CN111427369B

CN111427369B - Unmanned vehicle control method and device

Info

Publication number: CN111427369B
Application number: CN202010509758.3A
Authority: CN
Inventors: 樊明宇; 任冬淳; 白钰; 杨秋实; 夏华夏; 朱炎亮; 李鑫
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Online Technology Co Ltd
Priority date: 2020-06-08
Filing date: 2020-06-08
Publication date: 2020-11-03
Anticipated expiration: 2040-06-08
Also published as: CN111427369A

Abstract

The specification discloses a method and a device for controlling an unmanned vehicle, wherein the method comprises the steps of determining the position of each obstacle at the current moment according to acquired environment information, determining a plurality of predicted movement tracks of the obstacles in a specified time period in the future according to historical moments and the positions of the obstacles at the current moment, then determining risk distribution of the obstacles corresponding to the predicted movement tracks at each moment in the specified time period according to the predicted movement tracks of the obstacles aiming at each obstacle, and finally determining total space-time risk distribution according to the determined risk distribution in the specified time period so as to be used for determining a control strategy of the unmanned vehicle. Since the spatiotemporal risk distribution of the obstacle can be determined without determining the driving trajectory of the unmanned vehicle, the risk distribution is determined not based on the collision but based on the objective fact, so that the objective risk distribution caused by the obstacle is determined, and a more accurate control strategy can be determined based on the risk distribution.

Description

Unmanned vehicle control method and device

Technical Field

The application relates to the field of unmanned driving, in particular to an unmanned vehicle control method and device.

Background

At present, in an unmanned vehicle control method, a control strategy of an unmanned vehicle is generally determined based on a collision risk between an external obstacle and the unmanned vehicle.

In the method for determining the collision risk in the prior art, usually, the motion trajectory of an obstacle needs to be predicted first, and then whether the obstacle and the unmanned vehicle collide is determined by judging whether the motion trajectory of the unmanned vehicle intersects with the motion trajectory of the obstacle, so as to determine the collision risk. That is, it is determined whether the obstacle is at risk to the unmanned vehicle based on the determination result of whether the obstacle and the unmanned vehicle collide with each other, and therefore the collision risk in the prior art generally describes the relationship property between the obstacle and the unmanned vehicle.

However, the collision risk in the prior art is determined based on the predicted unmanned vehicle and obstacle movement trajectories, that is, the trajectory of the unmanned vehicle is required to be predicted in advance, so that the control strategy based on the collision risk determination in the prior art actually has the influence of the unmanned vehicle planning the trajectory in advance, but finally, whether the unmanned vehicle travels along the trajectory is uncertain. That is, the pre-planned trajectory of the unmanned vehicle actually also contributes to the impact risk, whereas the planned trajectory for the unmanned vehicle is not an objective fact, and thus the prior art does not determine the impact risk based solely on objective facts. This results in an inaccurate control strategy for determining the unmanned vehicle based on the risk of collision of the obstacle with the unmanned vehicle.

Disclosure of Invention

The unmanned vehicle control method and the unmanned vehicle control device provided by the embodiment of the specification are used for partially solving the problems in the prior art.

The embodiment of the specification adopts the following technical scheme:

the unmanned vehicle control method provided by the specification comprises the following steps:

determining the position of each obstacle at the current moment according to the acquired environmental information around the unmanned vehicle, wherein the environmental information at least comprises an image and laser radar information;

taking the position of each obstacle at the historical moment and the position of each obstacle at the current moment as input, inputting a pre-trained track prediction model, and respectively determining a plurality of predicted motion tracks of each obstacle in a specified time period in the future;

determining risk distribution corresponding to each predicted motion track of the obstacle at each moment in the specified time period according to the determined plurality of predicted motion tracks of the obstacle and a preset first Gaussian distribution parameter for each obstacle;

and determining total space-time risk distribution according to the determined risk distribution at each moment in the specified time period, and determining a control strategy of the unmanned vehicle according to the total space-time risk distribution.

Optionally, determining a total spatiotemporal risk distribution according to the determined risk distribution at each time in the specified time period, specifically including:

determining areas of laser radar information loss caused by obstacles at the current time as shielding areas at the current time according to the laser radar information, and determining the boundaries of the shielding areas;

determining risk distribution corresponding to each boundary at the current moment according to the determined coordinates of each boundary and a preset second Gaussian distribution parameter;

determining traffic identification around the unmanned vehicle according to the environment information, and determining risk distribution corresponding to the traffic identification in the appointed time period according to a risk area corresponding to the preset traffic identification and a preset risk probability;

determining traffic events around the unmanned vehicle according to the environment information, and determining corresponding risk distribution of the traffic events in the specified time period according to preset corresponding risk moments of the traffic events and a third Gaussian distribution parameter;

and determining total space-time risk distribution according to the determined risk distribution corresponding to each predicted movement track of the barrier in the specified time, the determined risk distribution corresponding to the traffic identification, the determined risk distribution corresponding to the traffic incident and the determined risk distribution corresponding to each boundary at the current moment.

Optionally, determining, according to the determined plurality of predicted movement trajectories of the obstacle and a preset first gaussian distribution parameter, risk distribution corresponding to each predicted movement trajectory of the obstacle at each time within the specified time period, specifically including:

determining the probability of the obstacle appearing in each coordinate at each moment in the specified time period according to the plurality of predicted motion tracks, wherein the probability of the obstacle appearing in each coordinate at any moment in the specified time period is normalized;

and determining the risk distribution of the obstacle at each coordinate at each moment in the specified time period according to the determined space-time occupation probability distribution of the obstacle and the preset first Gaussian distribution parameter, wherein for each coordinate, the higher the probability of the obstacle appearing at the coordinate is, and the higher the probability of the coordinate appearing risk caused by the obstacle is.

Optionally, determining, according to the determined space-time occupancy probability distribution of the obstacle and the preset first gaussian distribution parameter, the risk distribution of the obstacle at each coordinate at each time within the specified time period, specifically including:

for each moment in the specified time period, determining the speed and the direction of the obstacle at each coordinate at the moment according to a plurality of predicted motion tracks of the obstacle;

and for each coordinate in the moment, determining the risk distribution of the obstacle at the coordinate at the moment based on the first Gaussian distribution parameter according to the probability of the obstacle appearing at the coordinate, the speed and the direction of the obstacle at the coordinate.

Optionally, determining, according to the probability of the obstacle appearing at the coordinate, the speed and the direction of the obstacle at the coordinate, a risk distribution of the obstacle at the coordinate at the time based on the first gaussian distribution parameter, specifically including:

judging whether the obstacle is a static obstacle at the moment according to the speed of the obstacle at the coordinate;

if so, determining the risk distribution of the obstacle at the coordinate at the moment based on the first Gaussian distribution parameter, wherein the risk distribution is equiaxial two-dimensional Gaussian distribution taking the coordinate as a center;

if not, determining the direction of a long axis of the anisometric two-dimensional Gaussian distribution according to the orientation of the obstacle at the coordinate, determining the length of the long axis of the anisometric two-dimensional Gaussian distribution according to the speed of the obstacle at the coordinate, determining the length of a short axis of the anisometric two-dimensional Gaussian distribution according to the size of the obstacle, and determining the anisometric two-dimensional Gaussian distribution taking the coordinate as the center based on the first Gaussian distribution parameter, the determined short axis and the determined long axis as the risk distribution of the obstacle at the moment, wherein the larger the speed, the longer the long axis and the larger the size, the longer the short axis and the longer the short axis.

Optionally, the method further comprises:

and when the obstacle is not a static obstacle, according to the orientation of the obstacle at the coordinate, taking the Gaussian distribution on one side of the orientation in the anisometric two-dimensional Gaussian distribution determined by taking the coordinate as the center as the risk distribution of the obstacle at the coordinate at the moment.

