CN113306552B

CN113306552B - Ultra-low-speed creeping method for driverless cars in mixed road congestion

Info

Publication number: CN113306552B
Application number: CN202110876487.XA
Authority: CN
Inventors: 杨福刚; 陈子龙; 曾令洲; 胡声洋; 廖文俊; 李平飞; 牟森; 肖丰
Original assignee: Xihua University
Current assignee: Chengdu Greenhill Transportation Technology Co ltd
Priority date: 2021-07-31
Filing date: 2021-07-31
Publication date: 2021-10-01
Anticipated expiration: 2041-07-31
Also published as: CN113306552A

Abstract

The invention belongs to the technical field of unmanned vehicles, and particularly relates to an ultra-low speed creeping method of an unmanned vehicle under a mixed road congestion state. The specific technical scheme is as follows: analyzing the action tracks of pedestrians and non-motor vehicles in the range M around the vehicle pre-envelope area, and indicating whether the vehicle enters parking waiting, decelerating or running at original speed according to the distance between the front end of the vehicle and the pedestrians and the non-motor vehicles, whether the pre-action tracks of the pedestrians and the non-motor vehicles deviate from the vehicle pre-running track, and whether the pedestrians see the vehicle; the method comprehensively considers various factors, can ensure that pedestrians and non-motor vehicles can pass safely, can also ensure that vehicles keep moving ahead, avoids the situation that the vehicles cannot continue to run due to the principle of pedestrian safety, and greatly relieves the traffic jam condition when the flow of people is large in a mixed road.

Description

Ultra-low speed creeping method of unmanned vehicle under mixed road congestion state

Technical Field

The invention belongs to the technical field of unmanned vehicles, and particularly relates to an ultra-low speed creeping method of an unmanned vehicle under a mixed road congestion state.

Background

The urban traffic mixed road refers to the mixed running of motor vehicles, non-motor vehicles and pedestrians on the same road. Due to the low standard of the highway and the narrow road surface, the highway has no separation zone, and has no separation between a lane and a sidewalk, and between a fast vehicle and a slow vehicle. Under the mixed transportation mode, traffic accidents between people and vehicles are easy to occur, and the speed and the efficiency of all aspects are directly influenced. For example, when a car enters an underground parking lot of a shopping mall, the car is firstly transferred from a motor vehicle lane to an entrance of the underground parking lot, but the entrance of the parking lot is positioned in a sidewalk, so that the car entering the parking lot must pass through a section of mixed road, and the car and pedestrians pass through the mixed road at any time; there is a pedestrian path in front of the road, but there is no traffic light at the pedestrian path.

At present, drivers mostly adopt extremely slow speed to crawl under the condition of meeting the above conditions, and observe the walking track of pedestrians at any time, when the drivers observe that the pedestrians see the vehicle and generate an avoiding action, the drivers drive the vehicles to crawl continuously, the pedestrians avoid the vehicle to pass, and meanwhile, the pedestrians pay attention to the walking track of the pedestrians at any time during the passing period, so that the safety of the pedestrians is ensured; and when the driver observes that the pedestrian does not avoid on the way of the vehicle, the vehicle stops immediately, and the vehicle can pass through the driving mode under the condition of ensuring the safety of the pedestrian and the condition of incapability of driving the vehicle when the pedestrian flow is large.

Under the conditions, when the pedestrian flow is less, the unmanned automobile can better avoid pedestrians to drive into a parking lot or wait for the pedestrians to walk through a pedestrian road and then continue to drive, but when the pedestrian flow is greater, the unmanned automobile often stops according to the pedestrian safety principle and cannot continue to drive, and at the moment, serious vehicle congestion can be caused. Therefore, a driving method of an unmanned vehicle in a mixed road congestion state is needed.

Disclosure of Invention

In order to solve the technical problems, the invention aims to provide an ultra-low speed creeping method for an unmanned automobile in a mixed road congestion state, which ensures the safety and improves the passing performance of the automobile.

