Vehicle automatic driving decision method, device and storage medium

Abstract

The present invention discloses a method, device and storage medium for automatic driving decision-making of a vehicle. The method includes: obtaining candidate driving data of a first vehicle, wherein the candidate driving data is the acceleration and lane change information of the first vehicle in a future time period; obtaining driving data of the first vehicle and driving data of a second vehicle, wherein the driving data of the first vehicle is the position and speed of the first vehicle in a coordinate system at the current moment, the driving data of the second vehicle is the position, speed and lane change information of the second vehicle in a coordinate system at the current moment, and the second vehicle is a vehicle whose distance from the first vehicle in a future time period is less than a distance threshold; based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle, determining the driving decision of the first vehicle. The present invention solves the technical problem of low efficiency of automatic driving decision-making of vehicles.

Description

Technical Field

[0001] The present invention relates to the field of vehicles, and in particular to a method, device and storage medium for automatic driving decision-making of a vehicle. Background Art

[0002] At present, autonomous driving vehicles need to output reasonable decisions under the condition of uncertainty about the intentions of surrounding vehicles. The usual method is to use a local planning algorithm to give the vehicle's current driving path. However, this method does not conform to the actual situation and fails to consider the impact of the current driving path on the vehicle in future periods.

[0003] There is currently no effective solution to the problem of low efficiency in vehicle autonomous driving decision-making. Summary of the invention

[0004] Embodiments of the present invention provide a method, device and storage medium for autonomous driving decision-making of a vehicle, so as to at least solve the technical problem of low efficiency of autonomous driving decision-making of the vehicle.

[0005] According to one aspect of an embodiment of the present invention, a method for automatic driving decision-making of a vehicle is provided. The method includes: obtaining candidate driving data of a first vehicle, wherein the candidate driving data is acceleration and lane change information of the first vehicle in a future period; obtaining driving data of the first vehicle and driving data of a second vehicle, wherein the driving data of the first vehicle is the position and speed of the first vehicle in a coordinate system at the current moment, the driving data of the second vehicle is the position, speed and lane change information of the second vehicle in a coordinate system at the current moment, and the second vehicle is a vehicle whose distance from the first vehicle in a future period is less than a distance threshold; based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle, determining the driving decision of the first vehicle.

[0006] Optionally, obtaining candidate driving data of the first vehicle includes: obtaining first candidate driving data of the first vehicle, wherein the first candidate driving data is the acceleration of the first vehicle when traveling in a future time period; obtaining second candidate driving data of the first vehicle, wherein the second candidate driving data is lane changing data of the first vehicle when traveling in a future time period; and determining the candidate driving data based on the first candidate driving data and the second candidate driving data.

[0007] Optionally, obtaining first candidate driving data of the first vehicle includes: determining the first candidate driving data based on maximum speed data and minimum speed data of the road on which the first vehicle is traveling, the driving speed of the first vehicle, and the time when the first vehicle collides with the second vehicle in the same lane.

[0008] Optionally, obtaining the second candidate driving data of the first vehicle includes: determining the second candidate driving data based on the driving data of the first vehicle on the current road, the driving data of the second vehicle in an adjacent lane of the current road, and the time when the first vehicle and the second vehicle collide in different lanes.

[0009] Acquiring driving data of a first vehicle and driving data of a second vehicle includes: processing the driving data of the first vehicle and the driving data of the second vehicle based on a multi-layer perceptron to predict the probability of a collision between the first vehicle and the second vehicle.

[0010] Optionally, after predicting the probability of a collision between the first vehicle and the second vehicle, the method further includes: based on the probability of a collision between the first vehicle and the second vehicle, predicting driving data of the second vehicle in a future time period, wherein the driving data is lane changing information of the second vehicle in the future time period.

[0011] Optionally, based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle, the driving decision of the first vehicle is determined, including: using a solver to solve the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle to obtain the driving decision of the first vehicle.

[0012] Optionally, the method further includes: updating the driving data of the second vehicle in a future time period based on the driving data of the first vehicle under the driving decision and the driving data of the second vehicle at the time of the driving decision of the first vehicle.

[0013] According to another aspect of an embodiment of the present invention, there is also provided an automatic driving decision-making device for a vehicle, comprising: a first acquisition unit, for acquiring candidate driving data of a first vehicle, wherein the candidate driving data are acceleration and lane changing information of the first vehicle in a future time period; a second acquisition unit, for acquiring driving data of the first vehicle and driving data of a second vehicle, wherein the driving data of the first vehicle are the position and speed of the first vehicle in a coordinate system at a current moment, and the driving data of the second vehicle are the position, speed and lane changing information of the second vehicle in the coordinate system at a current moment, and the second vehicle is a vehicle whose distance from the first vehicle in a future time period is less than a distance threshold; a determination unit, for determining the driving decision of the first vehicle based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle.

[0014] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is further provided. The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the automatic driving decision method for a vehicle of an embodiment of the present invention.

