Automatic driving decision-making method and device

Abstract

The invention relates to the technical field of automatic driving decision planning, and discloses an automatic driving decision method and device. The automatic driving decision-making method comprises the steps of determining at least one candidate path trajectory based on path planning information, and obtaining trajectory parameters and obstacle parameters corresponding to the candidate path trajectories; determining a trajectory cost parameter and a behavior cost parameter of each candidate path trajectory according to the trajectory parameters and the obstacle parameters; and determining a comprehensive cost value of each candidate path trajectory according to the trajectory cost parameter and the behavior cost parameter, and determining a target decision trajectory from all the candidate path trajectories according to the comprehensive cost value. According to the method, the trajectory cost of the space level and the behavior cost of the time level are comprehensively considered, the trajectory cost parameters and the behavior cost parameters are utilized to realize space-time joint planning, the comprehensiveness of cost scoring is ensured, the flexibility and the accuracy of planning decision making are improved, and meanwhile, the reasonability of obstacle detouring is also improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving decision-making technology, and in particular to an autonomous driving decision-making method and device. Background Art

[0002] With the development of artificial intelligence and the Internet of Vehicles, more and more vehicles have the function of autonomous driving. In the prior art, when the autonomous driving solution is applied to a real vehicle, there is a planning scheme that first performs path planning and then speed planning. This scheme separates the path and speed in time and space, which may result in conservative not taking a detour when it is possible, and may also result in aggressive and wrong detours when it is not possible. It cannot use the detailed planning results to make the best decision and it is difficult to deal with dynamic obstacle scenarios. In the prior art, there is also a decision-making scheme that first deduces the trajectory of the vehicle and other vehicles and then makes a scoring decision. It needs to deduce the driving trajectory according to the rule algorithm, which is easy to lead to too late decision-making on obstacle avoidance. When facing low-speed urban conditions, there will be driving trajectory deduction errors. Therefore, the above-mentioned autonomous driving schemes in the prior art have the problems of inflexible planning decisions and unreasonable obstacle avoidance. Summary of the invention

[0003] Based on this, it is necessary to provide an autonomous driving decision-making method and device to address the above-mentioned technical problems, so as to solve the problems of poor planning and decision-making flexibility and unreasonable obstacle avoidance in existing autonomous driving solutions.

[0004] An automatic driving decision-making method, comprising: Determine at least one candidate path trajectory based on the path planning information, and obtain trajectory parameters and obstacle parameters corresponding to each of the candidate path trajectories; Determining trajectory cost parameters and behavior cost parameters of each of the candidate path trajectories according to the trajectory parameters and obstacle parameters; The comprehensive cost value of each of the candidate path trajectories is determined according to the trajectory cost parameter and the behavior cost parameter, and the target decision trajectory is determined from all the candidate path trajectories according to the comprehensive cost value.

[0005] An automatic driving decision-making device includes a processor and a memory, wherein: The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory to implement the above-mentioned automatic driving decision-making method.

[0006] In the above-mentioned autonomous driving decision-making method and device, the autonomous driving decision-making method determines at least one candidate path trajectory based on path planning information, and obtains trajectory parameters and obstacle parameters corresponding to each candidate path trajectory; determines the trajectory cost parameter and behavior cost parameter of each candidate path trajectory according to the trajectory parameters and obstacle parameters; determines the comprehensive cost value of each candidate path trajectory according to the trajectory cost parameter and behavior cost parameter, and determines the target decision trajectory from all candidate path trajectories according to the comprehensive cost value. The present invention comprehensively considers the trajectory cost at the spatial level and the behavior cost at the time level, obtains the comprehensive cost value by using the trajectory cost parameter and behavior cost parameter under the same priority, realizes the spatio-temporal joint planning, ensures the comprehensiveness of the cost scoring, and improves the flexibility and accuracy of the planning decision-making. At the same time, after determining the candidate path trajectory based on the path planning information, the present invention also considers the obstacle parameter in addition to the trajectory parameter, improving the rationality of obstacle avoidance. The present invention evaluates the cost using comprehensive elements and screens out the target decision trajectory, achieving a balance in terms of decision-making efficiency, accuracy, stability and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments of the present invention. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0008] Figure 1 is a flowchart of an autonomous driving decision-making method in an embodiment of the present invention; Figure 2 is a flowchart of step S10 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 3 is a flowchart of step S20 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 4 is a schematic diagram of an obstacle coordinate system of the autonomous driving decision-making method in an embodiment of the present invention; Figure 5 is another flowchart of step S20 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 6 is a flowchart of step S206 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 7 is another flowchart of step S20 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 8 is a schematic diagram of the positional relationship between an obstacle and a vehicle of the autonomous driving decision-making method in an embodiment of the present invention; Figure 9 It is a schematic flowchart of step S211 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 10 It is a schematic flowchart of step S30 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 11 It is another schematic flowchart of step S30 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 12 It is a schematic flowchart of step S303 of the autonomous driving decision-making method in an embodiment of the present invention; Figure 13 It is a schematic structural diagram of an autonomous driving decision-making device in an embodiment of the present invention. Detailed implementation manners

