to be avoided.
  • the above is a specific example of the application of the embodiments of the present application in complex and narrow traffic scenarios. It should be understood that, in combination with the lateral background cost potential field (ie, the first lateral background cost potential field) and the collision cost in the above-mentioned near-field range, the present application The embodiment can also be applied to fast obstacle avoidance in simple high-speed scenarios.
  • the lateral background cost potential field ie, the first lateral background cost potential field
  • the present application can also be applied to fast obstacle avoidance in simple high-speed scenarios.
  • FIG. 22 is an example diagram of a simple urban road scene provided by an embodiment of the present application.
  • FIG. 22 when the self-vehicle is driving fast on the structured road, there are a small number of dynamic and static obstacles (numbered 1-2) in the avoidable obstacle range. However, these obstacles are crossed on the road in front of the self-vehicle, and the self-vehicle needs to use the small S-curve to quickly avoid obstacles.
  • the reference path and lateral obstacle avoidance range as well as the occupancy map of all obstacles' size, location, speed and predicted trajectory, are known.
  • the complex and narrow scene is recorded as scene one, and the simple high-speed scene is recorded as scene two.
  • the path planning may be performed in combination with the lateral background cost potential field (ie, the first lateral background cost potential field) and the collision cost within the above-mentioned near-field range;
  • the lateral background cost potential field ie, the first lateral background cost potential field
  • the collision cost within the above-mentioned near-field range;
  • One or more of the screening of invalid grids, obtaining the final cost by synthesizing various other costs, path arbitration, and time domain smoothing, etc., are used for path planning.
  • the path planning method in the second scenario is also briefly described below in a preferred manner.
  • FIG. 23 is an example diagram of another alternative path generation scenario provided by an embodiment of the present application. The following describes the generation of alternative paths in scenario two in detail with reference to specific examples.
  • the reference path is the centerline of the current lane
  • the width of the lane and the obstacle avoidance range is 3.5m
  • the widths of the left and right sides of the lane centerline are 1.75m each.
  • the method of generating start points and goal points is basically the same as that in Scenario 1. The main difference is that 11 goal points can be generated in each row in Scenario 1, while in Scenario 2, 3 goal points can be generated in each row. goal point.
  • the path is divided into three sections, including the historical section, the fully connected section and the extended section.
  • the historical segment is behind the starting point, and is obtained by intercepting the historical path.
  • the fully connected segment is generated by the connection between the starting point and multiple rows of sampling points.
  • Each segment of the fully connected segment is linked by a cubic polynomial equation, which needs to satisfy the vector constraint passing through the goal point.
  • Extended segment Translate the reference path after the sampling point corresponding to the fully connected segment. The three paths are connected in sequence, and the entire alternative path is obtained by splicing. In this embodiment, a total of 243 alternative paths are generated.
  • the above is only an example, only to clearly describe the difference between the alternative path generation methods in the scenario 1 and the scenario 2, and not as a limitation of the present application. It should be understood that, the number of target points in each row, the number of rows of target points, and the number of alternative paths generated can be determined according to the actual situation, and are not limited.
  • step S520 for the evaluation of the alternative paths (step S520 ) and the output path arbitration (step S530 ), reference can be made to the description of the specific implementation of the scenario one.
  • step S540 the path time domain smoothing
  • the self-vehicle when the self-vehicle is in a scene where two obstacles are crossed, it has a high enough degree of freedom for lateral avoidance. Therefore, a long enough collision-free driving trajectory can be planned according to the above method, so that the self-vehicle can quickly avoid and pass through the traffic scene. Moreover, in the process of quickly avoiding obstacles, the self-vehicle can maintain a sufficient lateral distance from the obstacles and ensure the stability of the path.
  • this embodiment can generate an obstacle avoidance path as shown in FIG. 24 . Since the fully connected method used in this embodiment generates an alternative path, there are many alternative forms of the path, and a fully expanded solution space can obtain a collision-free trajectory for multiple objects around obstacles. And the path is evaluated based on the collision cost, so that the currently planned obstacle avoidance path (optimal alternative path) can keep a safe distance of 0.7m from obstacles 1 and 2 along the entire path. And in the process of driving, when the self-vehicle passes the side of the obstacle, the driving path will be adjusted based on the background cost, so that the self-vehicle and the obstacle keep a safe distance of 1m, so that the longitudinal range where the obstacles 1 and 2 are located.
  • the output path arbitration strategy will re-plan the historical path and the driving process. Arbitration is carried out on the path based on the previous path, so that the historical path is used, and the path is not switched to complete fast obstacle avoidance.
  • the generated alternative paths have fewer forms (as shown in Figure 7).
  • the driving path with collision will affect the speed planner at the back end, causing the ego car to decelerate, so that the ego car can only drive at low speed in this scene.
  • FIG. 25 is an example diagram of a device for planning an obstacle avoidance path of a traveling device according to an embodiment of the present application.
  • the planning apparatus 2500 includes an acquisition module 2510 and a processing module 2520 .
  • the obtaining module 2510 is used to obtain a distance map, and the distance map includes a plurality of first grids, and each first grid in the plurality of first grids is used to record the distance from the first grid to the nearest obstacle The distance of the first grid occupied by the object; the processing module 2520 is used to determine the first horizontal background cost potential field in the near field range according to the distance map, and the near field range is that the longitudinal distance from the driving device is less than or equal to the first The range of a threshold value, the first lateral background cost potential field is the first repulsive force field formed by the obstacle in the lateral position to the traveling device, and the first lateral background cost potential field includes the target minimum value point, the target minimum value The background cost on the lateral sides of the point is monotonic, and the background cost represents the near-field lateral collision cost to the traveling device caused by obstacles in the lateral position within the near-field range; according to the first lateral background cost potential field, determine the current avoidance cost. obstacle path.
  • determining the first horizontal background cost potential field within the near field range according to the distance map includes: determining a second horizontal background cost potential field within the near field range according to the distance map, where the second horizontal background cost potential field is: A second repulsive force field formed by the obstacles in the lateral position to the traveling device, the second lateral background cost potential field includes at least one minimum value point; the target minimum value point is determined from the at least one minimum value point; The background costs on both lateral sides of the target minimum point are subjected to monotonicity processing to generate the first lateral background cost potential field, and the slope of the monotonic processing is greater than or equal to the second threshold.
  • the acquiring module 2510 may also be used to: acquire an occupation map, where the occupation map includes a plurality of second grids, and each second grid in the plurality of second grids uses for recording the probability of the existence of obstacles in the second grid; the processing module 2520 may also be used to obtain a distance map according to the occupation map.
  • obtaining the distance map according to the occupation map includes: screening out the second grid located in the first area in the occupation map, the first area is located in front of the traveling device, if there is an obstacle in the first area The running device cannot avoid the obstacles in the first area; the distance map is obtained according to the occupied map after screening.
  • the processing module 2520 can also be used to: generate a feature vector, the feature vector including the path starting point and the sampling target point; generating a plurality of candidate paths with reference to the path, the historical path and the feature vector; the determining the second lateral background cost potential field within the near field range according to the distance map includes: according to the sampling target point, the reference path Perform lateral offset to generate multiple virtual parallel line paths; according to the distance map, calculate the distance between each waypoint in the near field range of each virtual parallel line path in the multiple virtual parallel line paths to the nearest obstacle to the waypoint. distance; calculate the background cost of the virtual parallel line according to the minimum value of the distance on the path of each virtual parallel line; generate a The second horizontal background cost potential field.
  • the processing module 2520 is further configured to: calculate the distance on each of the multiple candidate paths according to the distance map.
  • One or more of a collision cost, a constraint cost, a switching cost, a lateral offset cost, or a smoothness cost wherein the collision cost is used to evaluate whether a collision will occur when the traveling device travels on the alternative path, the The constraint cost is used to evaluate whether each waypoint on the candidate path is within the obstacle avoidance range, the switching cost is used to evaluate the deviation between the candidate path and the historical path, and the lateral offset cost is used to evaluate the The degree of lateral deviation of the alternative path from the road centerline, and the smoothness cost is used to evaluate the smoothness of the alternative path; according to the first lateral background cost potential field, determining the current obstacle avoidance path includes: according to the first lateral background The background cost in the cost potential field, or calculated according to the background cost in the first lateral background cost potential field and one or more of the
  • the processing module 2520 is further configured to: determine the final obstacle avoidance path according to the arbitration result of the historical path and the current obstacle avoidance path.
  • determining the final obstacle avoidance path according to the arbitration result of the historical path and the current obstacle avoidance path including: if the historical path has a collision but the current obstacle avoidance path does not have a collision, then use the current obstacle avoidance path as the final obstacle avoidance path; or, if both the historical path and the current obstacle avoidance path collide, and the absolute value of the difference between the collision distance of the current obstacle avoidance path and the collision distance of the historical path is greater than or equal to a third threshold , the current obstacle avoidance path and the historical path with the larger collision distance are used as the final obstacle avoidance path; or, if both the historical path and the current obstacle avoidance path collide, and the collision distance of the current obstacle avoidance path The absolute value of the difference between the collision distance and the historical path is less than the third threshold, then the historical path is used as the final obstacle avoidance path; or, if there is no collision on the historical path, and the final cost of the current obstacle avoidance path is equal to If the ratio of the historical path cost is less than or equal to
  • processing module 2520 can also be used to: perform time domain smoothing according to the historical path and the final obstacle avoidance path.
  • the planning apparatus 2500 may further include an output module for outputting the obstacle avoidance path to a speed planner, where the speed planner is configured to perform speed planning according to the final obstacle avoidance path and information on obstacles on the final obstacle avoidance path.
  • FIG. 26 is an exemplary block diagram of the hardware structure of an apparatus for planning an obstacle avoidance path provided by an embodiment of the present application.
  • the apparatus 2600 (the apparatus 2600 may specifically be a computer device) includes a memory 2610 , a processor 2620 , a communication interface 2630 and a bus 2640 .
  • the memory 2610, the processor 2620, and the communication interface 2630 realize the communication connection among each other through the bus 2640.
  • the memory 2610 may be a read only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM).
  • the memory 2610 may store a program, and when the program stored in the memory 2610 is executed by the processor 2620, the processor 2620 is configured to execute each step of the planning method of the embodiment of the present application.
  • the processor 2620 can be a general-purpose central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), a graphics processor (graphics processing unit, GPU), or one or more
  • the integrated circuit is used to execute the relevant program to realize the planning method of the method embodiment of the present application.
  • the processor 2620 can also be an integrated circuit chip with signal processing capability.
  • each step of the planning method of the present application may be completed by an integrated logic circuit of hardware in the processor 2620 or instructions in the form of software.
  • the above-mentioned processor 2620 may also be a general-purpose processor, a digital signal processor (digital signal processing, DSP), an application specific integrated circuit (ASIC), an off-the-shelf programmable gate array (field programmable gate array, FPGA) or other programmable logic devices, Discrete gate or transistor logic devices, discrete hardware components.
  • DSP digital signal processing
  • ASIC application specific integrated circuit
  • FPGA field programmable gate array
  • a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the steps of the method disclosed in conjunction with the embodiments of the present application may be directly embodied as executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor.
  • the software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art.
  • the storage medium is located in the memory 2610, and the processor 2620 reads the information in the memory 2610, and combines its hardware to complete the functions required by the modules included in the planning apparatus of the embodiments of the present application, or execute the planning methods of the method embodiments of the present application.
  • the communication interface 2630 implements communication between the apparatus 1000 and other devices or a communication network using a transceiving device such as, but not limited to, a transceiver.
  • a transceiving device such as, but not limited to, a transceiver.
  • the functions of the acquisition module 2510 and the output module described above are implemented.
  • the bus 2640 may include pathways for communicating information between the various components of the device 2600 (eg, the memory 2610, the processor 2620, the communication interface 2630).
  • Embodiments of the present application further provide a vehicle, including the above obstacle avoidance path planning device 2500 or 2600, capable of executing any of the above planning methods, so that the vehicle can run smoothly in a complex and narrow scene and be in harmony with vehicles in the lateral direction. Keep a safe distance.
  • the vehicle may be an electric vehicle, for example, a pure electric vehicle, an extended-range electric vehicle, a hybrid electric vehicle, a fuel cell vehicle, a new energy vehicle, etc., which are not specifically limited in this application.
  • the embodiment of the present application also provides a computer system on which any of the above-mentioned planning methods (algorithms) can be run, and the computing power of the computer should ensure that the update rate of any of the above-mentioned planning algorithms reaches a certain value, for example , 30HZ.
  • the path planning can be carried out according to the real-time environmental conditions, so as to ensure the safe and stable operation of the traveling device.
  • the disclosed system, apparatus and method may be implemented in other manners.
  • the apparatus embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented.
  • the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the functions, if implemented in the form of software functional units and sold or used as independent products, may be stored in a computer-readable storage medium.
  • the technical solution of the present application can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution, and the computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other media that can store program codes .

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Mechanical Engineering (AREA)
  • Transportation (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Human Computer Interaction (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Traffic Control Systems (AREA)
  • Control Of Driving Devices And Active Controlling Of Vehicle (AREA)

Abstract

一种行驶装置的避障路径的规划方法和装置，行驶装置可以是自动驾驶汽车、机器人或其他自主行驶装置。规划方法包括：获取距离地图（S410）；根据距离地图确定近场范围内的第一横向背景代价势场（S420），第一横向背景代价势场为横向位置上的障碍物对行驶装置构成的第一斥力场，第一横向背景代价势场包括目标极小值点，目标极小值点横向两侧的背景代价具有单调性，背景代价表示近场范围内横向位置上的障碍物对行驶装置构成的近场横向碰撞代价；根据第一横向背景代价势场，确定当前避障路径（S430）。避障路径的规划方法使得行驶装置在复杂狭窄交通场景中能够安全平稳的避障。