[0015] According to another aspect of an embodiment of the present invention, a processor is further provided. The processor is used to run a program, wherein the automatic driving decision method for a vehicle according to an embodiment of the present invention is executed when the program is run.

[0016] According to another aspect of an embodiment of the present invention, a vehicle is further provided, and the vehicle is used to execute the automatic driving decision method for the vehicle of an embodiment of the present invention.

[0017] In an embodiment of the present invention, candidate driving data of a first vehicle is obtained, wherein the candidate driving data is the acceleration and lane-changing information of the first vehicle in a future time period; the driving data of the first vehicle and the driving data of the second vehicle are obtained, wherein the driving data of the first vehicle is the position and speed of the first vehicle in the coordinate system at the current moment, the driving data of the second vehicle is the position, speed and lane-changing information of the second vehicle in the coordinate system at the current moment, and the second vehicle is a vehicle whose distance from the first vehicle in the future time period is less than a distance threshold; based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle, the driving decision of the first vehicle is determined. That is, the embodiment of the present invention first obtains the acceleration and lane-changing information of the first vehicle in the future time period, and then obtains the driving data of the first vehicle and the driving data of the second vehicle, and finally solves the acceleration and lane-changing information of the first vehicle in the future time period, the driving data of the first vehicle and the driving data of the second vehicle through a solver to obtain the driving decision of the first vehicle, thereby achieving the purpose of obtaining an accurate automatic driving decision of the first vehicle and solving the problem of vehicle

[0018] The technical problem of low efficiency of autonomous driving decision-making has been solved, achieving the technical effect of improving the efficiency of autonomous driving decision-making of vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0020] Figure 1 is a flow chart of an automatic driving decision-making method for a vehicle according to an embodiment of the present invention;

[0021] Figure 2 is a schematic diagram of a decision model for an autonomous driving vehicle according to an embodiment of the present invention;

[0022] Figure 3 is a schematic diagram of a multi-layer perceptron according to an embodiment of the present invention;

[0023] Figure 4 is a schematic diagram of an automatic driving decision-making device for a vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only embodiments of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] Example 1

[0027] According to an embodiment of the present invention, a method for automatic driving decision-making of a vehicle is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0028] Figure 1 is a flow chart of an automatic driving decision-making method for a vehicle according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0029] Step S101, obtaining candidate driving data of a first vehicle, wherein the candidate driving data is acceleration and lane change information of the first vehicle in a future period.

[0030] In the technical solution provided by the above step S101 of the present invention, the candidate action generation module of the first vehicle generates candidate driving data of the first vehicle, obtains longitudinal candidate driving data set 1 of the first vehicle according to the speed of the first vehicle on the driving road and the current maximum speed and minimum speed of the road, obtains longitudinal candidate driving data set 2 of the first vehicle according to the collision time of the first vehicle and the second vehicle in the same lane, extracts the intersection of longitudinal candidate driving data set 1 and longitudinal candidate data set 2, and obtains the final longitudinal candidate data set; wherein, candidate data set 1 and candidate data set 2 are composed of different accelerations; according to the collision time of the first vehicle and the second vehicle in different lanes, determines the lane change information of the vehicle, combines longitudinal candidate driving data set 1, longitudinal candidate driving data set 2 and lane change information, and obtains candidate driving data of the first vehicle. The lane change information of the first vehicle in the future period may be that the first vehicle changes lanes to the left or to the right on the road in the future period.

[0031] Step S102, obtaining driving data of the first vehicle and driving data of the second vehicle, wherein the driving data of the first vehicle is the position and speed of the first vehicle in the coordinate system at the current moment, and the driving data of the second vehicle is the position, speed and lane change information of the second vehicle in the coordinate system at the current moment, and the second vehicle is a vehicle whose distance from the first vehicle in the future time period is less than a distance threshold.

[0032] In the technical solution provided in the above step S102 of the present invention, the coordinate system includes a rectangular coordinate system, and the driving data of the first vehicle is the position of the first vehicle in the coordinate system at the current moment, including the horizontal and vertical positions of the first vehicle in the rectangular coordinate system at the current moment, and the horizontal and vertical speeds of the first vehicle in the rectangular coordinate system at the current moment. Optionally, the radar sensor of the first vehicle obtains the horizontal and vertical positions and horizontal and vertical speeds of the first vehicle in the rectangular coordinate system at the current moment, wherein the horizontal and vertical positions are the decomposition of the current position of the first vehicle in the rectangular coordinate system into the horizontal position and the longitudinal position, and the horizontal and vertical speeds are the decomposition of the current speed of the first vehicle in the rectangular coordinate system into the horizontal speed and the longitudinal speed; the radar sensor of the second vehicle obtains the horizontal and vertical positions, horizontal and vertical speeds and lane change information of the second vehicle in the rectangular coordinate system at the current moment, wherein the horizontal and vertical positions are the decomposition of the current position of the second vehicle in the rectangular coordinate system into the horizontal position and the longitudinal position, and the horizontal and vertical speeds are the decomposition of the current speed of the second vehicle in the rectangular coordinate system into the horizontal speed and the longitudinal speed, and the lane change information is the second vehicle changing lanes to the left, changing lanes to the right or driving in a straight line at the current moment.