In order to achieve the purpose of the invention, the technical scheme adopted by the invention is as follows: the ultra-low speed creeping method of the unmanned vehicle under the mixed road congestion state analyzes the action tracks of pedestrians and non-motor vehicles in the range M around the vehicle pre-envelope area, and indicates the vehicle to enter the parking waiting, the deceleration running or the original speed running according to the distance between the front end of the vehicle and the pedestrians and the non-motor vehicles, whether the pre-action tracks of the pedestrians and the non-motor vehicles deviate from the pre-running track of the vehicle or not, and whether the pedestrians see the vehicle or not;

the vehicle pre-enveloping area is an area formed by a vehicle body contour passing through a vehicle pre-running track; the range M is a rectangular area formed after the vehicle pre-enveloping area extends forwards and backwards by the distance a respectively.

Preferably, when the running speed V1 of the unmanned automobile is lower than 10km/h, the unmanned automobile enters an ultra-low speed crawling mode; let V1 be the first travel speed of the vehicle.

Preferably, the vehicle pre-travel track is an estimated travel route of the centroid or the geometric center of the vehicle within the vehicle pre-travel time T calculated according to the vehicle first travel speed V1; the length S of the pre-running track of the vehicle is 1-4 m.

Preferably, the pre-movement tracks of the pedestrians and the non-motor vehicles are predicted movement tracks formed by the pedestrians and the non-motor vehicles under the current speed and direction of the vehicles within the pre-movement time T of the vehicles.

Preferably, the method comprises the following steps,

a1, drawing a vehicle pre-enveloping area and a vehicle pre-enveloping range M, calculating the vehicle pre-running time T = S/V1, and analyzing pre-running tracks of pedestrians and non-motor vehicles;

a2, judging the sum N of the number of pedestrians and non-motor vehicles in the range M in real time, and entering the step A7 when N is smaller than a first threshold, wherein the value range of the first threshold is 2-5; when N is greater than a first threshold, determining whether the minimum distance L is greater than a first safe distance b1, if so, entering A3, and if not, entering a 4; l is the minimum distance between the front end of the vehicle and the pedestrian or the non-motorized vehicle in the range M;

a3, decelerating the vehicle to V2 and continuing to drive forwards, and entering A4 when the minimum distance L is less than the first safe distance b 1; v2 is the second running speed of the vehicle;

a4, judging whether the pre-running track of the pedestrian or the non-motor vehicle closest to the front end of the vehicle in the range M deviates from the running direction of the current running speed V of the vehicle, if so, entering A6, and if not, entering A5;

a5, parking, waiting, and monitoring in real time, entering A6 when the pre-running track of the pedestrian or the non-motor vehicle closest to the front end of the vehicle in the range M deviates from the pre-running track of the vehicle, and returning to A3 when the minimum distance L is greater than the first safety distance b 1;

a6, decelerating the vehicle to V3 and continuing to drive forwards until the minimum distance L is smaller than a second safety distance b2, stopping and waiting until the minimum distance L is larger than a second safety distance b2, and returning to the step A4; the V3 is the third running speed of the vehicle;

and A7, the vehicle continues to run, the value N in the range M is monitored in real time, and when the value N is larger than the first threshold value, the step A2 is returned.

Preferably, in the step a6, when the minimum distance L is less than the second safety distance b2, it is determined whether the pedestrian is a non-motor vehicle in front of the vehicle, and if the pedestrian is a non-motor vehicle, the vehicle is parked and waits until the minimum distance L is greater than the second safety distance b2, the process returns to the step a 4;

if the pedestrian is a pedestrian, judging whether the facing direction of the eyes of the pedestrian can see the vehicle or not, and if the pedestrian cannot see the vehicle, stopping the vehicle and waiting; if the vehicle can be seen, the vehicle decelerates to the speed of V4 and continues to run until the minimum distance L is less than the third safe distance b3, and the vehicle stops and waits; if the eyes of the pedestrian cannot be captured, stopping the vehicle for waiting; when the minimum distance L is larger than b2, returning to the step A4; the V4 is the fourth traveling speed of the vehicle.