Description

行驶装置的避障路径的规划方法和装置
本申请要求于2020年07月20日提交中国专利局、申请号为202010699919.X、申请名称为“行驶装置的避障路径的规划方法和装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及自动驾驶技术领域，并且更具体地，涉及一种行驶装置的避障路径的规划方法和装置。
背景技术
自动驾驶技术利用地图和定位模块获取导航信息，利用传感设备感知周围环境，结合决策、规划、控制模块实现车辆的自主导航。近年来随着计算机视觉、人工智能、传感设备的快速发展，自动驾驶技术得到了长足的发展，在军用、民用领域都得到了广泛的应用。在自动驾驶装置中，路径规划模块负责根据上层导航信息、决策指令及环境感知结果输出正确有效的行驶路径，引导自动驾驶装置的正确行驶。同时，路径规划模块能够结合障碍物信息进一步实现避障路径的规划，使得自动驾驶装置在动态环境中能够安全行驶，这也是路径规划的工作重点。
现有的避障路径的规划方法主要有：采样、优化、搜索、势场和特征点等，这些避障方法在简单场景中通常具有较好的规划效果，但对于真实的动态交通场景，尤其是在复杂狭窄的场景中，容易产生路径晃动，无法保证车辆安全平稳行驶。
因此，在面对复杂狭窄交通场景时，如何通过统一的规划方法输出安全平稳的避障路径是目前亟需解决的问题。
发明内容
本申请提供一种行驶装置的避障路径的规划方法和装置，使得行驶装置在复杂狭窄交通场景中能够安全平稳的避障。
第一方面，提供了一种行驶装置的避障路径的规划方法，该方法包括：获取距离地图，该距离地图包括多个第一栅格，该多个第一栅格中的每个第一栅格用于记录该第一栅格离最近被障碍物占据的第一栅格的距离；根据该距离地图确定近场范围内的第一横向背景代价势场，该近场范围为与行驶装置的纵向距离小于或等于第一阈值的范围，该第一横向背景代价势场为横向位置上的障碍物对该行驶装置构成的第一斥力场，该第一横向背景代价势场包括目标极小值点，该目标极小值点横向两侧的背景代价具有单调性，该背景代价表示近场范围内横向位置上的障碍物对该行驶装置构成的近场横向碰撞代价；根据该第一横向背景代价势场，确定当前避障路径。
可选地，上述行驶装置可以是自动驾驶车辆，也可以是机器人，又或是其他自主行驶 装置。
应理解，斥力场是由横向位置上的障碍物对行驶装置构成的，能够表示障碍物对行驶装置的碰撞趋势，会引导行驶装置避开障碍物，也可以称为碰撞场或风险场。该第一横向背景代价势场包括目标极小值点，该目标极小值点横向两侧的背景代价具有单调性，即意味着第一斥力场中的目标极小值点横向两侧的背景代价具有单调性。
应理解，该目标极小值点两侧的背景代价具有单调性，具体是在目标极小值点左侧的势场单调递减，在目标极小值点右侧的势场单调递增。
应理解，该目标极小值点横向两侧的背景代价具有单调性，也就意味着目标极小值点横向两侧的背景代价均比目标极小值点处的背景代价大。
可选地，第一横向背景代价势场可以呈V字型。
可选地，根据该第一横向背景代价势场，确定当前避障路径可以是根据第一横向背景代价势场中的目标极小值点所处的横向位置，确定当前避障路径。
在本申请实施例中，首先获取距离地图，然后根据距离地图构建近场范围内的第一横向背景代价势场，且该第一横向背景代价势场包括目标极小值点，该目标极小值点横向两侧的背景代价具有单调性；然后根据该第一横向背景代价势场，确定当前避障路径。使得在避障路径规划时，充分考虑到近场障碍物的信息，避免远处障碍物对规划路径产生影响；并且通过构建近场范围内的第一横向背景代价势场，有效体现行驶装置与障碍物的横向碰撞趋势，以此为基础规划的避障路径可以使行驶装置在复杂狭窄的环境与障碍物保持最佳的横向距离，从而能够提高避障路径的稳定性。
结合第一方面，在第一方面的某些实现方式中，根据该距离地图确定近场范围内的第一横向背景代价势场包括：根据该距离地图确定该近场范围内的第二横向背景代价势场，该第二横向背景代价势场为横向位置上的障碍物对该行驶装置构成的第二斥力场，该第二横向背景代价势场包括至少一个极小值点；从该至少一个极小值点中确定该目标极小值点；对该目标极小值点横向两侧的背景代价进行单调性处理，生成该第一横向背景代价势场，该单调性处理的斜率大于或等于第二阈值。
应理解，第二斥力场能够表示在单调性处理之前的障碍物对行驶装置构成的碰撞趋势。也就是说，对第二斥力场进行单调性处理会得到第一斥力场。
在本申请实施例中，首先根据距离地图构建近场范围内的第二横向背景代价势场，然后在第二横向背景代价势场中的至少一个极小值点中确定一个目标极小值点，并对该目标极小值点横向两侧的背景代价进行单调性处理，生成该第一横向背景代价势场，且该单调性处理的斜率大于或等于第二阈值。从而使得目标极小值点所在的横向位置的背景代价势场明显低于其他横向位置处的背景代价势场，当行驶装置侧面障碍物在横向上稍微靠近时，该行驶装置所处的横向位置的背景代价势场就会显著增大，从而缓慢平稳的避让并运行到背景代价势场最低的横向位置，保证与障碍物保持安全的距离。
结合第一方面，在第一方面的某些实现方式中，该目标极小值点与该行驶装置之间不存在障碍物。
换句话说，也就是该目标极小值点所在的横向位置与该行驶装置所在的横向位置之间在该近场范围内不存在障碍物。
可选地，该目标极小值点可以是与行驶装置之间在该近场范围内不存在障碍物的所有 极小值点中的最小极小值点。
在本申请实施例中，该目标极小值点与该行驶装置之间不存在障碍物，使得自车能够平稳地运行到目标极小值点所处的横向位置，避免因为行驶装置与目标极小值点所处的横向位置之间存在障碍物使得自车无法运行到目标极小值点所处的横向位置，从而能够提高路径的稳定性。
结合第一方面，在第一方面的某些实现方式中，在获取距离地图之前，该方法还包括：获取占据地图，该占据地图包括多个第二栅格，该多个第二栅格中的每个第二栅格用于记录该第二栅格中存在障碍物的概率；获取距离地图包括：根据该占据地图，得到该距离地图。
可选地，当第二栅格中存在障碍物的概率大于或等于一定阈值时，可以认为该第二栅格中存在障碍物；当第二栅格中存在障碍物的概率小于一定阈值时，可以认为该第二栅格中不存在障碍物。
可选地，该多个第二栅格中的每个第二栅格还可以直接用于记录该第二栅格中是否存在障碍物。
结合第一方面，在第一方面的某些实现方式中，根据所述占据地图，得到该距离地图包括：筛除该占据地图中位于第一区域的所述第二栅格，该第一区域位于该行驶装置的前方，若该第一区域内存在障碍物，该行驶装置无法避开该第一区域内的障碍物；根据筛除后的占据地图，得到该距离地图。
应理解，若该第一区域内存在障碍物，该行驶装置无法避开该第一区域内的障碍物，也可以描述为，若该第一区域内存在障碍物，该行驶装置无论选择哪一条避障路径都无法避开该第一区域内的障碍物。
由于在车前某一特定区域范围内，该行驶装置无论选择哪一条避障路径都无法避开该区域内的障碍物。又由于避障路径规划是根据上游模块提供的占据地图信息进行的，而上游模块提供的障碍物信息除了包括动静态障碍物的实际位姿和障碍物的预测轨迹投影，还有可能包括自车车头部位的误检导致的障碍物占据情况，或者是由于地面不平坦使得占据地图在该区域出现了障碍物信息。此时，如果还考虑这一区域的障碍物，就会影响路径规划，使得在行驶过程中总要避开这一区域的障碍物，从而引起路径的大幅震荡。因此，在本申请实施例中，在获取距离地图之前，可以先将占据地图中位于该区域的栅格筛除。从而有效避免了行驶装置因为障碍物的误检而进行的误避让。保证剩余的占据栅格只包含避障算法所需的必要信息，提高了避障路径的平稳性。
结合第一方面，在第一方面的某些实现方式中，在根据该距离地图确定该近场范围内的第二横向背景代价势场之前，该方法还包括：生成特征矢量，该特征矢量包括路径起始点和采样目标点；根据参考路径、历史路径和所述特征矢量，生成多条备选路径；根据该距离地图确定该近场范围内的第二横向背景代价势场包括：根据该采样目标点，对该参考路径进行横向偏移生成多条虚拟平行线路径；根据该距离地图计算该多条虚拟平行线路径中每条虚拟平行线路径在该近场范围内的每个路点离该路点最近障碍物的距离；根据该每条虚拟平行线路径上该距离的最小值计算该条虚拟平行线的背景代价；根据该多条虚拟平行线路径在该近场范围内的背景代价生成该近场范围内的第二横向背景代价势场。
结合第一方面，在第一方面的某些实现方式中，在根据该第一横向背景代价势场，确 定当前避障路径之前，该方法还包括：根据该距离地图计算该多条备选路径中的每条备选路径上的碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，其中，该碰撞代价用于评价该行驶装置在该备选路径上行驶时是否会发生碰撞，该约束代价用于评价该备选路径上的各个路点是否在可避障范围内，该切换代价用于评价该备选路径与该历史路径之间的偏差，该横向偏移代价用于评价该备选路径的横向偏离道路中心线的程度，该平滑度代价用于评价该备选路径的平滑程度；根据所述第一横向背景代价势场，确定当前避障路径包括：根据该第一横向背景代价势场中的背景代价，或者根据该第一横向背景代价势场中的背景代价以及该碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，计算每条备选路径上的最终代价；根据该最终代价，确定该当前避障路径。
在本申请实施例中，在确定当前避障路径时，除了考虑背景代价，还可以考虑碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，使得所确定的避障路径能够综合更多的因素，从而能够进一步提高避障路径的平稳性和安全性。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：根据历史路径和该当前避障路径的仲裁结果，确定最终避障路径。
在本申请实施例中，在确定当前避障路径时，根据当前避障路径和历史路径的差异，确定最终避障路径。避免了在复杂狭窄环境中运行时纯粹基于代价函数或纯逻辑得到的避障路径会产生帧间跳动的情况，有效降低了规划方法对代价函数的依赖，从而保证路径的稳定。
结合第一方面，在第一方面的某些实现方式中，根据历史路径和该当前避障路径的仲裁结果，确定最终避障路径包括：若该历史路径存在碰撞，但该当前避障路径不存在碰撞，则以该当前避障路径作为该最终避障路径；或者，若该历史路径和该当前避障路径都存在碰撞，且该当前避障路径的碰撞距离与该历史路径的碰撞距离之差的绝对值大于或等于第三阈值，则以该当前避障路径和该历史路径中碰撞距离较大者作为该最终避障路径；或者，若该历史路径和该当前避障路径都存在碰撞，且该当前避障路径的碰撞距离与该历史路径的碰撞距离之差的绝对值小于第三阈值，则以该历史路径作为该最终避障路径；或者，若该历史路径不存在碰撞，且该当前避障路径的该最终代价与该历史路径代价之比小于或等于第四阈值，则以该当前避障路径作为该最终避障路径；或者，若该历史路径不存在碰撞，且该当前避障路径的该最终代价与该历史路径代价之比大于第四阈值，则以该历史路径作为该最终避障路径；其中，该碰撞距离为每条备选路径上处于发生碰撞状态的所有的路点的纵向距离中的最小纵向距离。
在申请实施例中，如果当前避障路径明显优于历史路径，则切换至当前避障路径；如果当前避障路径没有明显优于历史路径，则沿用历史路径。避免了在复杂狭窄环境中运行时纯粹基于代价函数或纯逻辑得到的避障路径会产生帧间跳动的情况，从而保证路径的稳定。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：根据历史路径和最终避障路径进行时域平滑。
在本申请实施例中，做获取到最终避障路径时，还可以根据历史路径和最终避障路径进行时域平滑。此时，如果最终避障路径为沿用历史路径，时域平滑后的避障路径仍为历 史路径；如果最终避障路径为当前避障路径，则根据当前避障路径和历史路径进行时域平滑。有效避免了在复杂狭窄环境中切换避障路径时产生帧间跳动的情况，从而保证路径的稳定。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：将最终避障路径输出给速度规划器，该速度规划器用于根据最终避障路径与最终避障路径上障碍物信息，进行速度规划。
在本申请实施例中，输出最终避障路径后，速度规划器会根据最终避障路径上的障碍物信息进行速度规划，如果发现该路径上存在可能发生碰撞的障碍物，就会进行减速规划。从而提高路径的安全性。
第二方面，提供了一种行驶装置的避障路径的规划装置，该装置包括：获取模块和处理模块；该获取模块用于获取距离地图，该距离地图包括多个第一栅格，该多个第一栅格中的每个第一栅格用于记录该第一栅格离最近被障碍物占据的第一栅格的距离；该处理模块用于根据该距离地图确定近场范围内的第一横向背景代价势场，该近场范围为与该行驶装置的纵向距离小于或等于第一阈值的范围，该第一横向背景代价势场为横向位置上的障碍物对该行驶装置构成的第一斥力场，第一横向背景代价势场包括目标极小值点，该目标极小值点横向两侧的背景代价具有单调性，该背景代价表示近场范围内横向位置上的障碍物对该行驶装置构成的近场横向碰撞代价；根据该第一横向背景代价势场，确定当前避障路径。
结合第二方面，在第二方面的某些实现方式中，根据该距离地图确定近场范围内的第一横向背景代价势场包括：根据该距离地图确定近场范围内的第二横向背景代价势场，该第二横向背景代价势场为横向位置上的障碍物对该行驶装置构成的第二斥力场，该第二横向背景代价势场包括至少一个极小值点；从该至少一个极小值点中确定目标极小值点；对该目标极小值点横向两侧的背景代价进行单调性处理，生成该第一横向背景代价势场，该单调性处理的斜率大于或等于第二阈值。
结合第二方面，在第二方面的某些实现方式中，该目标极小值点与该行驶装置之间不存在障碍物。
结合第二方面，在第二方面的某些实现方式中，在获取距离地图之前，该获取模块还用于：获取占据地图，该占据地图包括多个第二栅格，该多个第二栅格中的每个第二栅格用于记录该第二栅格中存在障碍物的概率；该处理模块还用于：根据该占据地图，得到距离地图。
结合第二方面，在第二方面的某些实现方式中，该根据所述占据地图，得到距离地图包括：筛除该占据地图中位于第一区域的第二栅格，该第一区域位于行驶装置的前方，若该第一区域内存在障碍物，该行驶装置无法避开该第一区域内的障碍物；根据筛除后的占据地图，得到该距离地图。
结合第二方面，在第二方面的某些实现方式中，在该根据距离地图确定近场范围内的第二横向背景代价势场之前，该处理模块还用于：生成特征矢量，该特征矢量包括路径起始点和采样目标点；根据参考路径、历史路径和所述特征矢量，生成多条备选路径；该根据所述距离地图确定该近场范围内的第二横向背景代价势场包括：根据该采样目标点，对该参考路径进行横向偏移生成多条虚拟平行线路径；根据该距离地图计算该多条虚拟平行 线路径中每条虚拟平行线路径在该近场范围内的每个路点离该路点最近障碍物的距离；根据该每条虚拟平行线路径上该距离的最小值计算该条虚拟平行线的背景代价；根据该多条虚拟平行线路径在该近场范围内的背景代价生成该近场范围内的第二横向背景代价势场。
结合第二方面，在第二方面的某些实现方式中，在该根据所述第一横向背景代价势场，确定当前避障路径之前，该处理模块还用于：根据该距离地图计算该多条备选路径中的每条备选路径上的碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，其中，该碰撞代价用于评价该行驶装置在该备选路径上行驶时是否会发生碰撞，该约束代价用于评价该备选路径上的各个路点是否在可避障范围内，该切换代价用于评价该备选路径与该历史路径之间的偏差，该横向偏移代价用于评价该备选路径的横向偏离道路中心线的程度，该平滑度代价用于评价该备选路径的平滑程度；根据该第一横向背景代价势场，确定当前避障路径包括：根据该第一横向背景代价势场中的背景代价，或者根据该第一横向背景代价势场中的背景代价以及该碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，计算该每条备选路径上的最终代价；根据该最终代价，确定该当前避障路径。
结合第二方面，在第二方面的某些实现方式中，该处理模块还用于：根据历史路径和该当前避障路径的仲裁结果，确定最终避障路径。