[0009] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0010] The autonomous driving decision-making method provided in this embodiment can be applied to the path planning decision application scenario when the vehicle starts autonomous driving. Among them, the planning decision of the autonomous driving system is based on the fusion of multiple sensor information, and various possible driving paths are comprehensively planned by using the factors of the vehicle itself and the surrounding environment. Finally, a suitable driving path is selected through comparison, analysis and judgment to make a decision. The behavior decision (the block to which the present invention belongs) selects the best behavior. In addition, in addition to path planning and path decision-making, the autonomous driving system can also include a path optimization function, and finally control the vehicle to execute the optimized driving path, such as performing trajectory smoothing processing and other optimizations based on the selected driving path.

[0011] In an embodiment, as Figure 1 shown, an autonomous driving decision-making method is provided, including the following steps S10-S30: S10. Determine at least one candidate path trajectory based on the path planning information, and obtain the trajectory parameters and obstacle parameters corresponding to each of the candidate path trajectories.

[0012] Understandably, path planning information refers to the key data used to calculate possible driving paths in the context of autonomous driving applications, including map road information (such as road type, width, connection relationships, etc.), geographical information (such as terrain features, altitude, longitude and latitude, etc.), and obstacle information (such as the position and size of obstacles, whether they can be bypassed, etc.). Based on the path planning information, one or more candidate path trajectories can be determined. A candidate path trajectory refers to a possible driving path given after analyzing the path planning information. For each candidate path trajectory, the autonomous driving system obtains the trajectory parameters and obstacle parameters corresponding to each candidate path trajectory to facilitate subsequent comparative analysis of different candidate path trajectories using the trajectory parameters and obstacle parameters. Trajectory parameters refer to the driving path and status information of the vehicle in the candidate path trajectory, such as driving speed, acceleration, path length, etc. Obstacle parameters refer to the numerical values or information used to describe the characteristics and behaviors of obstacles, such as the position and size of obstacles.

[0013] The autonomous driving system adopts a method of triggering planning decisions frame by frame at fixed time intervals. For example, when the fixed time interval is 0.1 second, the interval between two adjacent frames is 0.1 second, that is, one or more candidate path trajectories are planned every 0.1 second, and one of these candidate path trajectories is selected as the target decision trajectory to be implemented at the current moment.

[0014] S20. Determine the trajectory cost parameters and behavior cost parameters of each of the candidate path trajectories according to the trajectory parameters and the obstacle parameters.

[0015] Understandably, for each candidate path trajectory, the autonomous driving system performs trajectory cost analysis and behavior cost analysis based on the trajectory parameters and obstacle parameters to obtain the trajectory cost parameters and behavior cost parameters. The trajectory cost parameter refers to the result of a cost function that characterizes the quality of the trajectory itself using multiple influencing factors under different weights. For example, the trajectory cost is related to factors such as the length of the path, driving speed, curvature, etc. The larger the trajectory cost parameter, the worse the performance of the candidate path trajectory. The behavior cost parameter refers to the result of a cost function that characterizes the rationality and stability of the vehicle's driving behavior over continuous time using multiple influencing factors under different weights. The behavior cost needs to consider the consistency of the trajectory behavior between the candidate path trajectory at the current moment and the actually selected trajectory at the previous moment to avoid frequent changes. The larger the behavior cost parameter, the worse the performance of the candidate path trajectory.

[0016] S30. Determine the comprehensive cost value of each of the candidate path trajectories according to the trajectory cost parameters and the behavior cost parameters, and determine the target decision trajectory from all the candidate path trajectories according to the comprehensive cost value.