Optionally, determining traffic identifiers around the unmanned vehicle according to the environment information, and determining risk distribution corresponding to the traffic identifiers in the specified time period according to a risk area corresponding to a preset traffic identifier and a preset risk probability, specifically including:

determining coordinates and traffic identifications of all traffic identifications around the unmanned vehicle according to the environment information;

and aiming at each traffic identification, determining a risk area corresponding to the traffic identification according to the shape, the size and the relative position of a risk area corresponding to the preset traffic identification and the coordinates of the traffic identification, wherein the traffic identification comprises a traffic sign and a traffic marking line.

Optionally, determining the risk distribution corresponding to the traffic event in the specified time period according to a preset risk time corresponding to the traffic event and a third gaussian distribution parameter, specifically including:

determining a traffic event occurring at the current moment and an occurring coordinate according to the environment information;

determining the risk time of the traffic event occurring at the current time according to the preset corresponding relation between each traffic event and the risk time;

and determining the risk distribution corresponding to each traffic event in the specified time period according to the historically determined risk time of other traffic events, the coordinates of the other traffic events, the risk time of the traffic event occurring at the current time, the coordinates of the traffic event occurring at the current time and a third Gaussian distribution parameter.

The present specification provides an unmanned vehicle control device including:

the obstacle determining module is used for determining the position of each obstacle at the current moment according to the acquired environment information around the unmanned vehicle, wherein the environment information at least comprises an image and laser radar information;

the track prediction module is used for inputting the historical time position and the current time position of each obstacle, inputting a pre-trained track prediction model and respectively determining a plurality of predicted motion tracks of each obstacle in a specified time period in the future;

the obstacle risk determining module is used for determining risk distribution corresponding to each predicted movement track of each obstacle at each moment in the specified time period according to the determined plurality of predicted movement tracks of the obstacle and a preset first Gaussian distribution parameter;

and the control strategy module determines total space-time risk distribution according to the determined risk distribution at each moment in the specified time period, and determines the control strategy of the unmanned vehicle according to the total space-time risk distribution.

A computer-readable storage medium, storing a computer program which, when executed by a processor, implements any of the methods described above.

The unmanned vehicle provided by the specification comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize any one of the methods.

The embodiment of the specification adopts at least one technical scheme which can achieve the following beneficial effects:

determining the position of each obstacle at the current moment according to the acquired environmental information, determining a plurality of predicted movement tracks of the obstacles in a future specified time period according to the historical moment and the position of the obstacle at the current moment, then determining the risk distribution corresponding to the predicted movement tracks of the obstacles at each moment in the specified time period according to each predicted movement track of the obstacles aiming at each obstacle, and finally determining the total space-time risk distribution according to the determined risk distribution in the specified time period so as to be used for determining the control strategy of the unmanned vehicle. Since the spatiotemporal risk distribution of the obstacle can be determined without determining the trajectory of the unmanned vehicle, the method provided by the present specification does not determine the risk on the basis of a collision, but rather determines the risk of an objective presence of the obstacle in the surroundings at the present moment on the basis of objective facts. The space-time occupation probability of the barrier is determined, and the distribution of the risk brought by the barrier on the space-time is determined, so that the subsequent determination of the unmanned vehicle control strategy is more accurate and objective, namely, only the total space-time risk distribution is considered from the aspect of driving safety. Therefore, the method provided by the specification can optimize the accuracy of determining the control strategy and improve the efficiency of determining the control strategy because the determined risk distribution is more accurate and the content is richer.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:

fig. 1 is a schematic view of a control process of an unmanned vehicle according to an embodiment of the present disclosure;

fig. 2 is a schematic diagram of an unmanned vehicle collecting environmental information according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram of a predicted motion trajectory provided by an embodiment of the present disclosure;

fig. 4 is a schematic diagram of a probability of a predicted motion trajectory provided in an embodiment of the present disclosure;

FIGS. 5 a-5 d are schematic diagrams of risk distributions provided by embodiments of the present disclosure;

FIG. 6 is a schematic diagram of an occlusion boundary provided by an embodiment of the present specification;

FIG. 7 is a schematic diagram of a risk distribution caused by boundaries provided by embodiments of the present specification;

FIG. 8 is a schematic diagram of a risk distribution of a moving obstacle according to an embodiment of the present disclosure;

fig. 9 and fig. 10 are schematic structural diagrams of an unmanned vehicle control device provided in an embodiment of the present specification;

fig. 11 is a schematic view of an unmanned vehicle for implementing an unmanned vehicle control method according to an embodiment of the present disclosure.

Detailed Description

In order to make the objects, technical solutions and advantages of the present disclosure more apparent, the technical solutions of the present disclosure will be clearly and completely described below with reference to the specific embodiments of the present disclosure and the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person skilled in the art based on the embodiments in the present specification without any inventive step are within the scope of the present application.

The technical solutions provided by the embodiments of the present application are described in detail below with reference to the accompanying drawings.

Fig. 1 is a schematic view of a control process of an unmanned vehicle provided in an embodiment of the present specification, including:

s100: and determining the position of each obstacle at the current moment according to the acquired environment information around the unmanned vehicle, wherein the environment information at least comprises an image and laser radar information.

In the field of unmanned driving at present, unmanned vehicles are mainly controlled by determining an unmanned vehicle control strategy according to sensed environmental information. Similarly, in this specification, in the process of controlling the unmanned vehicle, the environmental information around the unmanned vehicle may be acquired, wherein the environmental information may include at least an image and laser radar information.

In this specification, the unmanned vehicle or the server may execute the unmanned vehicle control process, the former may locally complete control of the unmanned vehicle itself on the unmanned vehicle, and the latter may acquire environmental information from the unmanned vehicle, and then send the environmental information to the server to determine a control policy, and then return the unmanned vehicle to be driven by the unmanned vehicle according to the control policy. For example, in order to reduce the resource consumption of the unmanned vehicle in the unmanned vehicle execution control process, the unmanned vehicle execution control process can be executed by the server, and in order to avoid instability caused by data transmission through a network, the unmanned vehicle execution control process can be executed by the unmanned vehicle.

For convenience of description, the unmanned vehicle is described as an example in the following, and specifically, the processing device on the unmanned vehicle may execute the control process, for example, a driving computer on the unmanned vehicle or a device dedicated to determining the control strategy.

Specifically, the unmanned vehicle can acquire environmental information around the unmanned vehicle through a sensor arranged on the unmanned vehicle. Because this environmental information includes image and lidar information at least in this description, consequently the sensor that this unmanned can the car set up can include image sensor and lidar at least, and wherein, lidar information specifically can be the laser point cloud information that lidar gathered. Then, the unmanned vehicle can determine the position of the obstacle around the unmanned vehicle according to the acquired environmental information. Since there are many mature schemes for determining the position of the obstacle according to the image or the laser point cloud information or the combination of the two, the present specification does not limit how to determine the position of the obstacle. The location may specifically be a coordinate in the world coordinate system. Or the coordinates of the unmanned vehicle under the coordinate system do not influence the subsequent steps by taking the unmanned vehicle as the origin.

Moreover, the collecting direction of the image sensor and the laser radar may be the surrounding of the unmanned vehicle, or the advancing direction of the unmanned vehicle, which is not limited in this specification, as shown in fig. 2. Fig. 2 is a schematic diagram of environment information collected by the unmanned vehicle provided in this specification, where a dotted line sector represents a collection range of laser radar information, a solid line sector represents an image collection range, and it is clear that the two kinds of environment information collection ranges are different, and it is needless to say that the determined control strategy may be affected by the difference in the collected environment information ranges since the control strategy of the unmanned vehicle is determined according to the collected environment information in the following.