结合第二方面，在第二方面的某些实现方式中，该根据历史路径和该当前避障路径的仲裁结果，确定最终避障路径，包括：若该历史路径存在碰撞，但该当前避障路径不存在碰撞，则以该当前避障路径作为该最终避障路径；或者，若该历史路径和该当前避障路径都存在碰撞，且该当前避障路径的碰撞距离与该历史路径的碰撞距离之差的绝对值大于或等于第三阈值，则以该当前避障路径和该历史路径中碰撞距离较大者作为该最终避障路径；或者，若该历史路径和该当前避障路径都存在碰撞，且该当前避障路径的碰撞距离与该历史路径的碰撞距离之差的绝对值小于第三阈值，则以该历史路径作为该最终避障路径；或者，若该历史路径不存在碰撞，且该当前避障路径的该最终代价与该历史路径代价之比小于或等于第四阈值，则以该当前避障路径作为该最终避障路径；或者，若该历史路径不存在碰撞，且该当前避障路径的该最终代价与该历史路径代价之比大于第四阈值，则以该历史路径作为该最终避障路径；其中，该碰撞距离为该每条备选路径上处于发生碰撞状态的所有的路点的纵向距离中的最小纵向距离。
结合第二方面，在第二方面的某些实现方式中，该处理模块还用于：根据历史路径和最终避障路径进行时域平滑。
结合第二方面，在第二方面的某些实现方式中，该装置还可以包括输出模块，用于将最终避障路径输出给速度规划器，该速度规划器用于根据最终避障路径与最终避障路径上障碍物信息，进行速度规划。
第三方面，提供了一种规划装置，包括输入输出接口、处理器和存储器，该处理器用于控制输入输出接口收发信号或信息，该存储器用于存储计算机程序，该处理器用于从存储器中调用并运行该计算机程序，使得该规划装置执行上述第一方面中的方法。
第四方面，提供了一种车辆，包括用于执行上述第一方面所述方法的各个模块。
第五方面，提供了一种计算机可读介质，其特征在于，所述计算机可读介质存储有程序代码，当所述计算机程序代码在计算机上运行时，使得计算机执行上述第一方面中的方 法。
第六方面，提供了一种芯片，所述芯片包括处理器与数据接口，所述处理器通过所述数据接口读取存储器上存储的指令，执行上述第一方面中的方法。
可选地，作为一种实现方式，所述芯片还可以包括存储器，所述存储器中存储有指令，所述处理器用于执行所述存储器上存储的指令，当所述指令被执行时，所述处理器用于执行上述第一方面中的方法。
上述芯片具体可以是现场可编程门阵列FPGA或者专用集成电路ASIC。
应理解，本申请中，第一方面的方法具体可以是指第一方面以及第一方面中各种实现方式中的任意一种实现方式中的方法。
附图说明
图1是本申请实施例提供的一种Frenet坐标系的示例图；
图2是本申请实施例提供的一种复杂狭窄的城市道路场景示例图；
图3是本申请实施例提供的一种避障路径规划的系统架构示例图；
图4是本申请实施例提供的一种行驶装置的避障路径的规划方法示例图；
图5是本申请实施例提供的一种避障路径的规划方法的核心流程示例图；
图6是本申请实施例提供的一种备选路径生成的场景示例图；
图7是本申请实施例提供的另一种备选路径生成的场景示例图；
图8是本申请实施例提供的一种无效占据栅格的示例图；
图9是本申请实施例提供的一种全碰撞三角区的示例图；
图10是本申请实施例提供的一种占据地图转换为距离地图的示例图；
图11是本申请实施例提供的一种V型的近场横向背景代价势场的生成方法示例图；
图12是本申请实施例提供的一种背景虚拟线的生成示例图；
图13是本申请实施例提供的一种近场横向背景代价势场单调性处理原理示例图；
图14是本申请实施例提供的一种车体多圆碰撞模型的示例图；
图15是本申请实施例提供的一种精确障碍物距离计算原理图；
图16是本申请实施例提供的一种横向分区示例图；
图17是本申请实施例提供的一种输出避障路径的仲裁方法示例图；
图18是本申请实施例提供的一种备选路径全碰撞时沿用历史路径的场景示例图；
图19是本申请实施例提供的一种备选路径全碰撞时切换至当前最优路径的场景示例图；
图20是本申请实施例提供的一种时域平滑示例图；
图21是本申请实施例提供的一种复杂狭窄场景避障路径生成结果示例图；
图22是本申请实施例提供的一种简单的城市道路场景示例图；
图23是本申请实施例提供的又一种备选路径生成的场景示例图；
图24是本申请实施例提供的一种简单的城市道路场景避障路径生成结果示例图；
图25是本申请实施例提供的一种行驶装置的避障路径的规划装置示例图；
图26是本申请实施例提供的一种避障路径的规划装置的硬件结构示例性框图。
具体实施方式
为了便于理解，首先对本申请各实施例中所涉及的一些技术术语进行介绍。
路径规划：自动驾驶装置在自主行驶过程中，根据上层导航信息、决策指令及环境感知结果输出正确有效的行驶路径，作为下层控制器的主要输入引导车辆的正确行驶，完成车道保持、换道、避障等多种动作。该路径的生成过程称为路径规划。应理解，在本申请实施例中，为便于描述，均用自车来表示自动驾驶装置。
占据地图：由栅格组成的，也称为占据栅格地图，该占据地图中包含自车附近给定区域内的障碍物信息，栅格值可表示为占据、未占据、未知等状态。
参考路径：自车用于路径规划的参考基准，如当前道路的中心线。该路径主要反应真实的地图信息，可以引导自车的行驶方向，但无避障功能。
可避障范围：自车在参考路径周围可以用于避障的行驶范围，为避障路径规划提供空间约束。
避让路径：在参考路径和可避障范围的基础上，自车根据障碍物信息进行路径规划，得到的可以对障碍物进行规避的路径。
Frenet坐标系(SL系)：图1是本申请实施例提供的一种Frenet坐标系的示例图，如图1所示，以自车后轴中点作为SL系原点，沿参考路径方向称为纵向(S方向)，垂直于路径方向称为横向(L方向)。其中，参考路径为道路的中心线。
为便于理解，再对本申请实施例涉及的背景技术进行详细介绍。
自动驾驶技术利用地图和定位模块获取导航信息，利用传感设备感知周围环境，结合决策、规划、控制模块实现车辆的自主导航。近年来随着计算机视觉、人工智能、传感设备的快速发展，自动驾驶技术得到了长足的发展，在军用、民用领域都得到了广泛的应用。
在自动驾驶车辆的各个模块中，规划模块负责根据导航信息、决策指令及环境感知结果输出正确有效的行驶路径，实现车辆的车道保持、换道、避障等动作。在路径规划模块中，通常会具备一条参考路径，它可以基于当前道路中心线生成，或者通过人为采集的方式获得，主要为自动驾驶车辆提供路径规划的基准线。在参考路径的基础上，结合障碍物信息需要进一步实现车辆的避障路径规划，实现在动态环境中车辆的安全行驶，这也是路径规划的工作重点。
目前，常用的避障路径规划方法有：采样、优化、搜索、势场和特征点等，各方法的主要特点如下：基于采样的避障方法计算效率高，便于工程转化，但是通常无法得到最优解，在复杂场景中可能会出现抖动；基于优化的方法可以充分考虑输入信息，可以得到近似理论最优解，但是消耗计算资源，且容易受到环境的影响，例如，自车远处障碍物的变动会直接影响自车近处的路径，因而使得自车在动态环境中行驶时会出现抖动或震荡；基于搜索的方法可以应对复杂场景，但是会大量消耗计算资源，且容易产生不平滑的规划路径，通常只适合低速场景，对动态场景的支持不够好；基于势场的方法可以综合考虑环境中的障碍物信息，但会产生局部折弯、震荡的情况；基于特征点的方法，在低速简单场景会产生稳定有效的避障路径，但是对于复杂动态场景会出现明显震荡。
综上，现有的避障路径规划方法在简单场景中通常具有较好的规划效果，但对于真实的动态交通场景，尤其是在复杂狭窄的场景中，容易产生路径晃动，无法保证车辆安全平稳行驶。
针对现有技术存在问题，本申请实施例通过构建近场范围内的第一横向背景代价势场，且该第一横向背景代价势场的目标极小值点横向两侧的背景代价具备单调性。有效的体现了自车与障碍物的横向碰撞趋势，以此为基础规划的避障路径可以使行驶装置在复杂狭窄的环境与障碍物保持最佳的横向距离，而且避免了所规划的避障路径受较远处环境变动的影响，从而能够提高避障路径的稳定性。
为了更好的理解本申请实施例的方案，在进行方法的描述之前，首先结合附图对本申请实施例能够应用的场景和系统构架进行简单的描述。
图2是本申请实施例提供的一种复杂狭窄的城市道路场景示例图。如图2所示，自车在道路上行驶时，在可避障范围内存在大量的动态障碍物(编号1-5)和/或静态障碍物(编号6)，构成狭窄拥挤的交通场景。其中，静态障碍物可以是停放在道路两边的车辆或其他障碍物，也可以是临时放置在道路中的其他障碍物；动态障碍物可以是在自车的可避障范围内行驶的其他车辆。在上述障碍物的影响下，可供自车行驶的空间狭窄，行驶路径复杂。应理解，在该场景中，参考路径和横向可避障范围，以及所有障碍物的尺寸、位置、速度及预测轨迹投影的占据地图都已知。
图3是本申请实施例提供的一种避障路径规划的系统架构示例图。该系统架构300包括参考路径规划器310、障碍物决策器320、避障路径规划器330以及速度规划器340。
其中，参考路径规划器310用于向避障路径规划器330提供参考路径和横向避障范围。在本申请实施例中，参考路径是指道路的中心线，应理解，参考路径也可以有其他的表现形式，在此不做限定。横向可避障范围指自车行驶过程中，在垂直于路径方向上可以进行避障的范围，例如，横向可避障范围可以是当前行驶道路的宽度。
障碍物决策器320用于向避障路径规划器330提供包含障碍物及障碍物预测轨迹投影的占据地图。应理解，本申请不限定障碍物决策器320获取障碍物及障碍物预测轨迹投影的占据地图的获取方式。
避障路径规划器330用于根据参考路径规划器310和障碍物决策器320提供的参考路径、横向避障范围以及障碍物及障碍物预测轨迹投影的占据地图进行避障路径的规划。
在本申请实施例中，避障路径规划器330包括备选路径生成模块331和备选路径评价模块332。其中，备选路径生成模块331用于根据参考路径和横向避障范围生成一簇备选路径，关于备选路径的生成方式将在下文详细描述，这里不再赘述；备选路径评价模块332用于根据备选路径本身及障碍物的占据信息对备选路径进行评价，选取当前帧最优路径作为当前避障路径。
在一种实现方式中，备选路径评价模块332所选取的当前帧最优路径可以作为当前输出的避障路径。在另一种实现方式中，避障路径规划器330还可以包括输出路径仲裁模块433，用于根据当前帧最优路径及历史(上一帧)路径决定最终输出的避障路径。
可选地，避障路径规划器330还可以包括路径时域平滑模块334，用于对最终输出的避障路径进行时域平滑。
避障路径规划器330将规划好的避障路径输出给后端的速度规划器340。在避障路径上如果自车与障碍物之间存在碰撞的可能性时，速度规划器会根据碰撞障碍物与自车的相对位置、速度进行减速规划。
图4是本申请实施例提供的一种行驶装置的避障路径的规划方法示例图。应理解，该 方法400可以应用在上述场景200中，还可以应用在上述系统架构300中。如图4所示，该方法400包括步骤S410-S430。下面对这些步骤进行详细描述。
S410，获取距离地图。
该距离地图包括多个第一栅格，多个第一栅格中的每个第一栅格用于记录该第一栅格离最近被障碍物占据的第一栅格的距离。
可选地，在获取距离地图之前，方法400还包括：获取占据地图，该占据地图包括多个第二栅格，多个第二栅格中的每个第二栅格用于记录该第二栅格中存在障碍物的概率。并根据占据地图，得到距离地图。
可选地，当第二栅格中存在障碍物的概率大于或等于一定阈值时，可以认为该第二栅格中存在障碍物；当第二栅格中存在障碍物的概率小于一定阈值时，可以认为该第二栅格中不存在障碍物。
可选地，该多个第二栅格中的每个第二栅格还可以直接用于记录该第二栅格中是否存在障碍物。示例性地，当该第二栅格中记录的数值为1时，表示存在障碍物；当该第二栅格中记录的数值为0时，表示不存在障碍物。
应理解，该占据地图是由上游模块提供的，例如上述障碍物决策器320。该占据地图中包含自车附近给定区域内的障碍物信息。还应理解，障碍物信息指障碍物或障碍物预测轨迹的信息，被障碍物占据的栅格是指被障碍物或障碍物预测轨迹投影所占据的栅格。
应理解，在获取占据地图之后，需要将占据地图转换为距离地图。关于转化方式将在下文具体实现方式中进行详细描述，这里不再赘述。
由于在车前某一特定区域范围内，该行驶装置无论选择哪一条避障路径都无法避开该区域内的障碍物。又由于避障路径规划是根据上游模块提供的占据地图信息进行的，而上游模块提供的障碍物信息除了包括动静态障碍物的实际位姿和障碍物的预测轨迹投影，还有可能包括自车车头部位的误检导致的障碍物占据情况，或者是由于地面不平坦使得占据地图在该区域出现了障碍物信息。此时，如果还考虑这一区域的障碍物，就会影响路径规划，使得在行驶过程中总要避开这一区域的障碍物，从而引起路径的大幅震荡。
因此，在本申请实施例中，可以筛除占据地图中位于第一区域的所述第二栅格。
该第一区域位于行驶装置的前方，若该第一区域内存在障碍物，该行驶装置无法避开该第一区域内的障碍物；然后根据筛除后的占据地图，得到距离地图。从而有效避免了行驶装置因为障碍物的误检而进行的误避让。保证剩余的占据栅格只包含避障算法所需的必要信息，提高了避障路径的平稳性。关于栅格筛除的方式将在下文具体实现方式中进行详细描述，这里不再赘述。
S420，根据距离地图确定近场范围内的第一横向背景代价势场。
其中，近场范围为与所述行驶装置的纵向距离小于或等于第一阈值的范围，第一横向背景代价势场为横向位置上的障碍物对行驶装置构成的第一斥力场，第一横向背景代价势场包括目标极小值点，目标极小值点横向两侧的背景代价具有单调性，该背景代价表示近场范围内横向位置上的障碍物对该行驶装置构成的近场横向碰撞代价。
可选地，上述行驶装置可以是自动驾驶车辆，也可以是机器人，又或是其他自主行驶装置，本申请对此不做限定。为便于描述，本申请均用自车来表示。
应理解，近场范围可以是从Frenet坐标系的原点出发沿着路径方向纵向延伸5m的范 围。在本申请实施例中，该原点可以为自车后轴中心。还应理解，通常自车的长度可能为3-6m，因此近场范围内横向位置上的障碍物可以理解为自车横向两侧的障碍物。该近场范围可以根据自车长度、行驶速度、场景复杂程度确定，本申请不做限定。
应理解，第一横向背景代价势场为横向位置上的障碍物对自车构成的第一斥力场。其中，斥力场是由横向位置上的障碍物对行驶装置构成的，能够表示障碍物对行驶装置的碰撞趋势，会引导行驶装置避开障碍物，也可以称为碰撞场或风险场。应理解，第一斥力场能够表示单调性处理后的障碍物对行驶装置构成的碰撞趋势。
也就是说，第一横向背景代价势场能够指示横向位置上自车与障碍物之间的碰撞趋势。通常，背景代价越大，说明自车与障碍物之间的距离越小，越容易发生碰撞；背景代价越小，说明自车与障碍物之间的距离越大，在该背景代价所处的横向位置上行驶时，能够与障碍物保持安全的距离。
应理解，该目标极小值点横向两侧的背景代价具有单调性，具体是在目标极小值点左侧的势场单调递减，在目标极小值点右侧的势场单调递增。意味着目标极小值点两侧的背景代价均比目标极小值点处的背景代价大。即就是说，目标极小值点所在的横向位置的势场最低，自车能够沿着该横向位置所在的路径稳定行驶。
可选地，第一横向背景代价势场可以呈“V”型，也可以呈“/”型,或者呈“\”型。还应理解，单调函数的斜率可以为定值，也可以是随横向位置不断变化的值，本申请不做限定。为便于描述，在下文具体实施例中，均采用V型的第一横向背景代价势场。
S430，根据第一横向背景代价势场，确定当前避障路径。
可选地，根据该第一横向背景代价势场，确定当前避障路径可以是根据第一横向背景代价势场中的目标极小值点所处的横向位置，确定当前避障路径。也就是说，根据势场最低的横向位置确定当前的避障路径，使得自车能够沿着势场最低的位置稳定行驶。
在本申请实施例中，规划当前避障路径时考虑到近场范围内的横向背景代价势场，一方面使得在规划避障路径时，仅考虑近场范围内的障碍物信息，避免采用现有避障方法规划路径时，远处障碍物发生变动，就会造成自车近处的路径变动，从而使得行驶不稳定；另一方面，通过构建近场范围内的第一横向背景代价势场，有效体现自车与障碍物的横向碰撞趋势，以此为基础规划的避障路径可以使自车在复杂狭窄的环境与障碍物保持最佳的横向距离，从而能够提高避障路径的稳定性。
可选地，可以先根据距离地图确定近场范围内的第二横向背景代价势场，该第二横向背景代价势场为横向位置上的障碍物对该行驶装置构成的第二斥力场，该第二横向背景代价势场包括至少一个极小值点；从至少一个极小值点中确定目标极小值点；对目标极小值点横向两侧的背景代价进行单调性处理，生成第一横向背景代价势场，单调性处理的斜率大于或等于第二阈值。例如可以取斜率为大于等于10/1m的任何一个数值，具体的取值需要结合实际情况确定，在此不做限定。
应理解，第二斥力场能够表示在单调性处理之前的障碍物对行驶装置构成的碰撞趋势。也就是说，对第二斥力场进行单调性处理会得到第一斥力场。
在本申请实施例中，第一横向背景代价势场的构建方式可以是：首先根据距离地图构建近场范围内的第二横向背景代价势场，然后在第二横向背景代价势场中的至少一个极小值点中确定一个目标极小值点，并对该目标极小值点横向两侧的背景代价进行单调性处 理，生成该第一横向背景代价势场。并通过设定该单调性处理的斜率大于或等于第二阈值，使得目标极小值点所在的横向位置的背景代价势场明显低于其他横向位置处的背景代价势场，从而能够保证自车在势场最低的路径上稳定行驶。换句话说，意味着当自车侧面障碍物在横向上稍微靠近一点点时，自车所处的横向位置的背景代价势场就会显著增大，使得自车缓慢平稳的避让并运行到背景代价势场最低的横向位置，与障碍物保持安全的距离。