[0017] Understandably, the autonomous driving system calculates a comprehensive cost value for each candidate path trajectory based on the trajectory cost parameter and the behavior cost parameter. It can be calculated by direct summation or by weighted calculation. The comprehensive cost value is a comprehensive consideration result used to characterize the quality of the trajectory itself and the rationality of the vehicle behavior. The autonomous driving system compares and analyzes all candidate path trajectories according to the comprehensive cost value, and finally selects the optimal candidate path trajectory as the target decision trajectory. The target decision trajectory refers to the path trajectory actually executed at the current moment. For example, when the candidate path trajectories at the current moment include a lane-keeping trajectory and a lane-changing trajectory, if their trajectory cost parameters are close, but the actually selected trajectory in the previous moment is the lane-changing trajectory, then the behavior cost parameter of the lane-keeping trajectory is greater at the current moment, and the behavior cost parameter of the lane-changing trajectory is smaller. That is, the lane-changing trajectory is preferably selected as the target decision trajectory at the current moment.

[0018] In this embodiment, at least one candidate path trajectory is determined based on the path planning information, and the trajectory parameters and obstacle parameters corresponding to each candidate path trajectory are obtained; the trajectory cost parameters and behavior cost parameters of each candidate path trajectory are determined according to the trajectory parameters and obstacle parameters; the comprehensive cost value of each candidate path trajectory is determined according to the trajectory cost parameter and the behavior cost parameter, and the target decision trajectory is determined from all candidate path trajectories according to the comprehensive cost value. This embodiment comprehensively considers the trajectory cost at the spatial level and the behavior cost at the time level, obtains the comprehensive cost value using the trajectory cost parameter and the behavior cost parameter under the same priority, realizes the spatio-temporal joint planning, ensures the comprehensiveness of the cost scoring, and improves the flexibility and accuracy of the planning decision. At the same time, after determining the candidate path trajectory based on the path planning information in this embodiment, in addition to the trajectory parameters, the obstacle parameters are also considered, which improves the rationality of obstacle bypassing. This embodiment evaluates the cost using comprehensive elements and screens out the target decision trajectory, achieving a balance in terms of decision-making efficiency, accuracy, stability, and safety.

[0019] In one embodiment, the path planning information includes an obstacle bypass list and map navigation information; as Figure 2 shown, in step S10, that is, determining at least one candidate path trajectory based on the path planning information includes: S101. Perform data analysis on the obstacle bypass list and map navigation information through a preset heuristic model to obtain at least one initial planned trajectory; S102. Perform trajectory search processing on each initial planned trajectory through a preset search model to obtain the candidate path trajectory and the trajectory parameters and obstacle parameters corresponding to each candidate path trajectory.

[0020] Understandably, the path planning information includes an obstacle detour list and map navigation information. The obstacle detour list refers to a list used to record obstacles that need to be detoured within a road or driving area. The map navigation information refers to various map data and guidance related to navigation. The autonomous driving system divides path planning into two stages: heuristic and search. In the heuristic stage, data analysis is performed on the obstacle detour list and map navigation information through a preset heuristic model to obtain at least one initial planning trajectory. The preset heuristic model is a pre-constructed mathematical model used to plan the possible driving trajectories of the vehicle during the decision-making frame at the current moment using the input information. The initial planning trajectory refers to a rough driving trajectory output by the preset heuristic model, and the initial planning trajectory lacks lateral detail information. In the search stage, trajectory search processing is performed on each initial planning trajectory through a preset search model to obtain candidate path trajectories, as well as trajectory parameters and obstacle parameters corresponding to each candidate path trajectory. The preset search model is a pre-constructed mathematical model used to refine the feasible kinematic trajectories of the initial planning trajectories. The candidate path trajectory is a refined path driving plan, including the recommended driving direction, the specific route of the driving path, traffic signs and signals that need attention, etc.