In addition, the specific range corresponding to the periphery of the unmanned vehicle is not limited in this specification, and for example, the unmanned vehicle may use environment information within a preset radius with the unmanned vehicle as the environment information around the unmanned vehicle, as a range defined by a circle filled with oblique lines in fig. 2. The length of the preset radius can be as required, for example, 20m, 50m, 100 m.

Of course, since the unmanned vehicle can determine the control strategy in the specified time period in the future by the method provided in the present specification, the maximum driving distance that the unmanned vehicle may reach in the specified time period can also be determined according to the moving speed of the unmanned vehicle and the duration of the specified time period, and this is used as the preset radius to determine the environmental information in the preset radius.

S102: and taking the historical time position and the current time position of each obstacle as input, inputting a pre-trained track prediction model, and respectively determining a plurality of predicted motion tracks of each obstacle in a specified time period in the future.

Since the risk of the surrounding environment of the unmanned vehicle is caused by the obstacle itself when the unmanned vehicle travels, that is, the risk is objectively present, the risk is caused as long as the obstacle is present, rather than being caused only when the obstacle intersects with the travel path of the unmanned vehicle. Therefore, in the specification, the unmanned vehicle based on the bayesian theory can determine the space-time risk distribution brought by the obstacle through the subsequent steps according to the predicted obstacle driving track. The simple understanding is that: the probability that the obstacle appears at each position at different moments in a specified time period brings risks to each coordinate in the surrounding range of the unmanned vehicle. Thus, first the unmanned vehicle can determine a preset movement trajectory within a specified time period of the obstacle.

Specifically, for each obstacle determined in step S100, the unmanned vehicle may first determine the position of the obstacle at each historical time before the current time, and then input a trajectory prediction model trained in advance using the current time and the positions of the obstacles determined at the historical times as inputs, to obtain a plurality of predicted movement trajectories of the obstacles output by the trajectory prediction model in a specified time period in the future.

For each obstacle, the current time and the past time acquired by the unmanned vehicle and the obstacle position may be used to determine the past movement trajectory of the obstacle, and therefore any model that predicts the future movement trajectory of the obstacle according to the past movement trajectory may be used to determine the predicted movement trajectory in step S102.

It should be noted that the current time and the historical time may be time points with unit time intervals, wherein the unit time may be 1s, 1ms, etc., and may be set as needed. For example, the unmanned vehicle acquires the environment information at intervals of 1 s.

In addition, since the predicted motion trajectory output by the trajectory prediction model is not absolutely accurate, in a real scene, the actual motion trajectory of the obstacle may not match the predicted motion trajectory due to an accident or a special case, and therefore, the trajectory prediction model may output a plurality of predicted motion trajectories in this specification. Of course, the present specification does not limit the specific form of the trajectory prediction model, and any trajectory prediction model that can output a plurality of predicted movement trajectories may be applied to step S102 in the present specification.

Further, when the model outputs multiple predicted motion trajectories, a confidence level for each predicted motion trajectory may also typically be output. The confidence coefficient is used for representing the probability of the obstacle traveling along the predicted motion trail, and generally, the higher the confidence coefficient of the predicted motion trail is, the higher the probability of the obstacle traveling along the predicted motion trail in a specified time period is. Of course, the confidence described in the present specification may be referred to as a different name such as an evaluation value in different models. If the model only outputs a plurality of predicted motion trajectories, the confidence of the predicted motion trajectories can be considered to be the same.

Furthermore, in this specification, if the model can only output one predicted motion trajectory according to the input historical motion trajectory, the unmanned vehicle may further add interference to the model parameters in the model, so that the model outputs several predicted motion trajectories that are not identical under the condition that the input of the model is not changed. The interference added to the model parameters can be randomly selected, and the number of the model parameters of the interference can also be randomly set. Of course, in order to reduce the influence of the interference on the accuracy of the model prediction result as much as possible, the number of the model parameters of the interference may be smaller than a preset value, and the interference added to the model parameters may not be larger than the preset proportion of the original model parameters. For example, if the preset proportion is 5% and a certain model parameter is 10, the value range of the model parameter after the interference is increased is 9.5-10.5, and the unmanned vehicle randomly selects a value in the value range as the model parameter.

In the present specification, the trajectory prediction model may be configured to determine a predicted movement trajectory of each obstacle from a historical movement trajectory of the obstacle for each obstacle, or may be configured to output a predicted movement trajectory of each obstacle separately from a historical movement trajectory of each obstacle.

In one embodiment provided in the present specification, the trajectory prediction model may be a prediction model based on a star network structure, that is, the trajectory prediction model may be a neural network model composed of two branch networks. The branch networks are a central branch network and an edge branch network respectively. The central branch network is used for determining the global interaction vector of each obstacle according to the input historical motion tracks of all the obstacles. The global interaction vector can be regarded as shared data, and then predicted motion tracks of different obstacles are determined according to historical motion tracks of the different obstacles and the global interaction vector through an edge branch network.

Fig. 3 is a schematic view of a predicted movement locus of an obstacle determined by an unmanned vehicle provided in the present specification, in which a triangle, a circle, and a square respectively represent positions of different obstacles at the current time, a graph filled with oblique lines represents traffic markings on the ground, and a thick line represents a road boundary. The dotted lines and the solid lines extending from the respective figures representing the obstacles in different forms respectively represent the predicted movement trajectories determined by the different unmanned vehicles, for example, 3 solid lines extending from the unmanned vehicle represented by a triangle represent 3 predicted movement trajectories of the unmanned vehicle.

S104: and determining risk distribution corresponding to each predicted motion track of the obstacle at each moment in the specified time period according to the determined plurality of predicted motion tracks of the obstacle and a preset first Gaussian distribution parameter for each obstacle.

In this specification, after determining each predicted motion trajectory corresponding to each obstacle, the unmanned vehicle may determine, for each obstacle, a space-time occupancy probability of the obstacle in a specified time period, that is, a probability of the obstacle appearing at different positions at different times, and then determine a risk distribution that obeys gaussian distribution, that is, a risk distribution within the specified time period brought by the obstacle, that is, a space-time risk distribution corresponding to the predicted motion trajectory, according to the space-time occupancy probability.

Firstly, for each obstacle, the unmanned vehicle can determine the probability of the obstacle appearing at each coordinate at each moment in the specified time period according to each predicted motion track of the obstacle, and the probability is used as the space-time occupation probability distribution of the obstacle.

In particular, the probability distribution of space-time occupancy can be formulated

Where t denotes the time t within a specified time period,

indicating an obstacle at point p, then

Indicating that an obstacle is present at point p at time t,

indicates the probability of the obstacle appearing at point p at time t, and the probability value range is assumed to be [ mu, 1%]μ is a predetermined constant, and is greater than 0 and less than 1, indicating a background risk, which can also be understood as the lowest risk that an obstacle can cause. The duration of the designated time period may be set according to needs, for example, 10S, 1min, and the like, and the present specification is not limited, and the time points in the designated time period may be determined at unit time intervals in the designated time period, for example, the unit time is 1S, the designated time period is 10S, and the time points in the designated time period include the 1 st, 2 nd, 3 rd, … … th S after the current time point, in accordance with the description in step S100.

Because the predicted motion trail has confidence or an evaluation value for evaluating the accuracy of prediction, the probability of the obstacle traveling along different predicted motion trails, namely the probability of the obstacle appearing on different trails can be determined according to the confidence. Specifically, normalization can be performed according to the confidence degrees of the predicted motion trajectories, and the probability that the obstacle travels along different predicted motion trajectories is determined.