Preferably, the safety distances b1, b2 and b3 have the size relationship of b1 > b2 > b 3.

Preferably, the vehicle running speeds V1, V2, V3 and V4 have magnitude relations of V1 > V2 > V3 > V4;

the V1 is 8-12 km/h, the V2 is 5-7 km/h, the V3 is 3-5 km/h, and the V4 is 0-3 km/h.

Preferably, the fourth traveling speed V4= V3 α, and α is an avoidance coefficient and has a value of 0.1 to 0.9.

Preferably, in the steps A3 to a6, the sum N of the numbers of pedestrians and non-motor vehicles in the range M is monitored in real time, and when N is smaller than the first threshold, the process directly proceeds to the step a 7.

Compared with the prior art, the invention has the following beneficial effects:

when the vehicle enters a mixed road, the vehicle runs at low speed and has more pedestrians and non-motor vehicles, the vehicle enters an ultra-low-speed crawling mode, and the vehicle is indicated to enter parking waiting, decelerating running or original-speed running according to the distance between the front end of the vehicle and the pedestrians and the non-motor vehicles, whether the action tracks of the pedestrians and the non-motor vehicles deviate from the pre-running track of the vehicle and whether the pedestrians see the vehicle.

Drawings

FIG. 1 is a flow chart of an ultra-low speed crawling method for an unmanned vehicle under a mixed road congestion state according to the present invention;

FIG. 2 is a schematic view of a hybrid road of the present invention driving a vehicle through a greater pedestrian volume;

FIG. 3 is a schematic diagram of a driving vehicle environment recognition module recognizing pedestrians and non-motor vehicles according to the present invention;

FIG. 4 is a schematic diagram of a vehicle pre-envelope and range M of the present invention;

FIG. 5 is a schematic diagram of a hybrid road state according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of a step A4 of determining a hybrid road status according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of the second step A6 of determining the process hybrid road status according to the embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Unless otherwise specified, the technical means used in the examples are conventional means well known to those skilled in the art.

It should be noted that, in the technical field, the running speed range of the vehicle for the ultra-low speed creeping is 2-12 km/h, the situation that the vehicle is difficult to control may occur below the speed range, and the situation that the running speed of the vehicle is higher than the speed range, the situation that the calculated speed is lower and cannot meet the use requirement may occur.

The invention discloses an ultra-low speed creeping method of an unmanned automobile in a mixed road congestion state, which specifically comprises the following steps: when the speed of the unmanned vehicle on the mixed road is lower than a certain speed, calculating a pre-driving track of the vehicle, and drawing a pre-enveloping area of the vehicle according to the pre-driving track of the vehicle; when the number of pedestrians and non-motor vehicles in the vehicle pre-envelope area and the range M of a certain distance exceeds a certain number, the vehicle enters an ultra-low speed crawling mode. The ultra-low speed crawling mode comprises the following steps: analyzing the action tracks of pedestrians and non-motor vehicles in the pre-envelope area of the vehicle and a certain distance range M of the pre-envelope area, and analyzing whether the vehicle enters a parking waiting state, is in deceleration running or is in original speed running according to the distance between the front end of the vehicle and the pedestrians and the non-motor vehicles, whether the action tracks of the pedestrians and the non-motor vehicles deviate from the pre-running track of the vehicle or not, and whether the pedestrians can see the vehicle or not.

It should be noted that, as shown in fig. 4, the vehicle pre-envelope area is an area formed by a vehicle body contour passing through a vehicle pre-travel track, that is, a vehicle pre-travel track in a two-dimensional horizontal plane is drawn according to a vehicle pre-travel route, and the area formed by the vehicle body contour passing through the vehicle pre-travel track is the vehicle pre-envelope area. The range M is a rectangular area formed by expanding the distance a forwards and backwards from the front side to the two sides of the pre-enveloping area of the vehicle respectively, the expansion distance a can be 1-4M, and the preferred expansion distance a is 2-3M. Of course, other larger ranges M are possible, the larger the range M, the more secure the pedestrian and non-motor vehicle can be, but the longer the vehicle takes to pass through the mixed road.