[0033] Step S103 , determining a driving decision of the first vehicle based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle.

[0034] In the technical solution provided in the above step S103 of the present invention, after obtaining the candidate driving data of the first vehicle, and obtaining the driving data of the first vehicle and the driving data of the second vehicle, the driving decision of the first vehicle can be determined based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle. The driving decision of the first vehicle can be obtained through a partially observable Markov, and the driving decision can be that the first vehicle changes lanes to the left, changes lanes to the right, or drives in a straight line at a specific speed. This is only an example here without specific limitation.

[0035] In the above steps S101 to S103 of the present application, in an embodiment of the present invention, the candidate driving data of the first vehicle is obtained, wherein the candidate driving data is the acceleration and lane change information of the first vehicle in the future time period; the driving data of the first vehicle and the driving data of the second vehicle are obtained, wherein the driving data of the first vehicle is the position and speed of the first vehicle in the coordinate system at the current moment, the driving data of the second vehicle is the position, speed and lane change information of the second vehicle in the coordinate system at the current moment, and the second vehicle is a vehicle whose distance with the first vehicle in the future time period is less than the distance threshold; based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle, the driving decision of the first vehicle is determined. That is to say, the embodiment of the present invention first obtains the acceleration and lane change information of the first vehicle in the future time period, then obtains the driving data of the first vehicle and the driving data of the second vehicle, and finally solves the acceleration and lane change information of the first vehicle in the future time period, the driving data of the first vehicle and the driving data of the second vehicle through the solver to obtain the driving decision of the first vehicle, thereby achieving the purpose of obtaining an accurate automatic driving decision of the first vehicle, solving the technical problem of low efficiency of automatic driving decision of the vehicle, and achieving the technical effect of improving the efficiency of automatic driving decision of the vehicle.

[0036] The above method of this embodiment is further introduced below.

[0037] As an optional implementation method, step S101, obtaining candidate driving data of the first vehicle, includes: obtaining first candidate driving data of the first vehicle, wherein the first candidate driving data is the acceleration of the first vehicle in a future time period; obtaining second candidate driving data of the first vehicle, wherein the second candidate driving data is lane change data of the first vehicle in a future time period; and determining candidate driving data based on the first candidate driving data and the second candidate driving data.

[0038] In this embodiment, the candidate action generation module of the first vehicle generates a driving acceleration of the first vehicle in a future time period based on the maximum speed and the minimum speed of the first vehicle on the current road, the driving speed of the first vehicle, and the collision time between the first vehicle and the second vehicle; the candidate generation module of the first vehicle determines whether the first vehicle will change lanes to the left or to the right in the future time period based on the collision time between the first vehicle and the second vehicle, and combines the driving acceleration of the first vehicle in the future time period and the lane change to the left or to the right to obtain candidate driving data.

[0039] As an optional implementation method, obtaining first candidate driving data of the first vehicle includes: determining the first candidate driving data based on the maximum speed data and the minimum speed data of the road on which the first vehicle is traveling, the driving speed of the first vehicle, and the time when the first vehicle collides with the second vehicle in the same lane.

[0040] In this embodiment, the maximum speed data, the minimum speed data and the driving speed of the first vehicle on the road are obtained by the radar sensor of the first vehicle. The difference between the maximum speed data and the driving speed of the first vehicle on the current driving road is within a fixed range, and candidate acceleration set one is obtained. The difference between the driving speed of the first vehicle on the current road and the minimum speed data of the current road is within a fixed range, and candidate acceleration set two is obtained. When the second vehicle is in front of the first vehicle and the time when the first vehicle collides with the second vehicle is within a fixed time range, candidate acceleration set three is obtained. When the second vehicle is behind the first vehicle and the time when the first vehicle collides with the second vehicle is within a fixed time range, candidate acceleration set four is obtained. The intersection of candidate acceleration set one, candidate acceleration set two, candidate acceleration set three and candidate acceleration set four is taken to obtain the first candidate driving data.

[0041] For example, when the maximum speed of the first vehicle traveling on the current road is 120 kilometers per hour, the minimum speed is 100 kilometers per hour, and the driving speed of the first vehicle is 110 kilometers per hour, when the difference between the driving speed and the maximum speed is -10 kilometers per hour, that is, -2.7 meters per second, -2.7 meters per second is between -5 meters per second and -3 meters per second, therefore, the candidate acceleration set 1 is {-2, -1.5, -1, -0.5, 0, 0.5, 1}; when the difference between the driving speed and the minimum speed is 10 kilometers per hour, that is, 2.7 meters per second, 2.7 meters per second is between 1 meter per second and 5 meters per second, therefore, the candidate acceleration set 2 is {-2, -1.5, -1, -0.5, 0, 0.5, 1}. 0.5, 0, 0.5, 1, 1.5, 2}; when the second vehicle is in front of the first vehicle, the time for the first vehicle to collide with the second vehicle is 1 second, which is between 0 seconds and 2 seconds. Therefore, the candidate acceleration set three is {-2, -1.5, -1, -0.5}; when the second vehicle is behind the first vehicle, the time for the first vehicle to collide with the second vehicle is 7 seconds, which is between 5 seconds and 10 seconds. Therefore, the candidate acceleration set four is {-1, -0.5, 0, 0.5, 1, 1.5, 2}; take the intersection of candidate acceleration set one, candidate acceleration set two, candidate acceleration set three and candidate acceleration set four to obtain the first candidate driving data {-1, -0.5}.