As shown in fig. 4, the graph filled with oblique lines indicates traffic markings on the ground, and the thick lines indicate road boundaries. Assuming that there are 3 predicted motion trajectories for an obstacle, i.e., the four-pointed star pattern in fig. 4, step S102 determines that there are 25%, 30%, and 45% probabilities of driving along the 3 predicted motion estimates, respectively, where the triangle, circle, square, and diamond are the positions of the obstacle on different predicted trajectories at different times within the specified time period. It can be seen that the probability of the obstacle appearing at each coordinate at any one time in a given time period is normalized.

And then, the unmanned vehicle can determine the risk distribution of the obstacle at each coordinate at each moment in a specified time period according to the determined space-time occupation probability distribution of the obstacle and the preset first Gaussian distribution parameter. Wherein, for each coordinate, the higher the probability of the obstacle appearing at the coordinate, the higher the probability of the obstacle causing the risk of the coordinate appearing.

In particular, it can be based on the formula

Determining a risk distribution, the formula representing a risk distribution of a p point caused by the presence of the obstacle at the p ' point at the time t, and the distribution obeying a gaussian distribution, μ = p ' representing that the center of the risk distribution is the p ' point,

in the form of a covariance matrix,

the first gaussian distribution parameter is set as required, and the range of the risk distribution is larger when the numerical value is larger. Of course, the first gaussian distribution parameter may be normalized, so that the larger the distribution range is, the lower the maximum value of the risk distribution probability caused by the obstacle is. Subsequently, if not specifically stated, first to secondThe three-gauss distribution parameters can be normalized.

Fig. 5a is a schematic view of risk distributions provided in an embodiment of the present disclosure, and fig. 5a shows a risk distribution corresponding to a coordinate, that is:

and (4) risk distribution corresponding to the formula, wherein the central point is a point p'.

Fig. 5b is a schematic view of risk distributions provided in an embodiment of the present disclosure, in which fig. 5b shows, from top to bottom, risk distributions corresponding to 3 time instants in sequence, and the risk distributions at the 3 time instants are combined to form a spatio-temporal risk distribution. Wherein, due to the different confidence degrees of the different predicted motion trajectories, the probability of the risk (i.e. the peak height of the two-dimensional gaussian distribution) is not completely consistent.

In addition, the unmanned vehicle can determine the speed and the direction of the obstacle at each coordinate at each moment in a specified time period according to a plurality of predicted movement tracks of the obstacle. And then determining the risk distribution of the obstacle at each coordinate at the moment according to the probability of the obstacle appearing at the coordinate, the speed and the direction of the obstacle at the coordinate and based on the first Gaussian distribution parameter.

Specifically, the following formula can be used:

it is shown that, among others,

representing an anisometric two-dimensional gaussian distribution along the velocity direction,

representing the center of anisometric two-dimensional Gaussian distribution at point p' at time t, and risk representing the risk function positively correlated with speed, and using a formula

Meaning a function related to the velocity V of the obstacle at the position p' at time t.

Wherein, for the coordinate not on any predicted motion track, the occupation probability

At a minimum value mu, the velocity and orientation are zero. And for the coordinate corresponding to the predicted motion trail at the moment, the speed and the direction of the obstacle at the moment of the coordinate can be determined according to the predicted motion trail. For example, the moving speed of the obstacle is determined according to the distance of the unmanned vehicle moving along the predicted motion trail at other time adjacent to the time and the interval duration of the time. The orientation can be determined in the tangential direction of the predicted motion trajectory at that coordinate. Of course, if the trajectory prediction model in step S102 can directly output the speed and the direction, the speed and the direction output by the model can be directly used in step S104.

Further, since the faster the moving speed of the obstacle is, the higher the risk caused by the obstacle is, the wider the range in which the risk can be spread. Further, since the spread of the risk is also related to the orientation of the obstacle, the unmanned vehicle can determine a risk distribution that follows a gaussian distribution in the direction in which the obstacle is oriented.

Fig. 5c and 5d are schematic diagrams of risk distributions provided by an embodiment of the present specification, where fig. 5c shows an anisometric two-dimensional gaussian risk distribution whose coordinates are determined according to speed and orientation, a dark arrow indicates the speed orientation of an obstacle at the point p', a longer dark arrow indicates a higher speed, a similar arrow filled with oblique lines indicates a short axis length corresponding to the size of the obstacle, a longer arrow filled with oblique lines indicates a larger size of the obstacle, fig. 5d shows that the directions and ranges of risk distributions corresponding to different moments are not completely the same, and a dotted line is a predicted motion trajectory.

Furthermore, since the obstacle can be subdivided into a stationary obstacle and a moving obstacle according to the predicted movement trajectory of the obstacle, when determining the space-time risk distribution of each obstacle, the unmanned vehicle can also determine whether the obstacle is a stationary obstacle or not, and determine the space-time risk distribution by at least one of the two manners.

Specifically, the unmanned vehicle may determine, according to each predicted movement trajectory, a coordinate corresponding to each predicted movement trajectory at each time within the specified time period, determine, for each determined coordinate, whether the obstacle is a stationary obstacle at the time according to a speed of the obstacle at the coordinate, and if so, determine, based on the first gaussian distribution parameter, a risk distribution of the obstacle at the time at the coordinate, where the risk distribution is an equiaxial two-dimensional gaussian distribution centered on the coordinate, that is, a formula

The risk profile is shown, for a stationary obstacle, with a predicted motion trajectory at a fixed point, i.e. p'.

If not, determining the long axis direction of the anisometric two-dimensional Gaussian distribution according to the orientation of the obstacle at the coordinate, determining the long axis length of the anisometric two-dimensional Gaussian distribution according to the speed of the obstacle at the coordinate, determining the short axis length of the anisometric two-dimensional Gaussian distribution according to the size of the obstacle, determining the anisometric two-dimensional Gaussian distribution taking the coordinate as the center based on the first Gaussian distribution parameter, the determined short axis and the determined long axis, and taking the anisometric two-dimensional Gaussian distribution as the risk distribution of the obstacle at the time, wherein the larger the speed, the longer the long axis, the larger the size, the longer the short axis, namely, the formula:

the risk distribution represented. The risk distribution corresponding to the coordinate where the obstacle is located at different moments is represented by anisometric gaussian distribution, that is, for any coordinate around the unmanned vehicle, when the probability of the obstacle appearing at the coordinate is higher and the speed of the coordinate is higher, the probability of the risk distribution in the coordinate where the obstacle faces upward is higher. Wherein the static barrier does not move within the specified time period, so the determined risk scoreCloth at any one time, the center of risk distribution can be used

Indicating that moving obstacles at different times may be located at different positions, and therefore the center of the risk distribution is used

And (4) showing.

In the specification, the unmanned vehicle integrates the space-time risk distribution of each obstacle, and a formula can be used

Representing the sum of the cumulative effects of all possible obstacles p' on the risk of point p at time t.

S106: and determining total space-time risk distribution according to the determined risk distribution at each moment in the specified time period, and determining a control strategy of the unmanned vehicle according to the total space-time risk distribution.

In this specification, after the space-time risk distribution of each obstacle is determined, the unmanned vehicle may superimpose the space-time risk distribution according to the space-time risk distribution of all obstacles around the unmanned vehicle at the current time, determine the total space-time risk distribution, and finally determine the control strategy of the unmanned vehicle according to the total space-time risk distribution.

In addition, in this specification, the unmanned vehicle can determine the strategy, plan and control of unmanned vehicle driving according to the total space-time risk distribution, and in this specification, you are collectively called a control strategy. The planning may be regarded as determining a specific path planning according to the strategy, and the control is a specific output control signal determined according to the determined strategy and planning by combining hardware (such as a chassis structure, a power system and the like) of the unmanned vehicle.