[0021] In a specific embodiment, the preset heuristic model adopts a cross-lane speed dynamic programming algorithm, and the preset search model adopts a hybrid search framework combining the Hybrid A Star algorithm and the Monte Carlo Tree Search (MCTS). In the heuristic stage, the cross-lane speed dynamic programming is used to give possible rough trajectories to obtain the initial planning trajectories, and then in the search stage, the hybrid search framework is used to perform feasible kinematic trajectory search processing on each initial planning trajectory to obtain candidate path trajectories. Among them, in the heuristic stage, it is necessary to combine the obstacle detour list and map navigation information to jointly plan possible rough trajectories such as lane changes, lane keeping, and detouring through borrowed lanes. For example, when the map navigation information shows the existence of an obstacle, it is necessary to combine the obstacle detour list to determine whether the initial planning trajectory is a lane-keeping trajectory following the vehicle or a detouring trajectory through a borrowed lane. If the obstacle also exists in the obstacle detour list, it indicates that it is necessary to detour through the obstacle, and at this time, the initial planning trajectory is a detouring trajectory through a borrowed lane. If the obstacle detour list is empty or the obstacle does not exist, it indicates that it is not necessary to detour through the obstacle, and at this time, the initial planning trajectory is a lane-keeping trajectory following the obstacle.

[0022] By combining the heuristic model and the search model, this embodiment can effectively determine a set of feasible candidate path trajectories on the basis of considering obstacle detours and map navigation information, reflecting the idea from preliminary planning to fine planning, and helping to quickly and accurately plan a safe and feasible driving path.

[0023] In one embodiment, the trajectory parameters include the trajectory point speed, the trajectory point potential field coordinates, the trajectory duration, the trajectory lane-changing distance, and the trajectory lane-changing time, and the obstacle parameters include the obstacle potential field coordinates; as Figure 3 shown, in step S20, that is, determining the trajectory cost parameters of each of the candidate path trajectories according to the trajectory parameters and the obstacle parameters includes: S201. Obtain the speed limit of the trajectory point, and determine the trajectory point speed cost according to the speed limit of the trajectory point and the trajectory point speed; S202. Determine the trajectory point safety cost according to the obstacle potential field coordinates and the trajectory point potential field coordinates; S203. Determine the trajectory duration cost according to a preset trajectory duration threshold and the trajectory duration; S204. Determine the lane-changing distance cost according to a preset lane-changing distance threshold and the trajectory lane-changing distance; S205. Determine the lane-changing time cost according to a preset lane-changing time threshold and the trajectory lane-changing time; S206. Determine the trajectory cost parameters according to the trajectory point speed cost, the trajectory point safety cost, the trajectory duration cost, the lane-changing distance cost, and the lane-changing time cost.

[0024] Understandably, each candidate path trajectory is composed of multiple trajectory points, and the time interval between two adjacent trajectory points is preset. For example, the trajectory point time interval is default set to 0.1 second. The trajectory durations planned in different scenarios and road conditions are different. The trajectory duration refers to the continuous duration of the candidate path trajectory starting from the current moment. For example, the trajectory duration in the vehicle driving scenario is 6 seconds. Starting from the current moment as 0 second, with one trajectory point every 0.1 second, the candidate path trajectory includes 60 trajectory points. Another example is that the trajectory duration in the vehicle parking scenario is 2 seconds. Starting from the current moment as 0 second, the candidate path trajectory only has 20 trajectory points.

[0025] In one embodiment, the trajectory cost parameters include the costs of various influencing factors. In the same candidate path trajectory, each trajectory point has its corresponding speed, acceleration, distance, and other information, and the trajectory cost parameters such as the trajectory point speed cost, the trajectory point safety cost, the trajectory duration cost, the lane-changing distance cost, and the lane-changing time cost can be calculated. When it comes to the influencing factors of the trajectory points, the costs of each trajectory point can be calculated first and then the average value is taken as the trajectory cost parameter of the entire candidate path trajectory. In addition, each influencing factor corresponds to a different weight, and the weights corresponding to different influencing factors are preset based on empirical data and road tests. For example, the trajectory point speed cost needs to consider the corresponding trajectory point speed weight, the trajectory duration cost needs to consider the corresponding trajectory duration weight, and the lane-changing distance cost needs to consider the corresponding lane-changing distance weight.

[0026] Based on the speed limit of the trajectory point and the speed of the trajectory point, the speed cost of the trajectory point can be determined. The speed limit of the trajectory point refers to the speed limit value of the road where each trajectory point is located in the candidate path trajectory, and the speed of the trajectory point refers to the predicted speed value of each trajectory point in the candidate path trajectory. The speed cost of the trajectory point is a cost value used to characterize the influence degree of the speed of the trajectory point on the quality of the candidate path trajectory itself. Specifically, when calculating the speed cost of each trajectory point, calculate the difference between the speed limit of the trajectory point and the speed of the trajectory point, and judge whether the difference is a positive number. If the difference is a positive number, it means that the speed is low and the speed limit is not reached. At this time, directly use the difference multiplied by the weight corresponding to the speed cost of the trajectory point as the speed cost of the single trajectory point. If the difference is negative, it means that the trajectory point is speeding. At this time, use the square of the difference multiplied by the weight corresponding to the speed cost of the trajectory point as the speed cost of the single trajectory point. Finally, after calculating the speed cost of each trajectory point in the candidate path trajectory, take the average value as the speed cost of the entire candidate path trajectory. The role of the speed cost of the trajectory point is to avoid the vehicle from speeding and driving too slowly.