可选地，目标极小值点与行驶装置之间不存在障碍物。
换句话说，也就是该目标极小值点所代表的横向位置与该行驶装置所在的横向位置之间在该近场范围内不存在障碍物。
可选地，该目标极小值点可以是与行驶装置之间在该近场范围内不存在障碍物的所有极小值点中的最小极小值点。
应理解，在实际场景中，可能存在多个极小值点所在的横向位置与自车所在的横向位置之间都不存在障碍物。此时，可以将这些极小值点中的最小极小值点作为目标极小值点；也可以根据实际情况选取其中的一个作为目标极小值点，例如，最小极小值点和次小极小值点的背景代价很接近，且次小极小值点所在的横向位置与自车更接近，此时可以将次小极小值点作为目标极小值点。为便于描述，在后述的实施例中，均采用与行驶装置所在的横向位置之间在该近场范围内不存在障碍物的所有极小值点中的最小极小值点作为目标极小值点。
在本申请实施例中，该目标极小值点与该行驶装置之间在该近场范围内不存在障碍物，使得自车能够平稳地运行到目标极小值点所处的横向位置，避免因为行驶装置所处的横向位置与目标极小值点所处的横向位置之间存在障碍物使得自车无法运行到目标极小值点所处的横向位置，从而能够提高路径的稳定性。而且，当目标极小值点可以是与行驶装置所在的横向位置之间在该近场范围内不存在障碍物的所有极小值点中的最小极小值点时，能够使得自车运行过程中与障碍物保存最佳的安全距离。
可选地，在根据距离地图确定近场范围内的第二横向背景代价势场之前，方法400还可以包括：生成特征矢量，所述特征矢量包括路径起始点和采样目标点；根据参考路径、历史路径和所述特征矢量，生成多条备选路径。
应理解，在本申请实施例的路径规划中，首先可以根据自车所处的位置生成多条备选路径，随后对多个备选路径进行评价选取出最优的备选路径作为当前的避障路径。关于备选路径的生成方式将在下文的实施例中进行详细描述，这里不再赘述。
可选地，可以根据采样目标点，对参考路径进行横向偏移生成多条虚拟平行线路径；根据距离地图计算所述多条虚拟平行线路径中每条虚拟平行线路径在近场范围内的每个路点离该路点最近障碍物的距离；根据每条虚拟平行线路径上距离的最小值计算该条虚拟平行线的背景代价；根据多条虚拟平行线路径在近场范围内的背景代价生成近场范围内的第二横向背景代价势场。应理解，关于第一横向背景代价势场和第二横向背景代价势场的构建方式将在下文的具体实现方式中详细描述，这里不再赘述。
可选地，在本申请实施例中，对备选路径进行评价时，除了根据背景代价进行评价，还可以结合碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，使得所确定的避障路径能够综合更多的因素，从而能够进一步提高避障路径的平稳性和安 全性。
因此，在本申请实施例中，在根据第一横向背景代价势场，确定当前避障路径之前，还可以根据距离地图计算多条备选路径中的每条备选路径上的碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项。
其中，碰撞代价用于评价行驶装置在备选路径上行驶时是否会发生碰撞，约束代价用于评价备选路径上的各个路点是否在可避障范围内，切换代价用于评价备选路径与历史路径之间的偏差，横向偏移代价用于评价所述备选路径的横向偏离道路中心线的程度，平滑度代价用于评价备选路径的平滑程度。
在进行备选路径的评价时，可以根据第一横向背景代价势场中的背景代价，或者根据第一横向背景代价势场中的背景代价以及碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，计算每条备选路径上的最终代价；然后根据最终代价，确定当前避障路径。
应理解，在本申请实施例中，可以根据各项代价的加权求和计算最终代价。应理解，为了使得自车能够与障碍物保持安全的距离，可以将背景代价和碰撞代价的权重取得较大一些。应理解，关于上述各种代价的获取以及具体评价方式将在下文具体实现方式中详细描述，这里不再赘述。
可选地，方法400还包括：根据历史路径和当前避障路径的仲裁结果，确定最终避障路径。
应理解，历史路径是指自车上一帧所行驶的路径，当前避障路径是指当前所规划的路径。根据历史路径和当前避障路径的仲裁结果，确定最终避障路径，也就是说，在历史路径和当前避障路径中选取一个路径，作为最终输出的避障路径。
在本申请实施例中，在确定当前避障路径时，根据当前避障路径和历史路径的差异，确定最终避障路径。避免了在复杂狭窄环境中运行时纯粹基于代价函数或纯逻辑得到的避障路径会产生帧间跳动的情况，有效降低了规划方法对代价函数的依赖，从而保证路径的稳定。
可选地，如果当前避障路径明显优于历史路径，则切换至当前避障路径；如果当前避障路径没有明显优于历史路径，则沿用历史路径。避免了在复杂狭窄环境中运行时纯粹基于代价函数或纯逻辑得到的避障路径会产生帧间跳动的情况，从而保证路径的稳定。关于具体的仲裁条件将在下文具体实现方式中进行详细描述，这里不再赘述。
可选地，方法400还可以包括：根据历史路径和最终避障路径进行时域平滑。在本申请实施例中，做获取到最终避障路径时，还可以根据历史路径和最终避障路径进行时域平滑。此时，如果最终避障路径为沿用历史路径，时域平滑后的避障路径仍为历史路径；如果最终避障路径为当前避障路径，则根据当前避障路径和历史路径进行时域平滑。有效避免了在复杂狭窄环境中切换避障路径时产生帧间跳动的情况，从而保证路径的稳定。
可选地，方法400还可以包括：将最终避障路径输出给速度规划器，该速度规划器用于根据最终避障路径与最终避障路径上障碍物信息，进行速度规划。在本申请实施例中，输出最终避障路径后，速度规划器会根据最终避障路径上的障碍物信息进行速度规划，如果发现该路径上存在可能发生碰撞的障碍物，就会进行减速规划。从而提高路径的安全性。
下面将结合附图，对本申请实施例提供的一种优选的避障路径的规划方法做详细描 述。
作为一种优选方式，图5是本申请实施例提供的一种避障路径的规划方法的核心流程示例图。应理解，图5所示的规划方法可以应用于图2所示的复杂狭窄的场景中。如图5所示，该核心流程500包括步骤S510、S520、S530和S540。下面对这些步骤进行详细描述。
S510，备选路径的生成。
备选路径的生成包括步骤S511、S512和S513。图6是本申请实施例提供的一种备选路径生成的场景示例图，下面结合附图6对这些步骤进行详细描述。
S511，特征矢量生成。
具体地，基于历史路径和参考路径规划器输入的参考路径和可避障范围，生成包含路径起始点(start点)，采样目标点(goal点，简称采样点)在内的特征矢量，如图6中所示的箭头。
应理解，start点可以在历史路径中寻找。在本申请实施例中，将上一帧自车后轴中心所经过的路径作为历史路径，因此可以在历史路径中找到离自车后轴中心点最近的路点，则该点的实际位置和朝向可以构成start点特征矢量。应理解，在每条路径上都包括很多的路点，路点与路点之间的距离可以根据实际情况确定。例如，可以每隔0.1m设置一个路点，也可以每隔0.1*v 自车设置一个路点，其中，v 自车为自车的行驶速度。
goal点在车体前方的可避障范围内，以参考路径为基准，通过采样方法得到。还应理解，一条备选路径可以包含多个goal点，但只包含一个start点。start点和goal点都通过矢量的方式表征。
S512，备选路径拼接。
具体地，利用参考路径、历史路径和特征矢量连接生成多段式备选路径，如图6所示。第一段：在start点后方，通过截取历史路径得到。第二段：利用光滑曲线连接start点和goal点，使得曲线满足矢量约束生成路径。第三段：将参考路径平移到在第二段对应的goal之后。三段路径依次连接，通过拼接的方式得到整条备选路径。
S513，路径空间域平滑。
对每条备选路径进行空间平滑，使得各段路径光滑连接。可选地，可以在大地直角坐标系下，每间隔几个路点生成一个平滑控制点，基于控制点列和3阶B样条生成平滑的备选路径。可选地，可以基于三个路点生成一个平滑控制点，也可以基于四个、五个、六个等，本申请对此不做限定。
示例性地，在如图2所示的场景中可以按照图7所示方式生成多条备选路径。具体地，假设参考路径为当前车道的中心线，车道宽度即可避障范围为3.5m，车道中心线的左右两边的宽度各为1.75m。此时可以在历史路径中找到离自车后轴中心点最近的路点，以该点的实际位置和朝向构成start点特征矢量。再生成3排goal点，各排goal点s坐标分别为1.2*max(车速，4.17m/s)，2.0*max(车速，4.17m/s)，3.5*max(车速，4.17m/s)，每排共11个goal点，在对应的纵向位置处的可避障范围内等距分布着与参考路径朝向相同的goal点。其中，max(车速，4.17m/s)是指取当前车速和4.17m/s中较大的数值；1.2*max(车速，4.17m/s)是指1.2乘当前车速和4.17m/s中较大的数值得到第一排goal点的s坐标。应理解，以上数据仅仅作为一种示例，不能作为对本申请的限定。
在确定好特征矢量之后，利用参考路径、历史路径和特征矢量连接生成三段式备选路径，包括历史段，多项式段和延伸段。历史段：在start点后方，通过截取历史路径得到。多项式段：利用start点和多排goal点的矢量约束生成路径。具体方法为以任一goal点结合start点的矢量约束，计算Frenet系下的三次多项式方程，离散化为路点得到第二段路径。延伸段：将参考路径平移到在第二段对应的goal点之后。三段路径依次连接，通过拼接的方式得到整条备选路径。每条路径与goal点一一对应，在该示例中，共产生33条备选路径。
最后对每条备选路径进行空间平滑。具体做法为：在大地直角坐标系下，每间隔4个路点生成一个平滑控制点，基于控制点列和3阶B样条生成平滑的备选路径。
在生成备选路径之后，需要对备选路径进行评价，也就是在这些备选路径中选择一条合适的备选路径作为当前避障路径。下面对备选路径的评价过程做以详细的描述。
S520，备选路径评价。
在本实施例中，备选路径的评价主要包括步骤S521、S522和S523，下面结合附图对这些步骤进行详细描述。
S521，无效占据栅格筛除。
基于上述生成的备选路径簇和自车位姿，容易看出，在车前某一特定区域范围内，无论自车选择哪一条备选路径，都无法避开该区域内的障碍物。又由于避障路径规划是根据上游模块提供的占据地图信息进行的，而上游模块提供的障碍物信息除了包括动静态障碍物的实际位姿和障碍物的预测轨迹投影，还有可能包括自车车头部位的误检出现了障碍物占据的情况，或者是由于地面不平坦使得占据地图在该区域出现了障碍物信息。此时，如果还考虑这一区域的障碍物，会影响路径规划，使得在行驶过程中总要避开误检的障碍物，从而使得路径规划产生大幅度变化，从而引起路径的震荡。
因此在避障路径规划中构建近场横向碰撞势场之前，可以不考虑这一区域的障碍物，在占据地图中可以将该区域的栅格作为无效栅格进行筛除，图8是本申请实施例提供的一种无效占据栅格的示例图。下面结合附图对无效占据栅格筛除的方法进行详细描述。
可选地，可以构建全碰撞三角区以筛选无效占据栅格。图9是本申请实施例提供的一种全碰撞三角区的示例图。可选地，占据地图覆盖范围可以为车体前方50m，后方10m，左右20m，分辨率为0.2m*0.2m(即每个格子的占据面积)，占据地图中被占据的部分栅格值为1，未被占据部分栅格值为0。应理解，全碰撞三角区是指在某个三角区域内，无论自车选择哪个路径，都无法避开该区域内的障碍物。三角区的顶点为备选路径的start点所在位置，其底边等于自车宽度，三角区底边的两个顶点位于最外侧的备选路径上。将落在该全碰撞三角区的占据栅格直接筛除。
可选地，在获取到筛除无效占据栅格后的占据地图后，可以将占据地图转换为距离地图。图10是本申请实施例提供的一种占据地图转换为距离地图的示例图。可选地，可以通过openCV的距离转换DistanceTransform函数将占据地图转换为距离地图。如图10所示，在占据地图中，栅格值为1的部分表示该栅格被占据，即存在障碍物；栅格值为0的部分表示该栅格没有被占据，即不存在障碍物。占据地图所转换的距离地图中的栅格大小与占据地图相等，距离地图中每个栅格的值为该栅格到最近被占据栅格点的距离。可以看出，距离地图中栅格值为0的部分表示该栅格到障碍物的距离为0，即该栅格处被占据； 距离地图中栅格值为0.2的部分表示该栅格到最近障碍物的距离为0.2m。应理解，距离地图中每个栅格的值为该栅格到最近被占据栅格点的距离是指如果一个栅格周围通常会存在多个障碍物，但是该栅格中只记录该栅格到最近障碍物的距离。可选地，该距离可以为直线距离。
在本申请实施例中，通过对车前无效栅格的筛除，有效避免了自车因为障碍物的误检而进行的误避让。保证剩余的占据栅格只包含避障算法所需的必要信息，提高了避障路径的稳定性。
还应理解，在避障路径的规划中将无效栅格筛除，意味着在避障路径规划中不考虑这一区域的障碍物。但在实际场景中，如果这个区域存在真实的障碍物，速度规划模块会做相应的减速处理以避免碰撞的发生。
S522，V型的近场横向背景代价势场生成。
在本实施例中，可以基于上述筛除无效栅格后的距离地图信息，构建V型的近场横向背景代价势场，即构建上述近场范围内的第一横向背景代价势场。下面结合附图对V型近场横向碰撞势场的构建过程做以详细描述。
图11是本申请实施例提供的一种V型的近场横向背景代价势场的生成方法示例图。该方法即步骤S522包括步骤S5221、S5222和S5223，下面对这些步骤进行详细描述。
S5221，根据距离地图，计算近场横向位置上的背景代价。
应理解，在根据距离地图，计算近场横向位置上的背景代价之前，首先需要基于图6或7中所生成的多条备选路径上的采样目标点，对参考路径进行横向偏移生成多条虚拟平行线路径。该生成的虚拟平行线路径如图12所示，即将参考路径横向平移到采样目标点所在的横向位置上。在本申请实施例中，该虚拟平行线路径也可以称为是背景代价虚拟线，自车在每条背景代价虚拟线的碰撞趋势可以用该条背景代价虚拟线的背景代价来表述。背景代价越大，发生碰撞的可能性越大；背景代价越小，发生碰撞的可能性越小。
在生成背景代价虚拟线之后，计算每条背景代价虚拟线上的背景代价。
具体地，利用距离地图信息，计算每条背景代价虚拟线上在近场范围内的所有路点离其最近障碍物的距离，以其距离最小值计算该条背景代价虚拟线的背景代价。
示例性地，根据距离最小值计算背景代价的公式可以如下：
Figure PCTCN2021090523-appb-000001
其中，cost back为背景代价；d min.i为某条背景代价虚拟线上在近场范围内的所有路点离其最近障碍物的距离中的最小值，并将具有该距离最小值的路点记为路点i；width为车宽。在上述公式(1)中临界值为
Figure PCTCN2021090523-appb-000002
还应理解，上述公式(1)中的临界值仅作为一种示例，在实际应用中，临界值还可以取
Figure PCTCN2021090523-appb-000003
中任一数值，只要能保证自车在横向上与障碍物处于一定的安全距离即可。
示例性地，某条背景代价虚拟线在近场范围内共有4个路点，分别离最近障碍物的距离为0.6、0.5、0.7和1.2，则以最小值0.5计算该条背景代价虚拟线的背景代价。又因为
Figure PCTCN2021090523-appb-000004
因此该条背景代价虚拟线的背景代价
Figure PCTCN2021090523-appb-000005
还应理解，虚拟平行线路径上包括多个路点，路点与路点之间的距离可以根据实际情况确定。例如，可以每隔0.1m设置一个路点，也可以每隔0.1*v 自车设置一个路点，其中，v 自车为自车的行驶速度。本申请对此不做限定。
可选地，计算每条虚拟平行线路径上在近场范围内的某个路点离最近障碍物的距离可以直接取该路点所在栅格的距离值，也可以由其周围的四个栅格中的距离信息通过双线性插值方法计算得到，本申请对此不做限定。
S5222，基于近场横向位置上的背景代价，生成初始的近场横向背景代价势场。
具体地，根据多条背景代价虚拟线在近场范围内的背景代价生成初始的近场横向背景代价势场，即上述近场范围内的第二横向背景代价势场。
S5223，根据初始的近场横向背景代价势场，构建V型的近场横向背景代价势场。
具体地，图13是本申请实施例提供的一种近场横向背景代价势场单调性处理原理示例图。