Based on the unmanned vehicle control method shown in fig. 1, the position of each obstacle at the current time is determined according to the acquired environment information, a plurality of predicted movement tracks of the obstacle in a future specified time period are determined according to the historical time and the position of the obstacle at the current time, then, for each obstacle, the risk distribution corresponding to the predicted movement track of the obstacle at each time in the specified time period is determined according to each predicted movement track of the obstacle, and finally, the total space-time risk distribution is determined according to the determined risk distribution in the specified time period, so as to be used for determining the control strategy of the unmanned vehicle. Since the spatiotemporal risk distribution of the obstacle can be determined without determining the trajectory of the unmanned vehicle, the method provided by the present specification does not determine the risk on the basis of a collision, but rather determines the risk of an objective presence of the obstacle in the surroundings at the present moment on the basis of objective facts. The space-time occupation probability of the barrier is determined, and the distribution of the risk brought by the barrier on the space-time is determined, so that the subsequent determination of the unmanned vehicle control strategy is more accurate and objective, namely, only the total space-time risk distribution is considered from the aspect of driving safety. Therefore, the method provided by the specification can optimize the accuracy of determining the control strategy and improve the efficiency of determining the control strategy because the determined risk distribution is more accurate and the content is richer.

In addition, in one or more embodiments provided herein, the unmanned vehicle may also refer to other risk-causing factors in the environment surrounding the unmanned vehicle when determining the overall spatiotemporal risk distribution. The method specifically comprises the following steps: a risk distribution of risks caused by occlusion of an obstacle, a risk distribution of risks caused by traffic identification, and a risk distribution of risks caused by a traffic event.

Specifically, in step S106, the unmanned vehicle may further perform the following steps:

s1061: and determining the areas of laser radar information loss caused by each obstacle at the current time as the occlusion areas at the current time according to the laser radar information, and determining the boundaries of the occlusion areas.

In this specification, since the environment information acquired by the unmanned vehicle depends on its own sensor, such as the image and the laser radar information described in step S100, these pieces of information are generally acquired by using the propagation characteristics of the wave, and the wave on which the information is based is not a long wave, and therefore, in principle, the information is easily blocked. For example, the image sensor cannot bypass the obstruction to acquire and display information behind the obstruction, the laser emitted by the laser radar is difficult to return information behind the obstruction, and the millimeter wave radar (used for distance measurement) which is common in the field of unmanned vehicles cannot return information behind the obstruction. Thus, since it is not possible to determine whether or not there is a risk in the occlusion region, the boundary of the occlusion region may be considered to be at risk, and a spatiotemporal risk distribution corresponding to the boundary of the occlusion region may be determined.

Specifically, the unmanned vehicle can determine the area where the laser radar information is lost due to each obstacle at the current time according to the acquired laser radar information, and the area is used as the shielding area at the current time, determines the boundary of each shielding area, and determines the coordinate corresponding to the boundary. Fig. 6 is a schematic diagram of a blocking boundary provided in an embodiment of this specification, and fig. 6 is a schematic diagram of an image formed by laser point cloud information, where a vehicle in the center is an unmanned vehicle, vehicles around the unmanned vehicle are obstacles, a curve and a circle represent laser signals that can be received by a laser radar of the unmanned vehicle, an area without the curve is an area where laser radar information is missing due to the obstacle, and a thick line represents a boundary of the determined blocking area.

Moreover, since the range that can be detected by the laser radar is limited, and it is also shown in step S100 of the present application that the unmanned vehicle can acquire environmental information (including laser radar information) within a preset range around the unmanned vehicle, the boundary of the occlusion region is generally composed of the boundary shown by the thick line in fig. 6 and the boundary of the preset range. However, since the boundary of the prediction range is limited in advance, it is an area where there is no objective risk, and therefore, in the present specification, only the boundary of the blocked area caused by blocking of an obstacle is used as a boundary that may cause a risk, and the risk distribution is determined.

The unmanned vehicle can determine the coordinates corresponding to each boundary after determining the boundary of each shielding area, so that the risk distribution brought by each boundary can be determined subsequently.

In addition, in this specification, since the occlusion region may change as the obstacle moves for the moving obstacle, the occlusion regions may not completely coincide at different times within a specified time period, that is, the occlusion region may change, and the boundary of the occlusion region may also change. Therefore, in this specification, the unmanned vehicle may determine, according to each predicted motion trajectory of each obstacle determined in step S102, the boundary of the occlusion region at different times within the specified time period and the probability of existence of the boundary of the occlusion region, and determine the boundary of the occlusion region at each time, which will be described in detail later in this specification. Alternatively, the unmanned vehicle may determine only the boundary of the occlusion region at the current time in order to simplify the calculation.

S1062: and determining the risk distribution corresponding to each boundary at the current moment according to the determined coordinates of each boundary and a preset second Gaussian distribution parameter.

In this specification, after the boundary of each occlusion region is determined, the risk distribution corresponding to the coordinate of each edge boundary at the current moment can be determined according to the coordinate of each edge boundary and the preset second gaussian distribution parameter.

Specifically, since the boundary of the occlusion region exists at the current moment, there is no probability distribution, and thus a formula can be adopted

Indicating that point p' at time t is the boundary of the occlusion region,

the coordinates of the boundary of the occlusion region determined in step S106 are defined by the coordinates of the other non-boundary of the region around the unmanned vehicle, indicating that the point p' at time t is not the boundary of the occlusion region.

Then, the formula is adopted

To represent a risk distribution caused by the boundary of the occlusion region, wherein,

which is a second gaussian parameter, the larger the value the larger the range of the risk distribution,

the risk probability is preset and can be specifically set according to needs. In this specification, the higher the probability of risk, the greater the risk. For example,

if the value is greater than 1, the risk caused by the boundary of the occlusion area is greater than the risk caused by the static obstacle.

Refer to the formula of risk distribution due to static obstacle in step S104

The coordinates on the visible boundary are similar in form to the risk distribution caused by the stationary obstacle, except that the risk probability of the stationary obstacle is 1 in the formula, which is of course a preset value. In the boundary-induced risk distribution, the risk probability QUOTE

Either equal to 1 or not equal to 1, and, for the same reason,

can be equal to

Or may not be equal to

. If the risk probability and the gaussian parameter are the same, each coordinate on the boundary can be regarded as a static obstacle, and the risk distribution brought by the static obstacle is equal to that of the static obstacle.

Based on the schematic boundary diagram of the occlusion area shown in fig. 6, the present specification further provides a schematic risk distribution diagram caused by the boundary shown in fig. 7, where the left vehicle is an unmanned vehicle, the right vehicle is an obstacle, the thick line is a determined boundary, the risk distribution corresponding to the visible boundary is similar to two "beams," a "wall" with high risk distribution is formed, and the control strategy for selecting a passing boundary can be reduced in the subsequent steps of the unmanned vehicle.

S1063: and determining traffic identification around the unmanned vehicle according to the environment information, and determining risk distribution corresponding to the traffic identification in the appointed time period according to a risk area corresponding to the preset traffic identification and a preset risk probability.

In this specification, since the traffic environment is generally complex, besides the risk caused by the obstruction of the obstacle and the obstacle, the risk may be caused by the traffic sign and the traffic event, for example, the traffic sign is a speed limit sign, a sign with a school in front, etc., and the risk may be caused by the behavior change of the traffic participant, and the same risk may be caused by the traffic event.

In order to provide traffic identification and space-time risk distribution caused by traffic events based on Bayesian theory, so that the space-time risk distribution caused by various factors can be fused into uniform risk distribution for subsequent determination of the control strategy of the unmanned vehicle, the unmanned vehicle in the specification can determine surrounding traffic identification according to environmental information, and determine the risk distribution corresponding to the traffic identification in a specified time period according to a preset risk area corresponding to the traffic identification and a preset risk probability.

Specifically, the coordinates and the traffic signs of all traffic signs around the unmanned vehicle are determined according to the environment information.