[0042] As an optional implementation method, obtaining the second candidate driving data of the first vehicle includes: determining the second candidate driving data based on the driving data of the first vehicle on the current road, the driving data of the second vehicle in an adjacent lane of the current road, and the time when the first vehicle and the second vehicle collide in different lanes.

[0043] In this embodiment, the radar sensor of the first vehicle detects the driving speed of the first vehicle on the current road and the driving speed of the second vehicle in the adjacent lane of the current road. When the first vehicle is in front of the left of the second vehicle, the collision time between the first vehicle and the second vehicle is one. When the first vehicle is behind the left of the second vehicle, the collision time between the first vehicle and the second vehicle is two. When the collision time one and the collision time two are both less than the fixed threshold, the first vehicle changes lanes to the left. When the first vehicle is in front of the right of the second vehicle, the collision time between the first vehicle and the second vehicle is three. When the first vehicle is behind the right of the second vehicle, the collision time between the first vehicle and the second vehicle is four. When the collision time three and the collision time four are both less than the fixed threshold, the first vehicle changes lanes to the right.

[0044] For example, the radar sensor of the first vehicle detects that the driving data of the first vehicle on the current road is 50 kilometers per hour, and the driving data of the second vehicle in the adjacent lane of the current road is 60 kilometers per hour. When the first vehicle is in front of the left of the second vehicle, the collision time between the first vehicle and the second vehicle is 6 seconds, and when the first vehicle is behind the left of the second vehicle, the collision time between the first vehicle and the second vehicle is 7 seconds. Therefore, the first vehicle changes lanes to the left; when the first vehicle is in front of the right of the second vehicle, the collision time between the first vehicle and the second vehicle is 6 seconds, and when the first vehicle is behind the right of the second vehicle, the collision time between the first vehicle and the second vehicle is 7 seconds. Therefore, the first vehicle changes lanes to the right.

[0045] As an optional implementation method, step S102, obtaining driving data of the first vehicle and the driving data of the second vehicle, includes: processing the driving data of the first vehicle and the driving data of the second vehicle based on a multi-layer perceptron, and predicting the probability of a collision between the first vehicle and the second vehicle.

[0046] In this embodiment, the driving data of the first vehicle and the second vehicle within a fixed time period is used as the input of the multilayer perceptron, and the probability of collision between the first vehicle and the second vehicle within a short time period is predicted as the output of the multilayer perceptron.

[0047] For example, the position, velocity and acceleration of the first vehicle and the second vehicle in the rectangular coordinate system at time 0 are used as the input of the multi-layer perception sensor. When the first vehicle collides with the second vehicle at time 0, it is recorded as 1, and when the first vehicle and the second vehicle do not collide at time 0, it is recorded as 0. 0 and 1 are used as labels, and the probability of the first vehicle and the second vehicle colliding within 3 seconds is 80% as the output of the multi-layer perception sensor.

[0048] As an optional implementation mode, after predicting the probability of a collision between the first vehicle and the second vehicle, the method further includes: based on the probability of a collision between the first vehicle and the second vehicle, predicting driving data of the second vehicle in a future time period, wherein the driving data is lane changing information of the second vehicle in the future time period.

[0049] In this embodiment, after the probability of the first vehicle colliding with the second vehicle is obtained, when the probability of the first vehicle colliding with the second vehicle is greater than a fixed probability threshold, the second vehicle changes lanes to the left or to the right in a future period.

[0050] As an optional implementation method, step S103, based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle, determines the driving decision of the first vehicle, including: using a solver to solve the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle to obtain the driving decision of the first vehicle.

[0051] In this embodiment, the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle are used as parameters in the partially observable Markov decision process. The partially observable Markov is solved according to the determined sparse partially observable tree (Determinized Sparse Partially Observable Tree, abbreviated as DESPOT) to obtain the current optimal driving strategy of the first vehicle. The optimal driving strategy may be changing lanes to the right at a speed of 50 kilometers per hour, which is only used as an example here.

[0052] As an optional embodiment, the method further includes: updating the driving data of the second vehicle in a future time period based on the driving data of the first vehicle under the driving decision and the driving data of the second vehicle when the driving decision of the first vehicle is made.