应理解，在实际的交通场景中，多个背景代价虚拟线的背景代价所形成的初始的近场横向背景代价势场在横向(沿L方向)分布上通常有多个极小值。需要先找到初始的近场横向背景代价势场中的多个极小值点，并记录横向位置和对应背景代价值。示例性地，如图13中的左图所示，在背景代价的横向分布中，存在3个极小值点。随后分别找到当前自车的横向位置以及三个极小值的位置。例如，当前自车横向位置为L＝-0.5；点1的坐标为L＝1.3，Cost back＝0.13；点2的坐标为L＝-1.2，Cost back＝0.53；点3的坐标为L＝-0.2，Cost back＝1.7。除此之外，从图13中还可以看出，在实际交通场景的近场范围内还存在最大背景代价，应理解，最大背景代价所在的横向位置处的背景代价虚拟线上在近场范围内的某个路点距离障碍物的距离为0，即该路点上存在障碍物。在本实施例中，由于离障碍物的距离为0时根据公式(1)无法计算背景代价，因此可以采用自定义最大背景代价的方式，例如取最大背景代价为10 9。
可以看出，在上述3个极小值点中，点1的背景代价最低，意味着如果在该背景代价虚拟线上行驶，能够与障碍物保持最佳的横向距离，如果按照传统的优化算法选择路径时，会以点1的横向位置所在的背景代价虚拟线作为最优路径。但在实际交通场景中，如图13中的左图所示，点1与自车中间存在障碍物。又因为在近场范围(0-5m)内，自车长度通常高达3-6m，因此在实际场景中，由于中间障碍物的存在，自车并不能成功行驶到背景代价最低的路径。也就是说，在近场横向背景代价势场中背景代价最低的点并不一定是最合适的点。
因此，基于以上问题，在本申请实施例中，还需要在多个极小值中确定目标极小值，该目标极小值是与自车所在的横向位置之间在该近场范围内不存在障碍物的所有极小值点中的最小极小值点。
具体地，将极小值点按背景代价值从小到大排列，从最小的极小值出发，依次检查该点与自车横向位置之间是否存在障碍物，即是否存在最大背景代价(与障碍物距离为0)的点。如果存在，则对应极小值点视为非法点，丢弃该点；如果不存在，则以此点为基准，将其两侧背景代价以固定步长处理为单调函数，形成V形近场横向碰撞势场。简单来讲， 先将极小值从小到大进行排列，然后从当前最小的极小值出发进行检查，如果当前最小的极小值与自车之间存在障碍物，则丢弃该点，再检查下一个最小的极小值点；如果当前最小的极小值与自车之间不存在障碍物则以该点为基准，将其两侧的背景代价以一定斜率构建成单调函数，不必再考虑之前两侧的背景代价。而且，每条背景代价虚拟线的背景代价都可以直接由对应横向的位置根据斜率计算得到。
示例性地，如图13左图所示，点1与自车横向之间存在最大背景代价，因此丢弃该点；点2与自车之间不存在最大背景代价，因此该点为合法极小值点，以该点为基础，将点两侧的背景代价做单调性处理得到V型的近场横向碰撞势场，且单调函数的斜率为10/1m，如图13右图所示。
应理解，做单调性处理时的函数斜率应该大于等于一定阈值，例如可以取斜率为大于等于10/1m的任何一个数值，具体的取值需要结合实际情况确定，在此不做限定。以使最优横向位置的背景代价远低于其他路径，在实际自车行驶过程中若稍微偏移最优横向位置时，最终代价便会明显增大，从而使得自车能够在背景代价最低的横向位置稳定行驶，从而提供自车行驶的稳定性和安全性。
S523，路径代价计算。
优选地，在本实施例中，备选路径的评价可以通过背景代价结合碰撞代价、约束代价、切换代价、横向偏移代价、平滑度代价进行。下面结合公式和附图对这些代价进行一一描述。
其中，碰撞代价，用于评价自车在备选路径上行驶时与障碍物发生碰撞的可能性。具体地，计算所有备选路径中每条备选路径上每个路点的碰撞代价，再对每条备选路径上所有的路点的碰撞代价求和得到该条备选路径的最终代价。应理解，每个路点的碰撞代价由其离最近障碍物的距离计算得到，距离越近碰撞代价越大。
示例性地，公式(2)是本申请实施例提供的一种碰撞代价的计算方式：
Figure PCTCN2021090523-appb-000006
其中，cost obj为碰撞代价；d min.i为备选路径上第i个路点离障碍物的距离；max(d min.i/0.7,0.2)为取d min.i/0.7和0.2中较大的值；pt i.s为备选路径上第i个路点的纵坐标；width为车宽。
在上述公式(2)中临界值为
Figure PCTCN2021090523-appb-000007
当d min.i大于临界值时视为不发生碰撞，当d min.i小于等于临界值时视为发生碰撞。还应理解，上述公式中的临界值仅作为一种示例，在实际应用中，临界值还可以取
Figure PCTCN2021090523-appb-000008
中任一数值，只要能保证自车在该路径行驶不发生碰撞即可。
上述提到每个路点的碰撞代价是由其离最近障碍物的距离计算得到，因此在计算碰撞代价之前，需要先计算每个路点离其最近障碍物的距离。关于路点离最近障碍物的距离的计算方式，将结合图14和图15进行详细描述。
图14是本申请实施例提供的一种车体多圆碰撞模型的示例图。在本实施例中，将车体等效为多圆模型，将车体多圆碰撞模型的后轴中点所在圆心与备选路径上的每个路点重合(应理解，自车的行驶路径为自车后轴中心所经过的路径)，车头朝向与路点朝向一致。此时，对应备选路径上的路点离障碍物的距离为多圆模型中多个圆心离障碍物的最小距离。当任一个圆心离障碍物距离小于一定距离时，该路点状态视为碰撞，该点的碰撞代价由该距离定义。
示例性地，以图14中的路点i为例对路点i到障碍物的距离进行详细描述，如图14所示，备选路径上某个路点i与四圆模型(C1-C4)的后轴中点所在圆心重合，此时路点i离障碍物的距离d min.i为四圆模型中4个圆心离障碍物的最小距离，即圆C4的圆心离障碍物的距离。若该距离小于一定阈值，则认为在该路点i的状态为发生碰撞；若大于一定阈值，则认为该路点i的状态为不发生碰撞。
进一步地，按照上述方法，遍历某条备选路径上的所有路点，若任一路点的状态为碰撞，则整条路径的状态为碰撞。在本申请实施例中，将每条备选路径上处于发生碰撞状态的所有的路点的纵向距离中的最小值记为该备选路径的碰撞距离。基于上面的描述可以得知，碰撞距离与路点离障碍物的距离d min.i并不相同。
可选地，在本申请实施例中，可以利用距离地图计算多圆模型中的多个圆心离最近障碍物的距离。且在计算距离时，为了避免栅格地图导致的精度丢失，可以利用双线性插值的方法计算具体圆心所在位置与最近障碍物的精确距离，如图15所示。应理解，点P可以为多圆模型中某个圆的圆心，点ABCD分别为距离地图里栅格地图所代表的路点位置以及离最近障碍物的距离值。因而，P点的距离值可由ABCD四个点的距离信息通过双线性插值方法计算得到。在通过上述方式分别计算出多个圆心到障碍物的距离后，取最小的距离为路点i离障碍物的距离d min.i。
约束代价，用来判断备选路径上的各个路点是否在可避障范围约束内，若超出可避障范围，则计算该点的约束代价，否则约束代价为0。具体地，每条备选路径的整体约束代价可以由各个路点的约束代价求和得到，如公式(3)：
Figure PCTCN2021090523-appb-000009
其中，cost margin为约束代价；pt i.s为备选路径上第i个路点的纵向坐标值；pt i.l为第i个路点的横向坐标值；LM为左侧约束，即可避障范围的左侧边界线的横向坐标值；RM为右侧约束，即可避障范围的右侧边界线的横向坐标值。可见，某条备选路径上在可避障范围之外的路点越多且在可避障范围之外的路点的纵向坐标值越小，该条路径的约束代价就越大；若备选路径上所有的路点都在可避障范围内时，约束代价就为0。因此，在约束代价为0的备选路径上行驶，能够保证自车在可避障范围内行驶。
切换代价，用于描述备选路径与历史路径之间的偏差。分为偏移切换代价和区域切换代价。
其中，偏移切换代价描述备选路径横向路径相对于历史路径的变化情况。在本实施例中，可以取当前备选路径goal点的横向偏移与上一帧历史输出路径的goal点横向偏移差 的绝对值。如下公式所示：
cost lchange＝|goal old.l-goal now.l|   (4)
其中，cost lchange为偏移切换代价；goal old为历史路径的goal点横向偏移，即历史路径的goal点的横坐标；goal now为当前备选路径goal点的横向偏移，即当前备选路径的goal点的横坐标。
应理解，横向偏移即偏移道路中心线的程度。
区域切换代价描述备选路径横向大范围的变化程度。当路径在相邻区域中跳动时，区域切换代价较小，当路径在左右两个区域之间跳动时，区域切换代价很大，此时区域切换代价可以用于评价路径的大幅变化。
示例性地，以横向距离为依据，可以将可避障范围划分为如图16所示的左Z1、中Z2、右Z3三个区域。当横向可避障范围为3m，道路中心线的横坐标为0时，三个区域可以按照如下方式进行划分。即，当goal点的横坐标小于-0.5时，goal点位于区域1(Z1)；当goal点的横坐标大于-0.5且小于0.5时，goal点位于区域2(Z2)；当goal点的横坐标大于0.5时，goal点位于区域3(Z3)。
Figure PCTCN2021090523-appb-000010
其中，zone表示区域，goal.l为goal点的横坐标值。
基于上面的描述，备选路径与历史路径之间的区域切换代价值可以按照如下公式(6)计算：
Figure PCTCN2021090523-appb-000011
其中，cost zone为区域切换代价，zone old为历史路径的goal点所在的区域；zone now为备选路径的goal点所在的区域。
应理解，在实际操作中，也可以划分为四个区域、五个区域、六个区域等，本申请对划分区域的数目不做限制。为便于描述，在本申请实施例中，均划分为三个区域。还应理解，区域的划分可以是等长度划分也可以不等，本申请对此不作限定。
应理解，当横向路径划分为更多区域时，例如，四个区域。可以采用1或2作为临界值。举例来讲，历史路径的goal点所在的区域以及备选路径的goal点所在的区域差的绝对值大于1时，区域切换代价为1；小于等于1时，区域切换代价为0。还可以是，历史路径的goal点所在的区域以及备选路径的goal点所在的区域差的绝对值大于2时，区域切换代价为1；小于等于2时，区域切换代价为0。具体区域的划分以及临界值的确定应结合实际道路情况确定，不做限定。
横向偏移代价，用于描述备选路径的横向偏离道路中心线的程度。具体地，可以使用备选路径的goal点的横向偏移描述该代价，如公式(7)所示。
cost offset＝|goal now.l|   (7)
其中，cost offset为横向偏移代价；goal now.l为当前备选路径goal点的横向偏移。
平滑度代价，描述备选路径的平滑程度，利用备选路径曲率表示。具体地，可以使用 备选路径各路点曲率平方的积分描述该代价，如公式(8)所示：
Figure PCTCN2021090523-appb-000012
其中，cost smooth为平滑度代价；pt i.curve为备选路径上第i个路点处的曲率，pt i.s为备选路径上第i个路点的纵向坐标值。应理解，曲率越大，代价越大。
优选地，在本申请实施例中，在计算出每条备选路径的上述各个代价之后，可以通过每条备选路径上的上述各个代价的加权求和结果作为该条备选路径的最终代价。随后根据所有备选路径的最终代价值，以最终代价最小的备选路径作为当前帧最优路径，当前帧最优路径即上述当前避障路径。
具体地，每条备选路径的最终代价可以按照公式(9)进行计算：
cost total＝w back*cost back+w obj*cost obj+w margin*cost margin+      (9)
w lchange*cost lchange+w zone*cost zone+w offset*cost offset+w smooth*cost smooth
其中，cost total为备选路径的最终代价；cost back为背景代价；w back为背景代价的权重；cost obj为碰撞代价；w obj为碰撞代价的权重；cost margin为约束代价；w margin为约束代价的权重；cost lchange为偏移切换代价；w lchange为偏移切换代价的权重；cost zone为区域切换代价；w zone为区域切换代价的权重；cost offset为横向偏移代价；w offset为横向偏移代价的权重；cost smooth为平滑度代价；w smooth为平滑度代价的权重。
应理解，根据本申请实施例的方案在计算最终代价时，综合了多种代价的影响，使得当前帧最优路径能够满足各个方面的要求。
而且，在计算过程中，可以将背景代价的权重和碰撞代价的权重取得较大一些。一方面，使得最优路径的最终代价远低于其他备选路径。在实际自车行驶过程中若稍微偏移最优路径时，最终代价便会明显增大，从而使得自车能够在代价最低的路径稳定行驶。另一方面，代价最低的路径其碰撞代价和背景代价也最优，使得自车在行驶过程中，当自车侧面障碍物逐渐逼近时，自车可以缓慢避让，从而与障碍物保持安全的距离。
示例性地，本申请实施例中设置各项代价的权重可以为：
w back＝1.0×10 6
w obj＝1.0×10 6
w margin＝1.0×10 -1
w lchange＝1.0×10 -6
w zone＝1.0×10 -3
w offset＝1.0×10 -3
w smooth＝1.0×10 -6
应理解，在本申请实施例中，可以根据实际情况结合上述一项或多项代价评价备选路径。还应理解，各项代价的权重可以根据实际情况取值，在此不做限定。
S530，输出路径仲裁。
在获取到当前帧最优路径即上述当前避障路径时，可以基于历史路径和当前帧最优路径进行仲裁，输出最终避障路径。具体仲裁方式如图17所示。应理解，当前帧最优路径即就是上述的当前避障路径。
图17是本申请实施例提供的一种输出避障路径的仲裁方法示例图。如图17所示，该仲裁方法即步骤S530包括步骤S531-S536。下面对这些步骤进行详细描述。
S531，判断历史路径当前是否存在碰撞。若存在碰撞，转步骤S532；若不存在碰撞，转步骤S533。
S532，判断当前最优路径是否存在碰撞。若存在碰撞，转步骤S534；若不存在碰撞，转步骤S535。
S533，判断当前最优路径的代价是否明显优于历史代价。若明显优于历史代价，转步骤S535；若不明显优于历史代价，转步骤S536。
S534，判断当前最优路径的碰撞距离是否明显优于历史碰撞距离。若明显优于，转步骤S535；若不明显优于，转步骤S536。
S535，输出当前最优路径。
S536，输出历史路径。
具体地，在进行仲裁之前，首先要根据上一帧生成历史路径所使用的goal点的信息，在当前帧重新生成历史路径，并计算历史路径的最终代价和碰撞状态。
如果历史路径在当前不存在碰撞，则根据当前最优路径的最终代价和历史路径的最终代价决定输出的避障路径。如果当前最优路径的最终代价明显优于历史路径的最终代价，则当前最优路径作为输出的避障路径，否则，继续沿用历史路径。应理解，当前最优路径的最终代价优于历史路径的最终代价是指当前最优路径的最终代价小于历史路径的最终代价。
如果历史路径当前存在碰撞，但是当前帧最优路径不发生碰撞，则使用当前帧最优路径作为当前输出的最终路径。还应理解，如果当前帧最优路径发生碰撞，则表示当前帧所有路径都发生碰撞。在这种情况中，如果当前帧最优路径的碰撞距离与历史碰撞距离接近，如图18所示，则沿用上一帧历史路径；如果当前帧最优路径的碰撞距离明显优于历史碰撞距离，如图19所示，则以当前最优路径作为输出的最终避障路径。
应理解，当前帧最优路径的碰撞距离明显优于历史碰撞距离，是指当前帧最优路径的碰撞距离减去历史碰撞距离的差值大于或等于一定阈值，例如，5m。
当前最优路径的最终代价明显优于历史路径的最终代价，是指当前最优路径的最终代价与历史路径的最终代价的比值小于或等于一定阈值，例如60％。
S540，路径时域平滑。
本实施例中，经过仲裁得到最终避障路径后，还可以对该避障路径进行时域平滑。
且本实施例中，每条路径都对应唯一goal点，因此可以利用goal点对避障路径进行时域平滑。具体地，根据上一帧平滑后的采样目标点goal old(即上一帧历史路径的采样目标点goal old)和当前帧选取的最终采样目标点goal final(即经过仲裁选取的最终避障路径的采样目标点goal final)，利用惯性滤波生成时域平滑的采样目标点goal filter。
应理解，当最终避障路径为历史路径时，当前帧选取的最终采样目标点为历史路径的采样目标点，此时，经过时域平滑和不经过时域平滑的结果都为历史路径的采样目标点。当最终避障路径为当前帧最优路径时，当前帧选取的最终采样目标点为当前最优采样目标点，图20示出了该情况下的时域平滑示例图。在生成时域平滑的采样目标点goal filter后，基于平滑后的采样目标点goal filter重新生成三段式路径，得到最终输出路径。
平滑后的目标点goal filter的具体位置可以按照公式(10)和(11)进行计算：