In this specification, the unmanned vehicle may recognize the traffic sign first and then determine the coordinates of the traffic sign. Since the current identification of the traffic sign is a relatively mature technology, the description of how to identify the traffic sign and the type of the traffic sign through the image is not repeated, as long as the part of the image is the traffic sign and the coordinates of the traffic sign can be identified.

It should be noted that the determined coordinates of the traffic identifier may be determined according to the image and the laser radar information, that is, the image identifies the feature points of the traffic identifier, matches the feature points with the laser point cloud information, and determines the coordinates corresponding to the feature points of the traffic identifier as the coordinates of the traffic identifier.

Of course, if one traffic sign corresponds to a plurality of feature points, the center point coordinates of the coordinates corresponding to the plurality of feature points may be used as the coordinates of the traffic sign. In addition, the coordinates corresponding to the feature points and the traffic identification coordinates may not include coordinates of height dimension, that is, two-dimensional coordinates on the ground.

And then, determining the risk area corresponding to the traffic identification according to the shape, the size and the relative position of the risk area corresponding to the preset traffic identification and the coordinates of the traffic identification. Wherein, the shape, the size and the relative position of the coordinate with the traffic sign of the risk area corresponding to the traffic sign of different types are not identical. The traffic sign comprises a traffic sign and a traffic marking.

In this specification, the risk areas corresponding to different traffic signs may be predetermined. Since it is usually seen that the driver of the deceleration sign slows down, and the speed change of the vehicle easily causes an accident and causes a risk, the road sections before and after the traffic sign are the risk areas corresponding to the traffic sign. For example, the zebra crossing indicates that the pedestrian has the priority to pass through, and the motor vehicle needs to slow down to yield, so that the zebra crossing itself and the road section in front of the zebra crossing are the risk areas corresponding to the traffic marking.

Of course, since the relative position relationship between different traffic signs and the risk area may be different, the risk area corresponding to the traffic sign in this specification includes the shape and size of the risk area, and also includes the relative position relationship between the risk area and the coordinates of the traffic sign.

In addition, in the present specification, a formula may be adopted

Representing the risk distribution caused by the traffic sign,the risk probability can be defined as a constant value between 0 and 1 in a fixed area, namely alpha in the formula. Zone represents the corresponding coordinate of the risk area, that is, when the p coordinate is located in the risk area, the risk of the coordinate is α. Only the risk distribution within the risk area is a fixed value, while the risk distribution within the non-risk area is 0.

S1064: and determining the traffic events around the unmanned vehicle according to the environment information, and determining the risk distribution corresponding to the traffic events in the specified time period according to the preset corresponding risk time of the traffic events and the third Gaussian distribution parameter.

In this specification, based on the definition in the standard GB/T28789-2012 video traffic detector, a traffic event is an abnormal traffic condition and behavior that occur on a road and affect the traffic safety and traffic passing, and mainly refers to typical traffic event types such as a stop event, a reverse-driving event, a pedestrian event, a spill event, a congestion event, and a motor vehicle driving-off event. It can be seen that various types of traffic events are also one of the risk-causing factors, and therefore the distribution of risk caused by traffic events may also be considered in this description,

specifically, the risk may be caused immediately after the occurrence of the traffic event, or may be caused after a time delay, for example, after the side parking event occurs, the driver may need to open the door and get off the vehicle for a while, and this action of opening the door and getting off the vehicle may cause the risk, or in the event of a spill, the risk may be caused immediately after the vehicle throws the object. Because the time points of risks caused by different traffic events are different, after the unmanned vehicle determines the surrounding traffic events according to the environmental information acquired at the current time, the risk distribution corresponding to the traffic events in the specified time period can be determined according to the preset risk time corresponding to the traffic events and the third Gaussian distribution parameter.

Specifically, in this specification, the unmanned vehicle may determine a traffic event occurring at the current time and an occurring coordinate according to the environmental information acquired at the current time. When the traffic event occurring at the current moment is determined, the traffic event is determined to be of what type. In the present specification, the type of the traffic event specifically includes what can be set as required, and the present specification is not limited and can be subdivided as required. For example, stop events such as side parking, garage dumping parking, etc. are subdivided.

And secondly, aiming at each determined traffic event, determining the risk time of the traffic event occurring at the current time according to the preset corresponding relation between each traffic event and the risk time. In this specification, the risk time in the correspondence relationship may be determined by statistical calculation according to data of different types of traffic events mined from big data and risk times causing risks.

For example, for a side parking event, large data mining can be performed, a large number of side parking events are acquired through a traffic monitoring video, the time difference between the parking time and the time for a driver to open a door and get off the vehicle is determined, a statistical value (such as average time length) from the parking time to the door opening time is determined through statistical calculation, and the risk time corresponding to the side parking event is determined according to the average time length. Similarly, the corresponding relation between each traffic event and the risk time can be determined by counting the risk time corresponding to other types of traffic events.

Then, the unmanned vehicle can determine the risk distribution corresponding to each traffic event in the specified time period according to the historically determined risk time of other traffic events, the coordinates of the other traffic events, the risk time of the traffic event occurring at the current time, the coordinates of the traffic event occurring at the current time and the third gaussian distribution parameter.

That is, if the historical time also identifies the spatiotemporal risk distributions corresponding to several traffic events, then the spatiotemporal risk distributions of the traffic events in the historical time are also considered when identifying the risk distributions corresponding to the traffic events in the specified time period at the current time.

Specifically, each traffic event determined by the unmanned vehicle for the current moment can be determined according to a formula,

determining a spatiotemporal risk distribution corresponding to the traffic event, wherein

In order to preset the risk probability,

is the third gaussian parameter, and is,

the risk time is e, the coordinate of the occurrence of the traffic event, and the risk distribution caused by the traffic event at other times is 0. And determining the space-time risk distribution brought by the total traffic events by combining the space-time risk distribution of the traffic events determined at the historical moment and the space-time risk distribution of each traffic event determined at the current moment.

S1065: and determining total space-time risk distribution according to the determined risk distribution corresponding to each predicted movement track of the barrier in the specified time, the determined risk distribution corresponding to the traffic identification, the determined risk distribution corresponding to the traffic incident and the determined risk distribution corresponding to each boundary at the current moment.

And finally, determining the total space-time risk distribution by the unmanned vehicle according to various risk distributions within the specified time determined in the previous step, and determining the control strategy of the unmanned vehicle according to the total space-time risk distribution. Specifically, the formula is used for expressing as:

。

of course, the unmanned vehicles can also determine the total space-time risk distribution in a weighted summation mode, and the weighted values can be set according to needs. Of course, since the risk is caused by objective factors, the size and the range of the risk caused by the objective factors are considered when determining the risk distribution of different objective factors in each step, and therefore, the risk distribution can be directly superposed.

It should be noted that the risk distribution corresponding to the predicted movement trajectory of the obstacle, the risk distribution corresponding to the boundary, the risk distribution corresponding to the traffic sign, and the risk distribution corresponding to the traffic event, which are described in this specification, may be understood as the distribution probability of the risk caused by the objective factor. The risk distribution includes both the high and low risk probabilities and the range of risk distributions caused by objective factors.