[0053] In this embodiment, the partially observable Markov method is based on the driving data of the first vehicle in the current driving decision, and the second vehicle obtains new driving data according to the changes in the data of the first vehicle under the current driving decision, and updates the driving speed and driving trajectory of the second vehicle in the future time period.

[0054] This embodiment obtains the acceleration and lane-changing information of the first vehicle in the future time period, and then obtains the driving data of the first vehicle and the driving data of the second vehicle. Finally, a solver is used to solve the acceleration and lane-changing information of the first vehicle in the future time period, the driving data of the first vehicle, and the driving data of the second vehicle to obtain the driving decision of the first vehicle, thereby solving the technical problem of low efficiency of the vehicle's automatic driving decision-making and achieving the technical effect of improving the efficiency of the vehicle's automatic driving decision-making.

[0055] Example 2

[0056] The technical solution of the embodiment of the present invention is illustrated below in conjunction with preferred implementation modes.

[0057] When a vehicle is driving on a highway, an autonomous vehicle needs to make reasonable decisions under the condition of uncertainty about the intentions of surrounding vehicles. The traditional method is mainly to combine prediction information and select the path with the least cost, but this greedy strategy fails to consider the long-term impact of the current action.

[0058] In order to overcome the above problems, a related technology proposes that for planning in an uncertain environment, a partially observable Markov decision process (POMDP) ​​provides a basic framework. The POMDP framework has been able to solve the real-world sequential decision-making problem well. Considering that when a robot performs a task, the important factors that restrict the reliable operation of the robot are the uncertainty of the environment, sensors, etc. Therefore, it is an important topic in the field of robotics for the robot to achieve online and effective planning in an uncertain environment with a large state and observation space. In view of the defects and deficiencies of the prior art, the purpose of this invention is to provide an online planning method based on an efficient search strategy in an uncertain environment, which is suitable for online planning when the robot performs a task. It introduces an efficient search strategy on the basic framework of DESPOT and proposes a line planning algorithm. Consider combining the initial upper and lower limits of the belief node and then discounting them to represent the overall information of the current belief node. The core of the efficient search strategy is to consider the factors affecting the information redundancy and accuracy of the belief node in the forward search process. By combining the upper and lower limits of the belief node and then discounting them to represent the current belief information, and using this to construct heuristics for forward search, the best behavior and observation branches obtained by the efficient search strategy avoid the greed and suboptimality of a single upper limit search, and can expand a better belief tree, reducing the expansion of behavior and observation branches with poor quality. When constructing a belief tree, DESPOT uses a single upper limit as a heuristic for forward search, but for the current belief node, a single upper limit value cannot accurately represent the current belief information. Therefore, for the current belief node, a forward search is performed with discounted upper and lower limits to expand a better belief node, thereby obtaining a better belief tree.

[0059] In another related technology, a partially observable driving planning method based on adaptive particles and belief filling is proposed. The partially observable Markov decision process is used to model and solve the uncertainty. The autonomous driving problem is modeled as a POMDP problem, and an efficient online planning method is used. It can handle the uncertainty of the state well and solve the difficulties brought by the continuous observation space.

[0060] In another related technology, a method for dynamic planning of drone paths to minimize information age is proposed. Drones are introduced to solve the problem that update data packets generated cannot be delivered in time due to the limited power of IoT devices. The trajectory of the drone is planned through a deep reinforcement learning algorithm to achieve the optimization goal of minimizing the long-term weighted average information age.

[0061] In the above-mentioned related technologies, POMDP is introduced into decision making, but the solution efficiency is not high. Therefore, the embodiment of the present invention uses candidate actions and a method for finding potential interaction objects to improve the final solution efficiency. Figure 2 is a schematic diagram of a decision model for an autonomous driving vehicle according to an embodiment of the present invention. Figure 2 As shown, the model may include: a perception information module 201, a candidate action generation module 202, a dynamic obstacle interaction prediction module 203, and a POMDP solution module 204.

[0062] The perception information module 201 can be used by the radar sensor of the vehicle to detect the speed, position and acceleration of the vehicle and the vehicles around the vehicle.

[0063] Candidate action generation module 202, Table 1 is a longitudinal action space according to an embodiment of the present invention, as shown in Table 2, the longitudinal {a=±0.5m / s 2 , a=±1m / s 2 , a=±1.5m / s 2 , a=±2m / s 2 , a=0m / s 2} action space, Table 3 is the lateral action space according to an embodiment of the present invention, as shown in Table 2, the lateral action space of {changing lanes to the left, changing lanes to the right}.