goal filter.l＝w l*goal old.l+(1-w l)*goal final.l   (10)
goal filter.s＝w s*goal old.s+(1-w s)*goal final.s   (11)
其中，goal filter.l为平滑后的采样目标点的横向位置；goal filter.s为平滑后的采样目标点的纵向位置；goal old.l为上一帧平滑后的采样目标点的横向位置；goal old.s为上一帧平滑后的采样目标点的纵向位置；goal final.l为当前帧选取的最终采样目标点的横向位置；goal final.s为当前帧选取的最终采样目标点的纵向位置；w l是在横向上进行平滑的权重，w s是在纵向上进行平滑的权重。
应理解，在本实施例中，横向和纵向上进行平滑时所使用的权重可以根据实际情况设定。例如，横向和纵向上进行平滑的权重可以取1，即平滑后的采样目标点goal filter的位置和上一帧平滑后的采样目标点goal old的位置相同，也即沿用历史路径；横向和纵向上进行平滑的权重可以取0，即平滑后的采样目标点goal filter的位置和当前帧选取的最终采样目标点goal final的位置相同，也即根据路径仲裁选择的最终路径的采样目标点作为平滑后的采样目标点goal filter。当路径仲裁的结果为采用当前帧最优路径，即当前帧选取的最终路径为当前帧的最优路径时，权重也可以根据实际路况选取0-1中任何一个数值进行时域平滑，使得自车可以综合考虑历史路径和当前帧最优路径的情况，防止直接运行到最优路径时产生的帧间跳动。应理解，横向和纵向上的权重可以相同，也可以不相同。示例性地，在本实施例中，可以取：w l＝0.92，w s＝0.85。
在本实施例中，一方面，生成了近场范围内的横向背景代价势场，当自车侧面障碍物逐渐横向逼近时，自车可以缓慢避让，与障碍物保持安全距离。另一方面，通过对无效栅格的筛除，有效避免了自车因为此处的占据栅格造成的误避让。又一方面，背景代价结合路径仲裁策略可以保证自车的稳定避障。
示例性地，在图2所示的场景中，在4号和5号车处于图示位置时，自车可以沿着两车中间行驶；当右侧5号车车辆缓慢靠近自车时，背景代价形成的V形势场的最低值会逐渐向左移动，使得自车在可避让范围的约束内选择靠左侧的备选路径，与障碍物保持1m的横向距离保证自车安全。而且由于背景代价的权重很高，例如10 6，当障碍物稍微靠近时，背景代价就会急剧变化，使得最终代价也发生明显的变化，必然会使得每帧得到的最优路径代价会小于历史最优路径代价的60％，便会不断触发路径从历史路径切换至当前最优路径，使得自车及时避让。除此之外，在整个备选路径评价时，还会使用到碰撞代价，例如，1号障碍物的存在使得在1号障碍物附近的备选路径都形成碰撞，其碰撞代价也会很高，因此根据背景代价和碰撞代价等多种代价的加权求和，在该场景中最终输出的避障路径会如图21所示向左偏移，对需避让的障碍物进行有效避让。
上述是本申请实施例应用在复杂狭窄的交通场景中的具体示例，应理解，结合上述近场范围内的横向背景代价势场(即第一横向背景代价势场)和碰撞代价，使得本申请实施例还可以应用在简单高速场景中进行快速的避障。
图22是本申请实施例提供的一种简单的城市道路场景示例图。如图22所示，自车在结构化道路上快速行驶时，在可避障范围内存在少量的动静态障碍物(编号1-2)。但是这些障碍物在自车的前方道路上交叉排列，自车需要利用小型S弯进行快速的避障。应理 解，在该场景中，参考路径和横向可避障范围，以及所有障碍物的尺寸、位置、速度及预测轨迹的占据地图都已知。
为便于描述，将复杂狭窄的场景记为场景一，将简单高速场景记为场景二。
可选地，在针对场景二进行路径规划时，可以结合上述近场范围内的横向背景代价势场(即第一横向背景代价势场)和碰撞代价进行路径的规划；除此，还可以结合无效栅格的筛除、综合其他多种代价获取最终代价、路径仲裁、时域平滑等中的一项或多项进行路径的规划。
下面同样以一种优选的方式，对场景二中的路径规划方法进行简单描述。
在一种优选的实现方式中，图5所示的规划方法同样可以应用于图22所示的场景二中。但与在场景一中应用时的备选路径具体的生成方式(S510步骤)不同。图23是本申请实施例提供的又一种备选路径生成的场景示例图，下面结合具体示例对场景二的备选路径的生成进行详细描述。
示例性地，在场景二中，同样可以假设参考路径为当前车道的中心线，车道宽度即可避障范围为3.5m，车道中心线的左右两边的宽度各为1.75m。在场景二中，关于start点和goal点的生成与的方法与场景一中的基本一致，主要区别在于场景一中可以每排生成11个goal点，而在场景二中，每排生成3个goal点。
在确定好特征矢量之后，利用参考路径、历史路径和特征矢量连接生成全连接式备选路径。路径分为三段，包括历史段、全连接段和延伸段。其中，历史段在起始点后方，通过截取历史路径得到。全连接段由起始点和多排采样点之间连接生成。具体的方法为起始点和第1排采样点分别连接，再由第1排采样点和第2排采样点分别连接，以此类推，可以得到多排采样点之间的连接关系，例如每排采样点有3个点，共3排采样点，则可以得到3*(3*3)*(3*3)＝243条备选的轨迹。全连接段各段通过三次多项式方程链接，需要满足经过goal点的矢量约束。延伸段：将参考路径平移到在全连接段对应的采样点之后。三段路径依次连接，通过拼接的方式得到整条备选路径。在该实施例中，共产生243条备选路径。
最后对每条备选路径进行空间平滑。关于平滑的方式可参考场景一的描述，不再赘述。
应理解，以上仅仅作为一种示例，仅为了清楚的描述场景一和场景二中备选路径生成方式的不同，并不能作为对本申请的限定。应理解，上述每排目标点的个数和目标点的排数和备选路径生成的个数均可以结合实际情况确定，不做限定。
在场景二中，关于备选路径的评价(S520步骤)和输出路径仲裁(S530步骤)均可参考对于场景一的具体实现方式的描述。
另外，由于备选路径连接方式的不同，使得对于场景二和场景一的路径规划中在路径时域平滑(S540步骤)时也稍有区别。
具体地，在场景二中，记录上一帧平滑后的采样目标点簇goal old，此处的goal old包含3个goal点，每一排采样点都对应一个goal点，基于当前帧选取的最终采样目标点(3个)，利用惯性滤波生成时域平滑的采样目标点簇。基于平滑后的采样目标点簇重新生成三段式路径，得到最终输出路径。对应goal点时域平滑的公式和参数同公式(10)和(11)。
在该场景二中，当自车处于两个障碍物交叉排列的场景时，拥有足够高横向避让自由度，因此可以按照上述方式规划得到足够长的无碰撞行驶轨迹，使得自车快速避让通过该 交通场景。而且在快速避障障碍物的过程中，自车可以与障碍物保持足够侧向距离且保证路径稳定。
具体地，本实施例可产生如图24所示的避障路径。由于该实施例中采用的全连接的方法生成备选路径，其路径的备选形态多，充分拓展的解空间，可得到一条对多个物体进行绕障的无碰撞轨迹。且基于碰撞代价对路径进行评价使得当前所规划的避障路径(最优的备选路径)在整条路径上都能够与障碍物1和2保持0.7m的安全距离。且在行驶过程中，自车经过障碍物侧面时，基于背景代价会对行驶路径做出调整，使得自车与障碍物保持1m的安全距离，从而使得在障碍物1和2所处的纵向范围内构成无碰撞路径，快速通过两个障碍物。而且，如果自车在图示位置所规划的路径使得自车运行到障碍物1和2侧面时，就能够保持1m的安全距离时，输出路径仲裁策略就会对历史路径和行驶过程中重新规划的路径进行仲裁，从而沿用历史路径，不对路径进行切换，完成快速避障。
而在场景二中，如果采用场景一所使用的路径规划方法，生成的备选路径的形态较少(如图7所示)，此时在多个备选路径中进行选择时，无法直接获取到一条无碰撞的行驶路径，在行驶过程只能分别对1和2号障碍物进行避让，使得开始所规划的路径能够避让1号障碍物，但是在该路径的前方与2号障碍物存在碰撞(即2号障碍物在图7中最下面一条路径的第三段上)。此时有碰撞的行驶路径会影响后端的速度规划器，引起自车减速，导致自车只能低速在该场景种行驶。
图25是本申请实施例提供的一种行驶装置的避障路径的规划装置示例图。如图25所示，该规划装置2500包括获取模块2510和处理模块2520。
其中，该获取模块2510用于获取距离地图，该距离地图包括多个第一栅格，该多个第一栅格中的每个第一栅格用于记录该第一栅格离最近被障碍物占据的第一栅格的距离；该处理模块2520用于根据该距离地图确定近场范围内的第一横向背景代价势场，该近场范围为与该行驶装置的纵向距离小于或等于第一阈值的范围，该第一横向背景代价势场为横向位置上的障碍物对该行驶装置构成的第一斥力场，第一横向背景代价势场包括目标极小值点，该目标极小值点横向两侧的背景代价具有单调性，该背景代价表示近场范围内横向位置上的障碍物对该行驶装置构成的近场横向碰撞代价；根据该第一横向背景代价势场，确定当前避障路径。
可选地，根据该距离地图确定近场范围内的第一横向背景代价势场包括：根据该距离地图确定近场范围内的第二横向背景代价势场，该第二横向背景代价势场为横向位置上的障碍物对该行驶装置构成的第二斥力场，该第二横向背景代价势场包括至少一个极小值点；从该至少一个极小值点中确定目标极小值点；对该目标极小值点横向两侧的背景代价进行单调性处理，生成该第一横向背景代价势场，该单调性处理的斜率大于或等于第二阈值。
可选地，该目标极小值点与该行驶装置之间不存在障碍物。
可选地，在获取距离地图之前，该获取模块2510还可以用于：获取占据地图，该占据地图包括多个第二栅格，该多个第二栅格中的每个第二栅格用于记录该第二栅格中存在障碍物的概率；该处理模块2520还可以用于：根据该占据地图，得到距离地图。
可选地，该根据所述占据地图，得到距离地图包括：筛除该占据地图中位于第一区域的第二栅格，该第一区域位于行驶装置的前方，若该第一区域内存在障碍物，该行驶装置 无法避开该第一区域内的障碍物；根据筛除后的占据地图，得到该距离地图。
可选地，在该根据距离地图确定近场范围内的第二横向背景代价势场之前，该处理模块2520还可以用于：生成特征矢量，该特征矢量包括路径起始点和采样目标点；根据参考路径、历史路径和所述特征矢量，生成多条备选路径；该根据所述距离地图确定该近场范围内的第二横向背景代价势场包括：根据该采样目标点，对该参考路径进行横向偏移生成多条虚拟平行线路径；根据该距离地图计算该多条虚拟平行线路径中每条虚拟平行线路径在该近场范围内的每个路点离该路点最近障碍物的距离；根据该每条虚拟平行线路径上该距离的最小值计算该条虚拟平行线的背景代价；根据该多条虚拟平行线路径在该近场范围内的背景代价生成该近场范围内的第二横向背景代价势场。
可选的，在该根据第一横向背景代价势场，确定当前避障路径之前，该处理模块2520还用于：根据该距离地图计算该多条备选路径中的每条备选路径上的碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，其中，该碰撞代价用于评价该行驶装置在该备选路径上行驶时是否会发生碰撞，该约束代价用于评价该备选路径上的各个路点是否在可避障范围内，该切换代价用于评价该备选路径与该历史路径之间的偏差，该横向偏移代价用于评价该备选路径的横向偏离道路中心线的程度，该平滑度代价用于评价该备选路径的平滑程度；根据该第一横向背景代价势场，确定当前避障路径包括：根据该第一横向背景代价势场中的背景代价，或者根据该第一横向背景代价势场中的背景代价以及该碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，计算该每条备选路径上的最终代价；根据该最终代价，确定该当前避障路径。
可选的，该处理模块2520还用于：根据历史路径和该当前避障路径的仲裁结果，确定最终避障路径。
可选的，该根据历史路径和该当前避障路径的仲裁结果，确定最终避障路径，包括：若该历史路径存在碰撞，但该当前避障路径不存在碰撞，则以该当前避障路径作为该最终避障路径；或者，若该历史路径和该当前避障路径都存在碰撞，且该当前避障路径的碰撞距离与该历史路径的碰撞距离之差的绝对值大于或等于第三阈值，则以该当前避障路径和该历史路径中碰撞距离较大者作为该最终避障路径；或者，若该历史路径和该当前避障路径都存在碰撞，且该当前避障路径的碰撞距离与该历史路径的碰撞距离之差的绝对值小于第三阈值，则以该历史路径作为该最终避障路径；或者，若该历史路径不存在碰撞，且该当前避障路径的该最终代价与该历史路径代价之比小于或等于第四阈值，则以该当前避障路径作为该最终避障路径；或者，若该历史路径不存在碰撞，且该当前避障路径的该最终代价与该历史路径代价之比大于第四阈值，则以该历史路径作为该最终避障路径；其中，该碰撞距离为该每条备选路径上处于发生碰撞状态的所有的路点的纵向距离中的最小纵向距离。
可选地，该处理模块2520还可以用于：根据历史路径和最终避障路径进行时域平滑。
可选地，该规划装置2500还可以包括输出模块，用于将避障路径输出给速度规划器，该速度规划器用于根据最终避障路径与最终避障路径上障碍物信息，进行速度规划。
图26是本申请实施例提供的一种避障路径的规划装置的硬件结构示例性框图。该装置2600(该装置2600具体可以是一种计算机设备)包括存储器2610、处理器2620、通信接口2630以及总线2640。其中，存储器2610、处理器2620、通信接口2630通过总线 2640实现彼此之间的通信连接。
存储器2610可以是只读存储器(read only memory，ROM)，静态存储设备，动态存储设备或者随机存取存储器(random access memory，RAM)。存储器2610可以存储程序，当存储器2610中存储的程序被处理器2620执行时，处理器2620用于执行本申请实施例的规划方法的各个步骤。
处理器2620可以采用通用的中央处理器(central processing unit，CPU)，微处理器，应用专用集成电路(application specific integrated circuit，ASIC)，图形处理器(graphics processing unit，GPU)或者一个或多个集成电路，用于执行相关程序，以实现本申请方法实施例的规划方法。
处理器2620还可以是一种集成电路芯片，具有信号的处理能力。在实现过程中，本申请的规划方法的各个步骤可以通过处理器2620中的硬件的集成逻辑电路或者软件形式的指令完成。
上述处理器2620还可以是通用处理器、数字信号处理器(digital signal processing，DSP)、专用集成电路(ASIC)、现成可编程门阵列(field programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器2610，处理器2620读取存储器2610中的信息，结合其硬件完成本申请实施例的规划装置中包括的模块所需执行的功能，或者执行本申请方法实施例的规划方法。
通信接口2630使用例如但不限于收发器一类的收发装置，来实现装置1000与其他设备或通信网络之间的通信。例如，实现上述获取模块2510和输出模块的功能。
总线2640可包括在装置2600各个部件(例如，存储器2610、处理器2620、通信接口2630)之间传送信息的通路。
本申请实施例还提供了一种车辆，包括上述避障路径的规划装置2500或2600，能够执行上述任一种规划方法，从而使得车辆在复杂狭窄场景中能够平稳的行驶并与横向上的车辆保持安全的距离。应理解，车辆可以是电动汽车，例如，纯电动汽车、增程式电动汽车、混合动力电动汽车、燃料电池汽车、新能源汽车等，本申请对此不做具体限定。
本申请实施例还提供了一种计算机系统，上述任一项规划方法(算法)可运行在该计算机系统上，该计算机算力应保证对上述任一项规划算法的更新率达到一定数值，例如，30HZ。使得可以根据实时的环境情况进行路径规划，保证行驶装置安全平稳的运行。
除非另有定义，本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同。本文中在本申请的说明书中所使用的术语只是为了描述具体的实施例的目的，不是旨在于限制本申请。
需要说明的是，本文中涉及的第一、第二、第三或第四等各种数字编号仅为描述方便进行的区分，并不用来限制本申请实施例的范围。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及 算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(Read-Only Memory，ROM)、随机存取存储器(Random Access Memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。