Based on the unmanned vehicle control method shown in the steps S1061 to S1065, the unmanned vehicle determines the position of each obstacle at the current time according to the acquired environment information, determines a plurality of predicted movement trajectories of the obstacles in a specified time period in the future according to the historical time and the position of the obstacle at the current time, then determines the risk distribution corresponding to the predicted movement trajectories of the obstacles in the specified time period according to the predicted movement trajectories of the obstacles, determines the boundary of the sheltered area at the current time according to the laser radar information, determines the risk distribution corresponding to each boundary at the current time, determines the risk distribution corresponding to the traffic signs at each time and the risk distribution corresponding to the traffic events at each time in the specified time period according to the environment information, and finally determines the various types of risk distributions in the specified time period according to the determined risk distributions, an overall spatiotemporal risk distribution is determined for use in determining a control strategy for the unmanned vehicle. Since the spatiotemporal risk distribution of the obstacle can be determined without determining the driving trajectory of the unmanned vehicle, the method provided by the present specification determines the risk objectively in the surrounding environment at the present time, not based on collision determination, but on objective facts. The method comprises the steps of determining the space-time occupation probability of the barrier, determining the space-time distribution of risks brought by the barrier, and determining the space-time distribution of risks brought by other objective factors influencing the driving safety by adopting the same idea, wherein the space-time risk distribution is more favorable for determining the control strategy of the unmanned vehicle within a period of time, and the factors influencing the driving safety are uniformly embodied by the total space-time risk distribution, so that the subsequent determination of the control strategy of the unmanned vehicle is simpler, namely, only the total space-time risk distribution needs to be considered from the aspect of the driving safety. Therefore, the method provided by the specification can optimize the accuracy of determining the control strategy and improve the efficiency of determining the control strategy because the determined risk distribution is more accurate and the content is richer.

In addition, when the risk distribution corresponding to each predicted movement trajectory of the obstacle at each time within the specified time period is determined in step S104, the risk distribution may not be additionally calculated for the coordinates where the predicted movement trajectory does not pass, or may be determined according to a preset background risk. The distribution range of the preset background risk distribution may also be in accordance with two-dimensional gaussian distribution, and the maximum risk probability is usually much smaller than the maximum risk probability corresponding to the various objective factors described above, which may be specifically set as required, and the present application is not limited.

Further, since the vehicle does not pose a risk to the rear of the vehicle in the traveling direction when the vehicle travels, the risk distribution exists only in the front of the traveling direction (i.e., the vehicle heading direction), and there is no risk directly behind the vehicle, and therefore, the risk distribution corresponding to each coordinate on the predicted movement locus of the moving obstacle is described using a gaussian distribution cut in the long axis direction, as shown in fig. 8.

Furthermore, in the present specification, for moving obstacles, in order to avoid the variation of the "peak height" in the risk distribution due to the risk function, and the difference in the maximum risk due to the difference in the velocity at the position of the same occupation probability, it is generally required that the risk function is selected to satisfy the condition that at a certain time t,

wherein one option is

。

It should be noted that the risk distribution formula of the static obstacle may be

When is coming into contact with

The time is the risk distribution formula of the stationary obstacle in step S104.

In addition, in step S1061 of this specification, when the unmanned vehicle needs to determine the risk distribution corresponding to the boundary of the occlusion area at each time in the predetermined time period, the probability of occurrence of the obstacle at each position at each time may be determined for each obstacle based on the predicted motion trajectory of the obstacle, the occlusion area occluded by the obstacle when the obstacle occurs at the position may be determined for each position, and the risk distribution corresponding to the boundary of the occlusion area occluded by the obstacle may be determined based on the probability of occurrence of the obstacle at the position.

Further, in step S1063, since the traffic sign is usually fixed and thus does not change with time, the risk distribution of the traffic sign at the current time is determined, which is equivalent to determining the risk distribution of the traffic sign in the specified time period, so as to determine the spatiotemporal risk distribution of the traffic sign.

Further, in step S1064, since the risk time may be determined by statistical calculation, the risk time distribution corresponding to each type of event may also be determined. For example, the statistical distribution of the time from parking to door opening is determined by statistical calculation and is used as the risk time distribution corresponding to the side parking event.

In addition, the unmanned vehicle may be used for unmanned distribution in the present specification, and the control method provided in the present specification may be particularly applied to the field of distribution using an unmanned vehicle, and when the unmanned vehicle is distributed, the control method may determine surrounding environment information, determine a total spatiotemporal risk distribution caused by various objective factors according to the environment information, and determine a control strategy for the unmanned vehicle based on the total spatiotemporal risk distribution. The method can be particularly used in the delivery scene of express delivery, takeaway and the like by using the unmanned vehicle.

Further, in this specification, the main body of the unmanned vehicle Control process is not limited, and the robot having a moving function may determine a Control strategy by executing the unmanned vehicle Control process during the moving process, or the vehicle having an unmanned function may plan a moving state by executing the unmanned vehicle Control process when the unmanned function is implemented, or may plan a moving state by the process when the intelligent vehicle controls a moving state of the vehicle, for example, when the vehicle uses an Adaptive Cruise Control (ACC) function, the moving state of the vehicle is controlled by the moving state planning process, and the like. That is, the unmanned vehicle motion state planning process provided by the present specification is as follows: the process of planning the motion state of the equipment which can automatically control the motion state of the equipment. Therefore, the main body for executing the motion state planning process in this specification may be an unmanned vehicle, a robot, an intelligent vehicle, a vehicle with an unmanned assumed function, or the like, and this specification does not limit this.

Based on the unmanned vehicle control process shown in fig. 1, the embodiment of the present specification further provides a schematic structural diagram of the unmanned vehicle control device, as shown in fig. 9.

Fig. 9 is a schematic structural diagram of an unmanned vehicle control device provided in an embodiment of the present specification, where the device includes:

the obstacle determining module 200 is used for determining the position of each obstacle at the current moment according to the acquired environment information around the unmanned vehicle, wherein the environment information at least comprises an image and laser radar information;

the track prediction module 202 is used for inputting the historical time position and the current time position of each obstacle, inputting a pre-trained track prediction model, and respectively determining a plurality of predicted motion tracks of each obstacle in a specified time period in the future;

the obstacle risk determining module 204 is used for determining risk distribution corresponding to each predicted movement track of each obstacle at each moment in the specified time period according to the determined plurality of predicted movement tracks of the obstacle and a preset first Gaussian distribution parameter for each obstacle;

and the control strategy module 206 determines total space-time risk distribution according to the determined risk distribution at each moment in the specified time period, and determines the control strategy of the unmanned vehicle according to the total space-time risk distribution.

Alternatively, the control strategy module 206 may be composed of an occlusion information determination sub-module 2061, an occlusion risk determination sub-module 2062, an identification risk determination sub-module 2063, an event risk determination sub-module 2064, and a comprehensive control strategy sub-module 2065, as shown in fig. 10.

The occlusion information determining submodule 2061 determines, according to the laser radar information, an area where the laser radar information is missing due to each obstacle at the current time, as an occlusion area at the current time, and determines a boundary of each occlusion area;

the occlusion risk determining submodule 2062 determines risk distribution corresponding to each boundary at the current moment according to the determined coordinates of each boundary and a preset second gaussian distribution parameter;

the identification risk determining submodule 2063 is configured to determine the traffic identification around the unmanned vehicle according to the environment information, and determine the risk distribution corresponding to the traffic identification in the specified time period according to the risk area corresponding to the preset traffic identification and the preset risk probability;

the event risk determining submodule 2064 is configured to determine a traffic event around the unmanned vehicle according to the environment information, and determine a risk distribution corresponding to the traffic event in the specified time period according to a preset risk time corresponding to the traffic event and a third gaussian distribution parameter;

the comprehensive control strategy submodule 2065 determines a total spatiotemporal risk distribution according to the determined risk distribution corresponding to each predicted movement trajectory of the obstacle within the specified time, the determined risk distribution corresponding to the traffic identifier, the determined risk distribution corresponding to the traffic event, and the determined risk distribution corresponding to each boundary at the current time.

Optionally, the obstacle risk determining module 204 determines, according to the predicted motion trajectories, a probability of occurrence of the obstacle at each coordinate at each time in the specified time period as a space-time occupancy probability distribution of the obstacle, where the probability of occurrence of the obstacle at each coordinate at any time in the specified time period is normalized, and determines a risk distribution of the obstacle at each coordinate at each time in the specified time period according to the determined space-time occupancy probability distribution of the obstacle and the preset first gaussian distribution parameter, where, for each coordinate, the higher the probability of occurrence of the obstacle at the coordinate is, the higher the probability of occurrence of the coordinate is caused by the obstacle is.