[0064] Table 1 Longitudinal action space

[0065] -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 -3<V-Vmax<-1m / s √ √ √ √ √ -5<V-Vmax<-3m / s √ √ √ √ √ √ √ V-Vmax<-5m / s √ √ √ √ √ √ √ √ √ 3>V-Vmin>1m / s √ √ √ √ 5>V-Vmin>3m / s √ √ √ √ √ √ √ 5>V-Vmin>1m / s √ √ √ √ √ √ √ √ √ TTC of the preceding vehicle>10s √ √ √ √ √ √ √ √ √ 10s> TTC of the preceding vehicle> 5s √ √ √ √ √ √ √ 5s> TTC of the car ahead> 2s √ √ √ √ √ 2s> TTC of the preceding vehicle> 0s √ √ √ √ TTC of the following vehicle>10s √ √ √ √ √ √ √ √ √ 10s> TTC of the following car> 5s √ √ √ √ √ √ √ 5s> TTC of the following car> 2s √ √ √ √ √

[0066] Table 2 Horizontal action space

[0067] Change lane left Change lane right Left front TTC>5&&left rear TTC>5 √ Right front TTC>5&&Right rear TTC>5 √

[0068] As shown in Table 1, according to the maximum and minimum speed limit information of the highway, combined with the vehicle's driving speed information, candidate longitudinal actions 1 and 2 are obtained through comparison. According to the time-to-collision (TTC) of the front and rear vehicles in the same lane, candidate longitudinal actions 3 and 4 are obtained, and their intersection is taken as the final longitudinal candidate action.

[0069] As can be seen from Table 2, based on the driving information of the adjacent lanes, it is determined whether the lane change conditions are met, and the candidate actions for the left and right road conditions are taken as the union set as the final lateral candidate action.

[0070] The final candidate actions in horizontal and vertical directions are combined as the final candidate actions to achieve the purpose of fast calculation.

[0071] Dynamic obstacle interaction prediction module 203, this module is mainly based on the driving data of the self-vehicle and surrounding vehicles, and uses a neural network to determine whether the self-vehicle will interact with the surrounding vehicles. The driving data of the self-vehicle and surrounding vehicles during the vehicle's driving process is recorded for about 20 hours, with 3s as a small segment, marked as 0-3s, and the horizontal and vertical coordinates, horizontal and vertical speeds, and accelerations of the self-vehicle and surrounding vehicles at time 0s are recorded as the input of the multi-layer perceptron. Figure 3 is a schematic diagram of a multi-layer perceptron according to an embodiment of the present invention, such as Figure 3 As shown in the figure, within 0-3s, observe whether each surrounding vehicle interacts with the vehicle. If an interaction occurs, it is recorded as 1, otherwise 0. The interaction data set is used as a label to obtain an interaction data set, and the multi-layer perceptron is trained using the data set. Here, there are two hidden layers, and each neuron consists of linear operations and non-linear activations. The non-linear activation function uses a sigmoid function, and the output is connected to a softmax classifier to output the probability of interaction and non-interaction. In this way, the model and real-time perception data can be used to predict the probability of the vehicle interacting with a surrounding vehicle in the next 3s.

[0072] POMDP solution module 204, POMDP models the system, and then takes a series of actions under uncertainty to maximize the reward. Formally, POMDP is a tuple: S, A, Z are state space, action space and observation space respectively. At each time step, the system performs an action, the state migrates, and an observation is obtained. T is a conditional probability equation, which represents the probability distribution of the new state after taking the action. O is a conditional probability equation, which represents an observation containing noise. R represents the reward after performing a certain action in a certain state; γ is a discount factor. In a partially observable system, the system state cannot be fully known, so a confidence is maintained for the possible state, and it is updated based on the Bayesian criterion: online POMDP planning looks for a strategy that maximizes the expected total reward under the current confidence. DESPOT is a solver for solving POMDP. The nodes of the confidence tree are confidences, and the edges are action-observation pairs. We search for strategies through subsequent traversals in the confidence tree. The leaf nodes use the default strategy to achieve the lower limit value, and the internal nodes select the best action through the Bellman equation. The equation recursively calculates the maximum value of the action branch and the average value of the observation branch. The result is a near-optimal strategy under the current confidence level.

[0073] (1) State space

[0074] The lateral and longitudinal positions, lateral and longitudinal speeds of the vehicle and surrounding vehicles, and the driving intentions of surrounding vehicles.

[0075] (2) Action Space

[0076] Longitudinal {a=±0.5m / s 2 , a=±1m / s2 , a=±1.5m / s 2 , a=±2m / s 2 , a=0m / s 2} action space in the horizontal direction, and the action space of {change lane to the left, change lane to the right} in the horizontal direction.

[0077] (3) Observation space

[0078] The lateral and longitudinal positions and lateral and longitudinal speeds of the vehicle and surrounding vehicles.

[0079] (4) Reward Function

[0080] First of all, safety. If a collision occurs, the reward is -10. Comfort mainly considers the acceleration value. The larger the value, the greater the negative reward. The efficiency calculation formula is vv max / v max The larger the value, the better. We hope that the vehicle can travel at the highest possible speed.

[0081] In the generative model of POMDP, the adopted actions should be generated from the final set of candidate actions rather than the entire action space.