Further, the unmanned vehicle enters an ultra-low speed crawling mode according to the judgment condition: and judging the current running speed V of the vehicle, wherein the current running speed V of the vehicle is lower than 10km/h, and the running speed at the moment is the first running speed V1 of the vehicle. For the safety of the vehicle driving, the running speed of the vehicle at this time is preferably 12km/h, and V1 may be set to 8km/h, or a value selected within the range of 8-12 km/h.

Further, the vehicle pre-running track length S is 1-4 m, and the vehicle pre-running time T = S/V1 is calculated according to the vehicle pre-running track length S and the first running speed V1 of the vehicle. The vehicle pre-driving track is an estimated driving route of the centroid or the geometric center of the vehicle within the vehicle pre-driving time T calculated according to the first driving speed V1 of the vehicle. The length S of the pre-vehicle travel track may be other lengths, and the longer the length S of the pre-vehicle travel track is, the longer the pre-vehicle travel time T is, the larger deviation between the analyzed movement tracks of the pedestrian and the non-motor vehicle and the actual movement tracks of the pedestrian and the non-motor vehicle may exist.

Further, the pre-movement tracks of the pedestrians and the non-motor vehicles are predicted movement tracks formed by the pedestrians and the non-motor vehicles under the current speed and direction of the pedestrians and the non-motor vehicles within the pre-movement time T of the vehicle.

As shown in fig. 1, the invention discloses an ultra-low speed creeping method of an unmanned vehicle in a mixed road congestion state, which comprises the following steps:

a1, as shown in fig. 2, when the vehicle is driven in an unmanned manner, and more pedestrians cross the road in the front or collide with the pre-driving track of the vehicle, the unmanned system of the vehicle calculates the pre-driving time T = S/V1 of the vehicle according to the first driving speed V1 of the vehicle and the length S of the pre-driving track of the vehicle, draws the pre-envelope area of the vehicle and the range M thereof, the environment recognition module judges the driving directions of the pedestrians and the non-motor vehicles and calculates the speeds of the pedestrians and the non-motor vehicles, and calculates the pre-driving tracks of the pedestrians and the non-motor vehicles according to the pre-driving time T of the vehicle.

It should be noted that, the processor used by the unmanned system of the vehicle may be a central processing unit with data analysis, processing and storage functions, the environment recognition module may be a vehicle-mounted panoramic camera or a panoramic camera, and the environment recognition module is in electrical connection with the processor or in communication connection in a wireless manner. As shown in fig. 3, the environment recognition module may transmit the captured image to the processor, and the processor may predict the driving direction and the movement speed of the pedestrian and the non-motor vehicle according to the image and analyze the predicted movement track of the pedestrian and the motor vehicle in the processor. When the vehicle is close to the pedestrian, the eye position of the pedestrian can be captured according to the image, and therefore the next action of the pedestrian can be judged. The ultra-low speed crawling method is based on the existing unmanned control method, namely the existing method used by the vehicle with the unmanned mode, so that the unmanned basic driving method, the environment recognition and acquisition processes related in the application are the prior art, the method for capturing the positions of the eyes of the pedestrians according to the images can use a CNN convolutional neural network or a BP neural network or other mature deep learning methods, and according to the learning methods, the processor can recognize the portrait outline in the picture shot by the environment recognition module and recognize the positions of the eyes in the portrait outline according to the characteristics.

A2, judging the sum N of the number of pedestrians and non-motor vehicles in the range M in real time, and entering the step A7 when N is smaller than a first threshold, wherein the value range of the first threshold is 2-5; when N is greater than a first threshold, determining whether the minimum distance L is greater than a first safe distance b1, if so, entering A3, and if not, entering a 4; and L is the minimum distance between the front end of the vehicle and the pedestrian or the non-motor vehicle in the range M. The first safety distance b1 is 0.5-1 m, and is preferably 1m for ensuring the safety of pedestrians and non-motor vehicles.