Claims (21)

  1. 一种行驶装置的避障路径的规划方法，其特征在于，包括：
    获取距离地图，所述距离地图包括多个第一栅格，所述多个第一栅格中的每个第一栅格用于记录该第一栅格离最近被障碍物占据的第一栅格的距离；
    根据所述距离地图确定近场范围内的第一横向背景代价势场，所述近场范围为与所述行驶装置的纵向距离小于或等于第一阈值的范围，所述第一横向背景代价势场为横向位置上的障碍物对所述行驶装置构成的第一斥力场，所述第一横向背景代价势场包括目标极小值点，所述目标极小值点横向两侧的背景代价具有单调性，所述背景代价表示近场范围内横向位置上的障碍物对所述行驶装置构成的近场横向碰撞代价；
    根据所述第一横向背景代价势场，确定当前避障路径。
  2. 根据权利要求1所述的规划方法，其特征在于，所述根据所述距离地图确定近场范围内的第一横向背景代价势场包括：
    根据所述距离地图确定所述近场范围内的第二横向背景代价势场，所述第二横向背景代价势场为横向位置上的障碍物对所述行驶装置构成的第二斥力场，所述第二横向背景代价势场包括至少一个极小值点；
    从所述至少一个极小值点中确定所述目标极小值点；
    对所述目标极小值点横向两侧的背景代价进行单调性处理，生成所述第一横向背景代价势场，所述单调性处理的斜率大于或等于第二阈值。
  3. 根据权利要求1或2所述的规划方法，其特征在于，所述目标极小值点与所述行驶装置之间不存在障碍物。
  4. 根据权利要求1-3中任一项所述的规划方法，其特征在于，在所述获取距离地图之前，所述方法还包括：
    获取占据地图，所述占据地图包括多个第二栅格，所述多个第二栅格中的每个第二栅格用于记录该第二栅格中存在障碍物的概率；
    所述获取距离地图包括：
    根据所述占据地图，得到所述距离地图。
  5. 根据权利要求4所述的规划方法，其特征在于，所述根据所述占据地图，得到所述距离地图包括：
    筛除所述占据地图中位于第一区域的所述第二栅格，所述第一区域位于所述行驶装置的前方，若所述第一区域内存在障碍物，所述行驶装置无法避开所述第一区域内的障碍物；
    根据筛除后的占据地图，得到所述距离地图。
  6. 根据权利要求2所述的规划方法，其特征在于，在所述根据所述距离地图确定所述近场范围内的第二横向背景代价势场之前，所述方法还包括：
    生成特征矢量，所述特征矢量包括路径起始点和采样目标点；
    根据参考路径、历史路径和所述特征矢量，生成多条备选路径；
    所述根据所述距离地图确定所述近场范围内的第二横向背景代价势场包括：
    根据所述采样目标点，对所述参考路径进行横向偏移生成多条虚拟平行线路径；
    根据所述距离地图计算所述多条虚拟平行线路径中每条虚拟平行线路径在所述近场范围内的每个路点离该路点最近障碍物的距离；
    根据所述每条虚拟平行线路径上所述距离的最小值计算该条虚拟平行线的背景代价；
    根据所述多条虚拟平行线路径在所述近场范围内的背景代价生成所述近场范围内的第二横向背景代价势场。
  7. 根据权利要求6所述的规划方法，其特征在于，在所述根据所述第一横向背景代价势场，确定当前避障路径之前，所述方法还包括：
    根据所述距离地图计算所述多条备选路径中的每条备选路径上的碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，
    其中，所述碰撞代价用于评价所述行驶装置在所述备选路径上行驶时是否会发生碰撞，所述约束代价用于评价所述备选路径上的各个路点是否在可避障范围内，所述切换代价用于评价所述备选路径与所述历史路径之间的偏差，所述横向偏移代价用于评价所述备选路径的横向偏离道路中心线的程度，所述平滑度代价用于评价所述备选路径的平滑程度；
    所述根据所述第一横向背景代价势场，确定当前避障路径包括：
    根据所述第一横向背景代价势场中的背景代价，或者根据所述第一横向背景代价势场中的背景代价以及所述碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，计算所述每条备选路径上的最终代价；
    根据所述最终代价，确定所述当前避障路径。
  8. 根据权利要求1-7中任一项所述的规划方法，其特征在于，所述方法还包括：
    根据历史路径和所述当前避障路径的仲裁结果，确定最终避障路径。
  9. 根据权利要求8所述的规划方法，其特征在于，所述根据历史路径和所述当前避障路径的仲裁结果，确定最终避障路径包括：
    若所述历史路径存在碰撞，但所述当前避障路径不存在碰撞，则以所述当前避障路径作为所述最终避障路径；或者，
    若所述历史路径和所述当前避障路径都存在碰撞，且所述当前避障路径的碰撞距离与所述历史路径的碰撞距离之差的绝对值大于或等于第三阈值，则以所述当前避障路径和所述历史路径中碰撞距离较大者作为所述最终避障路径；或者，
    若所述历史路径和所述当前避障路径都存在碰撞，且所述当前避障路径的碰撞距离与所述历史路径的碰撞距离之差的绝对值小于第三阈值，则以所述历史路径作为所述最终避障路径；或者，
    若所述历史路径不存在碰撞，且所述当前避障路径的所述最终代价与所述历史路径代价之比小于或等于第四阈值，则以所述当前避障路径作为所述最终避障路径；或者，
    若所述历史路径不存在碰撞，且所述当前避障路径的所述最终代价与所述历史路径代价之比大于第四阈值，则以所述历史路径作为所述最终避障路径；
    其中，所述碰撞距离为所述每条备选路径上处于发生碰撞状态的所有的路点的纵向距离中的最小纵向距离。
  10. 一种行驶装置的避障路径的规划装置，其特征在于，所述装置包括：获取模块和处理模块；
    所述获取模块用于获取距离地图，所述距离地图包括多个第一栅格，所述多个第一栅格中的每个第一栅格用于记录该第一栅格离最近被障碍物占据的第一栅格的距离；
    所述处理模块用于根据所述距离地图确定近场范围内的第一横向背景代价势场，所述近场范围为与所述行驶装置的纵向距离小于或等于第一阈值的范围，所述第一横向背景代价势场为横向位置上的障碍物对所述行驶装置构成的第一斥力场，所述第一横向背景代价势场包括目标极小值点，所述目标极小值点横向两侧的背景代价具有单调性，所述背景代价表示近场范围内横向位置上的障碍物对所述行驶装置构成的近场横向碰撞代价；根据所述第一横向背景代价势场，确定当前避障路径。
  11. 根据权利要求10所述的规划装置，其特征在于，所述根据所述距离地图确定近场范围内的第一横向背景代价势场包括：根据所述距离地图确定所述近场范围内的第二横向背景代价势场，所述第二横向背景代价势场为横向位置上的障碍物对所述行驶装置构成的第二斥力场，所述第二横向背景代价势场包括至少一个极小值点；从所述至少一个极小值点中确定所述目标极小值点；对所述目标极小值点横向两侧的背景代价进行单调性处理，生成所述第一横向背景代价势场，所述单调性处理的斜率大于或等于第二阈值。
  12. 根据权利要求10或11所述的规划装置，其特征在于，所述目标极小值点与所述行驶装置之间不存在障碍物。
  13. 根据权利要求10-12中任一项所述的规划装置，其特征在于，在所述获取距离地图之前，所述获取模块还用于：
    获取占据地图，所述占据地图包括多个第二栅格，所述多个第二栅格中的每个第二栅格用于记录该第二栅格中存在障碍物的概率；
    所述处理模块还用于：
    根据所述占据地图，得到所述距离地图。
  14. 根据权利要求13所述的规划装置，其特征在于，所述根据所述占据地图，得到所述距离地图包括：
    筛除所述占据地图中位于第一区域的所述第二栅格，所述第一区域位于所述行驶装置的前方，若所述第一区域内存在障碍物，所述行驶装置无法避开所述第一区域内的障碍物；根据筛除后的占据地图，得到所述距离地图。
  15. 根据权利要求11所述的规划装置，其特征在于，在所述根据所述距离地图确定所述近场范围内的第二横向背景代价势场之前，所述处理模块还用于：
    生成特征矢量，所述特征矢量包括路径起始点和采样目标点；
    根据参考路径、历史路径和所述特征矢量，生成多条备选路径；
    所述根据所述距离地图确定所述近场范围内的第二横向背景代价势场包括：
    根据所述采样目标点，对所述参考路径进行横向偏移生成多条虚拟平行线路径；根据所述距离地图计算所述多条虚拟平行线路径中每条虚拟平行线路径在所述近场范围内的每个路点离该路点最近障碍物的距离；根据所述每条虚拟平行线路径上所述距离的最小值计算该条虚拟平行线的背景代价；根据所述多条虚拟平行线路径在所述近场范围内的背景代价生成所述近场范围内的第二横向背景代价势场。
  16. 根据权利要求15所述的规划装置，其特征在于，在所述根据所述第一横向背景代价势场，确定当前避障路径之前，所述处理模块还用于：
    根据所述距离地图计算所述多条备选路径中的每条备选路径上的碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，其中，所述碰撞代价用于评价所述行驶装置在所述备选路径上行驶时是否会发生碰撞，所述约束代价用于评价所述备选路径上的各个路点是否在可避障范围内，所述切换代价用于评价所述备选路径与所述历史路径之间的偏差，所述横向偏移代价用于评价所述备选路径的横向偏离道路中心线的程度，所述平滑度代价用于评价所述备选路径的平滑程度；
    所述根据所述第一横向背景代价势场，确定当前避障路径包括：
    根据所述第一横向背景代价势场中的背景代价，或者根据所述第一横向背景代价势场中的背景代价以及所述碰撞代价、约束代价、切换代价、横向偏移代价或平滑度代价中的一项或多项，计算所述每条备选路径上的最终代价；根据所述最终代价，确定所述当前避障路径。
  17. 根据权利要求10-16中任一项所述的规划装置，其特征在于，所述处理模块还用于：
    根据历史路径和所述当前避障路径的仲裁结果，确定最终避障路径。
  18. 根据权利要求17所述的规划装置，其特征在于，所述根据历史路径和所述当前避障路径的仲裁结果，确定最终避障路径，包括：
    若所述历史路径存在碰撞，但所述当前避障路径不存在碰撞，则以所述当前避障路径作为所述最终避障路径；或者，
    若所述历史路径和所述当前避障路径都存在碰撞，且所述当前避障路径的碰撞距离与所述历史路径的碰撞距离之差的绝对值大于或等于第三阈值，则以所述当前避障路径和所述历史路径中碰撞距离较大者作为所述最终避障路径；或者，
    若所述历史路径和所述当前避障路径都存在碰撞，且所述当前避障路径的碰撞距离与所述历史路径的碰撞距离之差的绝对值小于第三阈值，则以所述历史路径作为所述最终避障路径；或者，
    若所述历史路径不存在碰撞，且所述当前避障路径的所述最终代价与所述历史路径代价之比小于或等于第四阈值，则以所述当前避障路径作为所述最终避障路径；或者，
    若所述历史路径不存在碰撞，且所述当前避障路径的所述最终代价与所述历史路径代价之比大于第四阈值，则以所述历史路径作为所述最终避障路径；
    其中，所述碰撞距离为所述每条备选路径上处于发生碰撞状态的所有的路点的纵向距离中的最小纵向距离。
  19. 一种车辆，其特征在于，包括用于执行如权利要求1-9中任一项所述的规划方法的各个模块。
  20. 一种计算机可读介质，其特征在于，所述计算机可读介质存储有程序代码，当所述计算机程序代码在计算机上运行时，使得计算机执行如权利要求1-9中任一项所述的规划方法。
  21. 一种芯片，其特征在于，所述芯片包括处理器与数据接口，所述处理器通过所述数据接口读取存储器上存储的指令，执行如权利要求1-9中任一项所述的规划方法。
PCT/CN2021/090523 2020-07-20 2021-04-28 行驶装置的避障路径的规划方法和装置 WO2022016941A1 (zh)