[0082] In this embodiment, the speed, position, and acceleration of the vehicle and surrounding vehicles are obtained through the perception information module, the candidate action generation module obtains the final candidate action of the vehicle based on the lateral candidate action and the longitudinal candidate action, the dynamic obstacle interaction prediction module predicts the probability of the vehicle interacting with a surrounding vehicle in the future based on the driving data of the vehicle and surrounding vehicles, and the POMDP solution module obtains the final driving action of the vehicle based on the lateral and longitudinal positions, lateral and longitudinal speeds of the vehicle and surrounding vehicles, the intentions of the surrounding vehicles, and the final candidate action, thereby solving the technical problem of low efficiency of vehicle autonomous driving decision-making and achieving the technical effect of improving the efficiency of vehicle autonomous driving decision-making.

[0083] Example 3

[0084] According to an embodiment of the present invention, there is also provided an automatic driving decision-making device for a vehicle. It should be noted that the automatic driving decision-making device for a vehicle can be used to execute the automatic driving decision-making method for a vehicle in Example 1.

[0085] Figure 4 Schematic diagram of an automatic driving decision-making device for a vehicle according to an embodiment of the present invention. Figure 4 As shown, the vehicle's automatic driving decision-making device 400 may include: a first acquisition unit 401, a second acquisition unit 402, and a determination unit 403.

[0086] The first acquisition unit 401 is used to acquire candidate driving data of the first vehicle, wherein the candidate driving data is acceleration and lane change information of the first vehicle in a future period.

[0087] The second acquisition unit 402 is used to acquire the driving data of the first vehicle and the driving data of the second vehicle, wherein the driving data of the first vehicle is the position and speed of the first vehicle in the coordinate system at the current moment, and the driving data of the second vehicle is the position, speed and lane change information of the second vehicle in the coordinate system at the current moment, and the second vehicle is a vehicle whose distance from the first vehicle in the future time period is less than a distance threshold.

[0088] The determination unit 403 is used to determine the driving decision of the first vehicle based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle.

[0089] Optionally, the first acquisition unit 401 may include: a first acquisition module, used to acquire first candidate driving data of the first vehicle, wherein the first candidate driving data is the acceleration of the first vehicle in a future time period; a second acquisition module, used to acquire second candidate driving data of the first vehicle, wherein the second candidate driving data is lane changing data of the first vehicle in a future time period; and a determination module, used to determine the candidate driving data based on the first candidate driving data and the second candidate driving data.

[0090] Optionally, the first acquisition module may include: a first determination submodule, used to determine the first candidate driving data based on the maximum speed data and the minimum speed data of the first vehicle's driving road, the driving speed of the first vehicle, and the time when the first vehicle collides with the second vehicle in the same lane.

[0091] Optionally, the second acquisition module may include: a second determination submodule, used to determine the second candidate driving data based on the driving data of the first vehicle on the current road, the driving data of the second vehicle in an adjacent lane of the current road, and the time when the first vehicle and the second vehicle collide in different lanes.

[0092] Optionally, the second acquisition unit 402 may include: a first processing module, configured to process the driving data of the first vehicle and the driving data of the second vehicle based on a multi-layer perceptron, and predict the probability of a collision between the first vehicle and the second vehicle.

[0093] Optionally, the processing module may include: a prediction submodule, for predicting driving data of the second vehicle in a future time period based on the probability of a collision between the first vehicle and the second vehicle, wherein the driving data is lane changing information of the second vehicle in the future time period.

[0094] Optionally, the determination unit 403 may include: a second processing module, configured to use a solver to solve the candidate driving data of the first vehicle, the driving data of the first vehicle, and the driving data of the second vehicle to obtain the driving decision of the first vehicle.

[0095] Optionally, the device further includes: an updating unit, configured to update the driving data of the second vehicle in a future period based on the driving data of the first vehicle under the driving decision and the driving data of the second vehicle when the driving decision of the first vehicle is made.

[0096] In this embodiment, candidate driving data of the first vehicle is acquired by a first acquisition unit, wherein the candidate driving data is acceleration and lane changing information of the first vehicle in a future time period; driving data of the first vehicle and driving data of the second vehicle are acquired by a second acquisition unit, wherein the driving data of the first vehicle is the position and speed of the first vehicle in the coordinate system at a current moment, and the driving data of the second vehicle is the position, speed and lane changing information of the second vehicle in the coordinate system at a current moment, and the second vehicle is a vehicle whose distance from the first vehicle in the future time period is less than a distance threshold; the driving decision of the first vehicle is determined by a determination unit based on the candidate driving data of the first vehicle, the driving data of the first vehicle and the driving data of the second vehicle, thereby solving the technical problem of low efficiency of automatic driving decision-making of the vehicle and achieving the technical effect of improving the efficiency of automatic driving decision-making of the vehicle.

[0097] Example 4

[0098] According to an embodiment of the present invention, a computer-readable storage medium is also provided, which includes a stored program, wherein the program executes the automatic driving decision-making method for the vehicle in Example 1.

[0099] Example 5

[0100] According to an embodiment of the present invention, a processor is also provided, which is used to run a program, wherein the automatic driving decision method for the vehicle in Example 1 is executed when the program is running.