And A3, decelerating the vehicle to V2 and continuing to drive forwards, and if the minimum distance L is less than the first safety distance b1, entering A4. V2 is the second driving speed of the vehicle, V2 is 5-7 km/h, and 5km/h is preferable for ensuring the safety of pedestrians and non-motor vehicles.

A4, judging whether the pre-running track of the pedestrian or the non-motor vehicle closest to the front end of the vehicle in the range M deviates from the running direction of the current running speed V of the vehicle, if so, entering A6, and if not, entering A5.

And A5, parking, waiting and monitoring in real time, entering A6 when the pre-running track of the pedestrian or the non-motor vehicle closest to the front end of the vehicle in the range M deviates from the pre-running track of the vehicle, and returning to A3 when the minimum distance L is greater than the first safety distance b 1.

A6, decelerating the vehicle to V3 and continuing to drive forwards until the minimum distance L is smaller than a second safety distance b2, stopping and waiting until the minimum distance L is larger than a second safety distance b2, and returning to the step A4; the V3 is the third traveling speed of the vehicle, and the V3 is 3-5 km/h, preferably 3 km/h. The second safety distance b2 is 0.3-0.5 m, and preferably 0.5m for ensuring the safety of pedestrians and non-motor vehicles.

Further, in step a6, when the minimum distance L is less than the second safety distance b2, it is determined whether the pedestrian is a non-motor vehicle or the vehicle is ahead of the vehicle, and if the vehicle is a non-motor vehicle, the vehicle is parked and waits until the minimum distance L is greater than the second safety distance b2, the process returns to step a 4;

if the image is a pedestrian, judging whether the eyes of the pedestrian can be identified in the image at the moment, and if the eyes of the pedestrian cannot be identified, stopping the vehicle for waiting; if the eyes of the pedestrian can be identified, the vehicle decelerates to the speed of V4 and continues to run until the minimum distance L is less than the third safe distance b3, and then the vehicle stops for waiting; if the eyes of the pedestrian cannot be captured, stopping the vehicle for waiting; when the minimum distance L is larger than b2, returning to the step A4; the V4 is the fourth traveling speed of the vehicle.

The third safety distance b3 is 0.2-0.3 m, preferably 0.3 m. And the fourth running speed V4= V3 α, wherein α is an avoidance coefficient and has a value of 0.1-0.9. The avoidance coefficient alpha can be obtained by a deep learning method or a simulation method, or the two methods can be combined. For example, the avoidance coefficient α is calculated by a simulation method, one or more unmanned vehicles are first used to travel on a simulated road, a plurality of dummy models having a predetermined trajectory are provided around the simulated road, the plurality of dummy models simulate different traveling routes, α is sequentially evaluated in accordance with 0.1, 0.2, and 0.3 … … 0.9.9 under the different traveling routes, and the avoidance coefficient α corresponding to the fastest passing speed of the vehicle and no collision between the vehicle and the dummy models is obtained as an initial value. For example, the avoidance coefficient α is calculated by a deep learning method, a target value, that is, the shortest vehicle passing time is given, and then α and the travel trajectories of a plurality of pedestrians and non-motor vehicles are used as input parameters, and a deep learning model such as a CNN convolutional neural network is used for calculation, so that the optimal value of α is determined.

It should be noted that, in the steps A3 to a6, the sum N of the numbers of pedestrians and non-motor vehicles within the range M is monitored in real time, and when N is smaller than the first threshold, the process directly proceeds to the step a 7.

It should be noted that specific values of the safety distances b1, b2 and b3 can be set arbitrarily according to actual needs, but must satisfy b1 > b2 > b 3. Meanwhile, the specific settings of the vehicle running speeds V1, V2, V3 and V4 may be arbitrarily set according to actual needs, but must satisfy V1 > V2 > V3 > V4.