Optionally, the obstacle risk determining module 204 determines, for each time within the specified time period, a speed and an orientation of the obstacle at each coordinate at the time according to a number of predicted motion trajectories of the obstacle, and determines, for each coordinate at the time, a risk distribution of the obstacle at the time at the coordinate based on the first gaussian distribution parameter according to a probability of the obstacle appearing at the coordinate, the speed and the orientation of the obstacle at the coordinate.

Optionally, the obstacle risk determining module 204 determines whether the obstacle is a stationary obstacle at the time according to the speed of the obstacle at the coordinate, if so, determines the risk distribution of the obstacle at the time at the coordinate based on the first gaussian distribution parameter, if not, determines the long axis direction of the anisometric two-dimensional gaussian distribution according to the orientation of the obstacle at the coordinate, determines the long axis length of the anisometric two-dimensional gaussian distribution according to the speed of the obstacle at the coordinate, determines the short axis length of the anisometric two-dimensional gaussian distribution according to the size of the obstacle, determines the anisometric two-dimensional gaussian distribution centered on the coordinate based on the first gaussian distribution parameter, the determined short axis and the determined long axis, and determines the anisometric two-dimensional gaussian distribution as the risk distribution of the obstacle at the time at the coordinate, wherein the greater the velocity, the longer the major axis, the greater the dimension, the longer the minor axis.

Optionally, the obstacle risk determining module 204 is configured to, when the obstacle is not a stationary obstacle, determine, according to the orientation of the obstacle at the coordinate, a gaussian distribution of one side of the orientation in an anisometric two-dimensional gaussian distribution determined with the coordinate as a center as the risk distribution of the obstacle at the coordinate at the time.

Optionally, the identifier risk determining sub-module 2063 determines the coordinates of each traffic identifier and the traffic identifier around the unmanned vehicle according to the environment information, and determines, for each traffic identifier, the risk area corresponding to the traffic identifier according to the shape, the size, the relative position of the risk area corresponding to the preset traffic identifier and the coordinates of the traffic identifier, where the shape, the size, and the relative position of the risk area corresponding to different types of traffic identifiers are not exactly the same as those of the coordinates of the traffic identifier, and the traffic identifier includes a traffic sign and a traffic marking.

Optionally, the event risk determining sub-module 2064 determines the traffic event and the occurring coordinate at the current time according to the environment information, determines the risk time of the traffic event occurring at the current time according to a preset corresponding relationship between each traffic event and the risk time, and determines the risk distribution corresponding to each traffic event in the specified time period according to the historically determined risk time of other traffic events, the coordinates of the other traffic events, the risk time of the traffic event occurring at the current time, the coordinates of the traffic event occurring at the current time, and the third gaussian distribution parameter.

The present specification also provides a computer-readable storage medium storing a computer program, which is operable to execute any one of the above-described unmanned vehicle control methods.

Based on the unmanned vehicle control process provided in fig. 1, the embodiment of the present specification further provides a schematic structural diagram of the unmanned vehicle shown in fig. 11. As shown in fig. 11, on a hardware level, the unmanned vehicle includes a processor, an internal bus, a network interface, a memory, and a nonvolatile memory, and the processor reads a corresponding computer program from the nonvolatile memory into the memory and then runs the computer program to implement any one of the above-mentioned unmanned vehicle control methods. Of course, the unmanned vehicle can also comprise hardware required for realizing other functions of the unmanned vehicle. For example, a moving device required for the unmanned vehicle to move, a communication device required for communication, an electronic device that collects surrounding environment information, and the like.

Of course, besides the software implementation, the present specification does not exclude other implementations, such as logic devices or a combination of software and hardware, and the like, that is, the execution subject of the following processing flow is not limited to each logic unit, and may be hardware or logic devices.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Language Description Language), traffic, pl (core unified Programming Language), HDCal, JHDL (Java Hardware Description Language), langue, Lola, HDL, laspam, hardsradware (Hardware Description Language), vhjhd (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functions of the various elements may be implemented in the same one or more software and/or hardware implementations of the present description.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.

Claims

1. An unmanned vehicle control method, comprising:

determining risk distribution corresponding to each predicted motion track of the obstacle at each moment in the specified time period according to the determined plurality of predicted motion tracks of the obstacle and a preset first Gaussian distribution parameter for each obstacle, wherein the probability of the obstacle appearing at each coordinate at any moment in the specified time period is normalized;

determining total space-time risk distribution according to the determined risk distribution at each moment in the specified time period, and determining a control strategy of the unmanned vehicle according to the total space-time risk distribution, wherein the risk distribution at each moment in the specified time period further comprises risk distribution corresponding to each boundary at the current moment, and the risk distribution corresponding to each boundary is determined by adopting the following method:

and determining the risk distribution corresponding to each boundary at the current moment according to the determined coordinates of each boundary and a preset second Gaussian distribution parameter.

2. The method of claim 1, wherein determining a total spatiotemporal risk distribution based on the determined risk distributions at the respective times within the specified time period comprises:

and determining total space-time risk distribution according to the determined risk distribution corresponding to each predicted movement track of the barrier in the specified time, the determined risk distribution corresponding to the boundary, the determined risk distribution corresponding to the traffic identification and the determined risk distribution corresponding to the traffic incident.

3. The method according to claim 1, wherein determining, at each time within the specified time period, a risk distribution of the obstacle corresponding to each predicted motion trajectory according to the determined predicted motion trajectories of the obstacle and a preset first gaussian distribution parameter, specifically comprises:

4. The method according to claim 3, wherein determining the risk distribution of the obstacle at each coordinate at each time within the specified time period according to the determined space-time occupancy probability distribution of the obstacle and the preset first gaussian distribution parameter includes:

5. The method of claim 4, wherein determining the risk distribution of the obstacle at the coordinate at the time based on the first Gaussian distribution parameter according to the probability of the obstacle appearing at the coordinate, the speed and the orientation of the obstacle at the coordinate comprises:

6. The method of claim 5, wherein the method further comprises:

7. The method according to claim 2, wherein determining traffic signs around the unmanned vehicle according to the environment information, and determining risk distributions corresponding to the traffic signs in the specified time period according to a risk area corresponding to a preset traffic sign and a preset risk probability specifically include:

8. The method of claim 2, wherein determining the risk distribution corresponding to the traffic event in the specified time period according to a preset risk time corresponding to the traffic event and a third gaussian distribution parameter specifically comprises:

9. An unmanned vehicle control device, characterized by comprising:

the obstacle risk determination module is used for determining risk distribution corresponding to each obstacle in each predicted motion track at each moment in the specified time period according to the determined plurality of predicted motion tracks of the obstacle and a preset first Gaussian distribution parameter, wherein the probability of the obstacle appearing in each coordinate at any moment in the specified time period is normalized;

the control strategy module determines total space-time risk distribution according to the determined risk distribution at each moment in the specified time period, and determines the control strategy of the unmanned vehicle according to the total space-time risk distribution, wherein the risk distribution at each moment in the specified time period further comprises risk distribution corresponding to each boundary at the current moment, and the risk distribution corresponding to each boundary is determined by adopting the following method: and determining the areas of laser radar information loss caused by the obstacles at the current moment as the sheltered areas at the current moment, determining the boundaries of the sheltered areas, and determining the risk distribution corresponding to the boundaries at the current moment according to the determined coordinates of the boundaries and the preset second Gaussian distribution parameters.

10. A computer-readable storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method of any of the preceding claims 1-8.

11. An unmanned vehicle comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the method of any of claims 1-8.