[0101] Example 6

[0102] According to an embodiment of the present invention, a vehicle is also provided, which is used to execute the automatic driving decision-making method for the vehicle in Example 1.

[0103] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0104] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0105] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0106] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0107] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0108] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.

[0109] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A vehicle automatic driving decision-making method, characterized in that: include: Acquire candidate driving data of the first vehicle, wherein the candidate driving data is acceleration and lane change information of the first vehicle in a future time period; Acquire driving data of the first vehicle and driving data of the second vehicle, wherein the driving data of the first vehicle is the position and speed of the first vehicle in the coordinate system at the current moment, the second vehicle is a vehicle whose distance from the first vehicle in the future time period is less than a distance threshold, and the driving data of the second vehicle includes lane change information, lateral and longitudinal positions, and lateral and longitudinal speeds of the second vehicle at the current moment, the lane change information is used to indicate that the second vehicle changes lanes to the left, changes lanes to the right, or drives in a straight line at the current moment, the lateral and longitudinal positions include the lateral position and longitudinal position obtained by decomposing the position of the second vehicle at the current moment in the coordinate system, and the lateral and longitudinal speeds include the lateral speed and longitudinal speed obtained by decomposing the speed of the second vehicle at the current moment in the coordinate system; determining a driving decision of the first vehicle based on the candidate driving data of the first vehicle, the driving data of the first vehicle, and the driving data of the second vehicle; The method further includes: updating the driving data of the second vehicle in the future time period based on the driving data of the first vehicle under the driving decision and the driving data of the second vehicle when the first vehicle makes the driving decision.

2. The method according to claim 1, characterized in that Obtaining candidate driving data of the first vehicle, including: Acquire first candidate driving data of the first vehicle, wherein the first candidate driving data is the acceleration of the first vehicle in a future time period; Acquire second candidate driving data of the first vehicle, wherein the second candidate driving data is lane change data of the first vehicle in a future time period; The candidate driving data is determined based on the first candidate driving data and the second candidate driving data.

3. The method according to claim 2, characterized in that Acquiring first candidate driving data of the first vehicle includes: The first candidate driving data is determined based on the highest speed data and the lowest speed data of the road on which the first vehicle is traveling, the longitudinal speed of the first vehicle, and the time when the first vehicle collides with the second vehicle in the same lane.

4. The method according to claim 2, characterized in that: Acquiring second candidate driving data of the first vehicle, including: The second candidate driving data is determined based on the driving data of the first vehicle on the current road, the driving data of the second vehicle in an adjacent lane of the current road, and the time when the first vehicle and the second vehicle collide in different lanes.

5. The method according to claim 1, characterized in that Acquiring the driving data of the first vehicle and the driving data of the second vehicle, including: The driving data of the first vehicle and the driving data of the second vehicle are processed based on a multi-layer perceptron to predict the probability of a collision between the first vehicle and the second vehicle.

6. The method according to claim 5, characterized in that After predicting the probability of a collision between the first vehicle and the second vehicle, the method further includes: Based on the probability of a collision between the first vehicle and the second vehicle, driving data of the second vehicle in a future time period is predicted, wherein the driving data is lane changing information of the second vehicle in the future time period.

7. The method according to any one of claims 1, characterized in that Determining a driving decision of the first vehicle based on the candidate driving data of the first vehicle, the driving data of the first vehicle, and the driving data of the second vehicle includes: A solver is used to solve the candidate driving data of the first vehicle, the driving data of the first vehicle, and the driving data of the second vehicle to obtain a driving decision of the first vehicle.

8. A vehicle automatic driving decision-making device, characterized in that: include: a first acquisition unit, configured to acquire candidate driving data of a first vehicle, wherein the candidate driving data is acceleration and lane change information of the first vehicle in a future period; a second acquisition unit, configured to acquire driving data of the first vehicle and driving data of a second vehicle, wherein the driving data of the first vehicle is the position and speed of the first vehicle in the coordinate system at a current moment, the second vehicle is a vehicle whose distance from the first vehicle in the future time period is less than a distance threshold, and the driving data of the second vehicle includes lane change information, lateral and longitudinal positions, and lateral and longitudinal speeds of the second vehicle at a current moment, the lane change information is used to indicate that the second vehicle changes lanes to the left, changes lanes to the right, or drives in a straight line at the current moment, the lateral and longitudinal positions include the lateral position and longitudinal position obtained by decomposing the position of the second vehicle at the current moment in the coordinate system, and the lateral and longitudinal speeds include the lateral speed and longitudinal speed obtained by decomposing the speed of the second vehicle at the current moment in the coordinate system; a determining unit, configured to determine a driving decision of the first vehicle based on the candidate driving data of the first vehicle, the driving data of the first vehicle, and the driving data of the second vehicle; The device is also used to perform the following steps: based on the driving data of the first vehicle under the driving decision and the driving data of the second vehicle when the first vehicle makes the driving decision, updating the driving data of the second vehicle in the future time period.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

Images