Example one

As shown in FIG. 5, the vehicle enters the hybrid road at a vehicle first driving speed V1 of 8km/h, a vehicle second driving speed V2 of 6km/h, a vehicle third driving speed V3 of 4km/h, a vehicle fourth driving speed V4 of 2km/h, a first safety distance b1 of 1m, a second safety distance b2 of 0.5m, and a third safety distance b3 of 0.3 m.

A1, the length S of a vehicle pre-driving track is 2M, the vehicle pre-driving time T = S/V1=8/2=4S is calculated, the vehicle pre-enveloping area is drawn, the range M is a rectangular area formed after the vehicle pre-enveloping area extends forwards and backwards by a distance of 2M respectively, and a system on the vehicle analyzes the pre-driving tracks of pedestrians and non-motor vehicles according to the vehicle pre-driving time 4S;

a2, judging the sum N of the number of pedestrians and non-motor vehicles in a range M in real time, wherein 3 pedestrians exist in the range M, the minimum distance L is 1.5M, the minimum distance L is larger than 1M, and the pedestrian enters A3;

a3, decelerating the vehicle to 6km/h and continuing to drive forwards, and entering A4 when the minimum distance L is less than 1 m;

a4, as shown in fig. 6, when the pedestrian in front is closest to the vehicle within the range M and the action direction is deviated from the vehicle running direction, the vehicle enters a 6;

and A6, decelerating the vehicle to 4km/h, and continuing to drive forwards until the minimum distance L is less than 0.5m, wherein the pedestrian is still the pedestrian at the closest distance in front of the vehicle, and the pedestrian walks away from the vehicle, but the environment recognition module of the vehicle does not catch the eyes of the pedestrian, namely the pedestrian does not observe the vehicle, and then the vehicle stops and waits.

Example two

As shown in fig. 7, the vehicle enters the hybrid road at a vehicle first running speed V1 of 8km/h, a vehicle second running speed V2 of 6km/h, a vehicle third running speed V3 of 4km/h, a vehicle fourth running speed V4 of 2km/h, a first safety distance b1 of 1m, a second safety distance b2 of 0.5m, and a third safety distance b3 of 0.3 m.

After the vehicle continues to drive forward for a certain distance in the ultra-low-speed crawling mode, the pedestrian in the lower part of the range M walks towards the vehicle and is closest to the vehicle, and the minimum distance L is smaller than 0.5M, and the judgment process is in step A6: the environment recognition module of vehicle catches pedestrian's eyes, the vehicle slows down to 4km/h and continues to travel forward, the vehicle can further reduce speed and continues to travel to V4 this moment, the pedestrian mostly can take to change the walking direction this moment, carry out certain mode of dodging and walk, make the vehicle can travel with slower mode in the region that the flow of people is great, and the actual conditions that the pedestrian can take certain dodging measure after seeing the vehicle has been considered in the driving process and thus let the vehicle pass through smoothly, very big improvement the vehicle's under this scene ability of traveling, of course in order to guarantee pedestrian's safety, when minimum distance L is less than 0.3m, the vehicle stops and waits for the pedestrian to pass through or keeps away from the vehicle certain distance after and begins to travel again.

The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various changes, modifications, alterations, and substitutions which may be made by those skilled in the art without departing from the spirit of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. the ultra-low-speed creeping method of unmanned vehicles under the mixed road congestion state, it is characterized in that: analyze the action trajectories of pedestrians and non-motor vehicles in the range M around the vehicle pre-envelope area, according to the front end of the vehicle and pedestrians and non-motor vehicles. The distance between the workshops, whether the pre-movement trajectory of pedestrians and non-motor vehicles deviates from the pre-travel trajectory of the vehicle, whether the pedestrian sees the vehicle, and instructs the vehicle to stop and wait, slow down or drive at the original speed;

The vehicle pre-envelope area is an area formed by the contour of the vehicle body through the vehicle pre-travel; the range M is a rectangular area formed by extending the vehicle pre-envelope area forward and to both sides by a distance a;