Priority Applications (5)

Application Number Priority Date Filing Date Title
KR1020237005632A KR20230041752A (ko) 2020-07-20 2021-04-28 이동 장치의 장애물 회피 경로를 계획하기 위한 방법 및 장치
EP21847177.9A EP4180894A4 (en) 2020-07-20 2021-04-28 OBSTACLE AVOIDANCE PATH PLANNING METHOD AND DEVICE FOR MOBILE DEVICE
JP2023504056A JP2023535175A (ja) 2020-07-20 2021-04-28 移動装置の障害物回避経路を計画する方法及び装置
BR112023001066A BR112023001066A2 (pt) 2020-07-20 2021-04-28 Método e aparelho para planejar o percurso de desvio de obstáculos de um aparelho móvel
US18/156,703 US20230159056A1 (en) 2020-07-20 2023-01-19 Method and apparatus for planning obstacle avoidance path of traveling apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010699919.X 2020-07-20
CN202010699919.XA CN113960996A (zh) 2020-07-20 2020-07-20 行驶装置的避障路径的规划方法和装置

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/156,703 Continuation US20230159056A1 (en) 2020-07-20 2023-01-19 Method and apparatus for planning obstacle avoidance path of traveling apparatus

Publications (1)

Publication Number Publication Date
WO2022016941A1 true WO2022016941A1 (zh) 2022-01-27

Family

ID=79459614

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/090523 WO2022016941A1 (zh) 2020-07-20 2021-04-28 行驶装置的避障路径的规划方法和装置

Country Status (7)

Country Link
US (1) US20230159056A1 (zh)
EP (1) EP4180894A4 (zh)
JP (1) JP2023535175A (zh)
KR (1) KR20230041752A (zh)
CN (1) CN113960996A (zh)
BR (1) BR112023001066A2 (zh)
WO (1) WO2022016941A1 (zh)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114859895A (zh) * 2022-04-06 2022-08-05 高斯机器人(深圳)有限公司 Agv充电方法、装置与设备
CN115185286A (zh) * 2022-09-13 2022-10-14 上海仙工智能科技有限公司 一种移动机器人自主绕障规划方法及其任务调度系统
CN115782867A (zh) * 2022-11-17 2023-03-14 上海西井信息科技有限公司 轨迹碰撞风险评估方法、装置、电子设备和存储介质
CN116125995A (zh) * 2023-04-04 2023-05-16 华东交通大学 一种高铁巡检机器人的路径规划方法及系统
CN116859956A (zh) * 2023-09-04 2023-10-10 中船(北京)智能装备科技有限公司 一种无人艇航行路线确定方法、装置及设备
WO2023202485A1 (zh) * 2022-04-22 2023-10-26 武汉路特斯汽车有限公司 一种自动驾驶系统中的轨迹预测方法及系统
CN118192615A (zh) * 2024-05-16 2024-06-14 江苏三铭智达科技有限公司 一种机器人路径控制方法及系统

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230127002A1 (en) * 2021-10-25 2023-04-27 Moovita Pte Ltd Effective path planning systems and methods for autonomous vehicles
CN114995416A (zh) * 2022-05-30 2022-09-02 深圳市优必选科技股份有限公司 全局路径导航方法、装置、终端设备及存储介质
CN115014375B (zh) * 2022-06-06 2023-11-03 北京京深深向科技有限公司 碰撞检测方法、装置及电子设备、存储介质
JP7400912B1 (ja) 2022-09-26 2023-12-19 いすゞ自動車株式会社 自動運転装置
US20240166239A1 (en) * 2022-11-21 2024-05-23 Apollo Autonomous Driving USA LLC Trajectory planning for navigating small objects on road
CN116109658B (zh) * 2023-04-07 2023-06-20 山东金大丰机械有限公司 基于5g技术的收割机控制数据处理方法
CN117429448B (zh) * 2023-10-24 2024-10-25 北京易航远智科技有限公司 障碍物未来占据空间的预测方法、装置、电子设备及介质

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070156286A1 (en) * 2005-12-30 2007-07-05 Irobot Corporation Autonomous Mobile Robot
CN102880186A (zh) * 2012-08-03 2013-01-16 北京理工大学 基于稀疏a*算法和遗传算法的航迹规划方法
CN103760904A (zh) * 2014-02-13 2014-04-30 北京工业大学 一种语音播报式智能车辆路径规划装置与实施方法
EP2806288A1 (en) * 2013-05-24 2014-11-26 Advanced Scientific Concepts, Inc. Automotive auxiliary ladar sensor
CN105974917A (zh) * 2016-05-11 2016-09-28 江苏大学 一种基于新型人工势场法的车辆避障路径规划研究方法
CN107357293A (zh) * 2017-07-31 2017-11-17 上海应用技术大学 移动机器人路径规划方法和系统
CN108241370A (zh) * 2017-12-20 2018-07-03 北京理工华汇智能科技有限公司 通过栅格地图获取避障路径的方法及装置
CN108253984A (zh) * 2017-12-19 2018-07-06 昆明理工大学 一种基于改进a星算法的移动机器人路径规划方法
CN109947119A (zh) * 2019-04-23 2019-06-28 东北大学 一种基于多传感器融合的移动机器人自主跟随系统及方法
US20190235512A1 (en) * 2018-01-30 2019-08-01 Brain Corporation Systems and methods for precise navigation of autonomous devices

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4741292A (en) * 1986-12-22 1988-05-03 The Babcock & Wilcox Company Electro-impulse rapper system for boilers
JP2011253302A (ja) * 2010-06-01 2011-12-15 Toyota Motor Corp 車両用危険度算出装置
DE102013012324A1 (de) * 2013-07-25 2015-01-29 GM Global Technology Operations LLC (n. d. Ges. d. Staates Delaware) Verfahren und Vorrichtung zur Fahrwegfindung
JP6580087B2 (ja) * 2017-06-02 2019-09-25 本田技研工業株式会社 走行軌道決定装置及び自動運転装置
JP6901331B2 (ja) * 2017-06-20 2021-07-14 株式会社東芝 情報処理装置、移動体、情報処理方法、およびプログラム
CN108274465A (zh) * 2018-01-10 2018-07-13 上海理工大学 基于优化a*的人工势场机械臂三维避障路径规划方法
GB2570887B (en) * 2018-02-07 2020-08-12 Jaguar Land Rover Ltd A system for a vehicle
CN110111880B (zh) * 2019-04-22 2021-09-28 北京航空航天大学 柔性针的基于障碍分级的人工势场路径规划方法及装置
CN110244713B (zh) * 2019-05-22 2023-05-12 江苏大学 一种基于人工势场法的智能车辆换道轨迹规划系统及方法
CN110414803B (zh) * 2019-07-08 2022-01-07 清华大学 不同网联程度下自动驾驶系统智能水平的测评方法及装置
CN110865642A (zh) * 2019-11-06 2020-03-06 天津大学 一种基于移动机器人的路径规划方法

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070156286A1 (en) * 2005-12-30 2007-07-05 Irobot Corporation Autonomous Mobile Robot
CN102880186A (zh) * 2012-08-03 2013-01-16 北京理工大学 基于稀疏a*算法和遗传算法的航迹规划方法
EP2806288A1 (en) * 2013-05-24 2014-11-26 Advanced Scientific Concepts, Inc. Automotive auxiliary ladar sensor
CN103760904A (zh) * 2014-02-13 2014-04-30 北京工业大学 一种语音播报式智能车辆路径规划装置与实施方法
CN105974917A (zh) * 2016-05-11 2016-09-28 江苏大学 一种基于新型人工势场法的车辆避障路径规划研究方法
CN107357293A (zh) * 2017-07-31 2017-11-17 上海应用技术大学 移动机器人路径规划方法和系统
CN108253984A (zh) * 2017-12-19 2018-07-06 昆明理工大学 一种基于改进a星算法的移动机器人路径规划方法
CN108241370A (zh) * 2017-12-20 2018-07-03 北京理工华汇智能科技有限公司 通过栅格地图获取避障路径的方法及装置
US20190235512A1 (en) * 2018-01-30 2019-08-01 Brain Corporation Systems and methods for precise navigation of autonomous devices
CN109947119A (zh) * 2019-04-23 2019-06-28 东北大学 一种基于多传感器融合的移动机器人自主跟随系统及方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4180894A4

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114859895A (zh) * 2022-04-06 2022-08-05 高斯机器人(深圳)有限公司 Agv充电方法、装置与设备
WO2023202485A1 (zh) * 2022-04-22 2023-10-26 武汉路特斯汽车有限公司 一种自动驾驶系统中的轨迹预测方法及系统
CN115185286A (zh) * 2022-09-13 2022-10-14 上海仙工智能科技有限公司 一种移动机器人自主绕障规划方法及其任务调度系统
CN115185286B (zh) * 2022-09-13 2022-12-27 上海仙工智能科技有限公司 一种移动机器人自主绕障规划方法及其任务调度系统
CN115782867A (zh) * 2022-11-17 2023-03-14 上海西井信息科技有限公司 轨迹碰撞风险评估方法、装置、电子设备和存储介质
CN115782867B (zh) * 2022-11-17 2024-01-30 上海西井科技股份有限公司 轨迹碰撞风险评估方法、装置、电子设备和存储介质
CN116125995A (zh) * 2023-04-04 2023-05-16 华东交通大学 一种高铁巡检机器人的路径规划方法及系统
CN116859956A (zh) * 2023-09-04 2023-10-10 中船(北京)智能装备科技有限公司 一种无人艇航行路线确定方法、装置及设备
CN116859956B (zh) * 2023-09-04 2023-12-08 中船(北京)智能装备科技有限公司 一种无人艇航行路线确定方法、装置及设备
CN118192615A (zh) * 2024-05-16 2024-06-14 江苏三铭智达科技有限公司 一种机器人路径控制方法及系统

Also Published As

Publication number Publication date
JP2023535175A (ja) 2023-08-16
EP4180894A4 (en) 2023-12-27
EP4180894A1 (en) 2023-05-17
CN113960996A (zh) 2022-01-21
US20230159056A1 (en) 2023-05-25
KR20230041752A (ko) 2023-03-24
BR112023001066A2 (pt) 2023-03-07

Similar Documents

Publication Publication Date Title
WO2022016941A1 (zh) 行驶装置的避障路径的规划方法和装置
CN112068545B (zh) 一种无人驾驶车辆在十字路口的行驶轨迹规划方法、系统及存储介质
JP7479064B2 (ja) 動的障害物を有する環境における動作計画を容易にする装置、方法及び物品
Yu et al. Occlusion-aware risk assessment for autonomous driving in urban environments
WO2021142799A1 (zh) 路径选择方法和路径选择装置
US10948919B2 (en) Dynamic programming and gradient descent based decision and planning for autonomous driving vehicles
CN114270142A (zh) 非结构化的车辆路径规划器
EP4149814A1 (en) Unstructured vehicle path planner
US20190079523A1 (en) Dp and qp based decision and planning for autonomous driving vehicles
WO2019042295A1 (zh) 一种无人驾驶路径规划方法、系统和装置
CN110389584A (zh) 用于评估自动驾驶车辆的轨迹候选项的方法
US11586209B2 (en) Differential dynamic programming (DDP) based planning architecture for autonomous driving vehicles
CN114281084B (zh) 一种基于改进a*算法的智能车全局路径规划方法
US20220153296A1 (en) Trajectory planning with obstacle avoidance for autonomous driving vehicles
Wu et al. Trajectory prediction based on planning method considering collision risk
WO2021050745A1 (en) Dynamic collision checking
US20210262819A1 (en) A mixed regular and open-space trajectory planning method for autonomous driving vehicle
WO2024118997A1 (en) Prediction model with variable time steps
WO2024118765A1 (en) Vehicle trajectory tree search for off-route driving maneuvers
CN117184128A (zh) 基于路径规划的自动驾驶停车方法及装置
CN115420300A (zh) 一种运动调度方法、装置以及介质
Chang et al. On-road trajectory planning with spatio-temporal informed RRT
US20240174265A1 (en) Determining prediction times for a model
US20240092357A1 (en) Trajectory optimization in multi-agent environments
Yang et al. An End-to-end Obstacle Avoidance Algorithm for Autonomous Vehicles Based on Global Path Constraint

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21847177

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2023504056

Country of ref document: JP

Kind code of ref document: A

REG Reference to national code

Ref country code: BR

Ref legal event code: B01A

Ref document number: 112023001066

Country of ref document: BR

WWE Wipo information: entry into national phase

Ref document number: 202317006581

Country of ref document: IN

ENP Entry into the national phase

Ref document number: 2021847177

Country of ref document: EP

Effective date: 20230207

ENP Entry into the national phase

Ref document number: 20237005632

Country of ref document: KR

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 112023001066

Country of ref document: BR

Kind code of ref document: A2

Effective date: 20230119