Described ultra-low speed creeping method comprises the following steps,

A1. Draw the vehicle pre-envelope area and range M, calculate the vehicle pre-travel time T=S/V1, and analyze the pre-movement trajectories of pedestrians and non-motor vehicles;

A2. Judging in real time the sum N of the number of pedestrians and non-motor vehicles within the range M, when N is less than the first threshold, enter step A7, the value range of the first threshold is 2-5; when N is greater than The first threshold value, then judge whether the minimum distance L is greater than the first safety distance b1, if it is greater than the first safety distance b1, then enter A3, if it is less than the first safety distance b1, enter A4; the L is the front end of the vehicle and the range M The minimum distance between pedestrians or non-motorized vehicles within;

A3. The vehicle decelerates to V2 and continues to drive forward. When the minimum distance L is less than the first safety distance b1, it enters A4; V2 is the second driving speed of the vehicle;

A4. Judging whether the pre-movement trajectory of the pedestrian or non-motor vehicle closest to the front end of the vehicle within the range M deviates from the driving direction of the vehicle's current speed V, if it deviates, enter A6, if not, enter A5;

A5. Stop and wait, and monitor in real time. When the pre-action trajectory of the pedestrian or non-motor vehicle closest to the front end of the vehicle within the range M deviates from the pre-travel trajectory of the vehicle, enter A6, and when the minimum distance L is greater than the first safety distance b1, then return A3;

A6. The vehicle decelerates to V3 and continues to drive forward until the minimum distance L is less than the second safe distance b2, stop and wait until the minimum distance L is greater than the second safe distance b2, then return to step A4; the V3 is the third driving speed of the vehicle ;

A7. The vehicle continues to drive, and the N value within the range M is monitored in real time. When the N value is greater than the first threshold, return to step A2;

When the driving speed V1 of the unmanned vehicle is lower than 10km/h, it enters the ultra-low speed creep mode; let V1 be the first driving speed of the vehicle;

The pre-travel trajectory of the vehicle is a pre-travel route calculated by calculating the center of mass or the geometric center of the vehicle within the pre-travel time T of the vehicle at the first running speed V1 of the vehicle; the pre-travel length S of the vehicle is 1-4 m;

The pre-action trajectories of the pedestrian and the non-motor vehicle are the estimated action trajectories formed by the pedestrian and the non-motor vehicle under the respective current speed and direction within the vehicle pre-travel time T;

In the step A6, when the minimum distance L is less than the second safety distance b2, it is determined whether the front of the vehicle is a pedestrian or a non-motor vehicle. If it is a non-motor vehicle, stop and wait until the minimum distance L is greater than the second safety distance b2, then return to step A4. ;

If it is a pedestrian, judge whether the pedestrian's eyes can be recognized in the image at this time. If the pedestrian's eyes can not be recognized, stop and wait; if the pedestrian's eyes can be recognized, the vehicle will decelerate to the speed of V4 and continue to drive to the minimum When the distance L is less than the third safe distance b3, stop and wait; if the pedestrian's eyes cannot be caught, stop and wait; when the minimum distance L is greater than b2, return to step A4; the V4 is the fourth driving speed of the vehicle.

2. The ultra-low-speed creeping method for driverless vehicles in a state of mixed road congestion according to claim 1, wherein the size relationship of the b1, b2 and b3 is b1>b2>b3.

3. the ultra-low speed creeping method of unmanned vehicle under the mixed road congestion state according to claim 1, it is characterized in that: the magnitude relation of described V1, V2, V3 and V4 is V1>V2>V3>V4;

4. The ultra-low-speed creeping method for unmanned vehicles in a state of mixed road congestion according to claim 1, wherein the fourth driving speed of the vehicle is V4=V3*α, and the α is an avoidance coefficient, which is The numerical value is 0.1 to 0.9.

5. The ultra-low-speed creeping method for unmanned vehicles under the mixed road congestion state according to claim 1, characterized in that: in the described steps A3 to A6, real-time monitoring of pedestrians and non-motor vehicles within the range M is performed. The sum of the numbers is N. When N is less than the first threshold, go directly to step A7.