[0027] When there are no obstacles in the driving area of the candidate path trajectory, the safety cost of the trajectory point is 0. When there are no obstacles in the driving area of the candidate path trajectory, it is necessary to establish an artificial potential field by using the vehicle polygon and the obstacle polygon, and determine the safety cost of the trajectory point based on the obstacle potential field coordinates and the trajectory point potential field coordinates. The obstacle is represented by a rectangular frame, and the obstacle coordinate system as shown in Figure 4 is established. Among them, the coordinate origin O represents the center of the obstacle, the inner small rectangular frame represents the obstacle, and the outer large rectangular frame represents the dangerous range of the obstacle. At the same time, a third rectangular frame can also be used to represent the vehicle of each trajectory point (not shown in the figure). The obstacle potential field coordinates refer to the fixed-point coordinates used to characterize the range of the obstacle in the obstacle coordinate system, specifically the fixed-point coordinates of the four vertices of the rectangular frame corresponding to the obstacle. The trajectory point potential field coordinates refer to the fixed-point coordinates used to characterize the range of the vehicle itself in the obstacle coordinate system, specifically the fixed-point coordinates of the four vertices of the rectangular frame corresponding to the vehicle. The safety cost of the trajectory point is a cost value used to characterize the influence degree of the distance between the trajectory point vehicle and the obstacle on the quality of the candidate path trajectory itself. The linear difference of each fixed-point coordinate calculated based on the obstacle coordinate system can be used as the potential field cost, and the potential field cost is represented by a value between 0 and 1. If any point of the vehicle rectangular frame falls within the inner small rectangular frame of the obstacle ( Figure 4 ), the potential field cost is the ceiling value 1. If all points of the vehicle rectangular frame fall within the dangerous range of the obstacle ( Figure 4Specifically, the potential field cost is calculated based on the dangerous range rectangular box of the obstacle using the four fixed points of the potential field coordinates of the trajectory point, and then the potential field cost is calculated based on the rectangular box of the vehicle using the four fixed points of the potential field coordinates of the obstacle. Finally, the maximum value of all potential field costs multiplied by the weight corresponding to the safety cost of the trajectory point is selected as the safety cost of the trajectory point. That is, the closer the vehicle is to the obstacle, the greater the potential field cost.

[0028] Furthermore, when there is a trajectory point in the candidate path trajectory whose potential field cost is capped at 1, it is necessary to determine whether the trajectory point moment of the trajectory point is within 3 seconds from the current moment. 3 seconds is the preset safety duration, which can be adjusted as needed. If it is within 3 seconds, it means that the candidate path trajectory is relatively dangerous. At this time, the safety cost of the trajectory point will be multiplied by "(3-t)" and then multiplied by the preset amplification weight to obtain the enhanced and updated safety cost of the trajectory point, in order to avoid potential collision risks. Among them, t represents the trajectory point moment when the potential field cost is capped at 1.

[0029] The trajectory duration cost can be determined based on the preset trajectory duration threshold and trajectory duration. The trajectory duration refers to the total duration from the current time as 0 seconds to the end point of the candidate path trajectory. The preset trajectory duration threshold refers to the pre-set maximum critical value used to determine whether the trajectory duration meets the requirements. For example, the default value of the preset trajectory duration threshold can be set to 6 seconds, and can also be adjusted as needed. The trajectory duration cost is not reflected on the trajectory point. The difference between the preset trajectory duration threshold and the trajectory duration is directly calculated and multiplied by the weight corresponding to the trajectory duration cost as the trajectory duration cost. The trajectory duration cost is a cost value used to characterize the degree of influence of the trajectory duration on the quality of the candidate path trajectory itself. Since trajectories with too short trajectory duration are not very reliable (such as encountering obstacles or exceeding the limit of movement in a short time), candidate path trajectories that are too short can be excluded in this way.

[0030] The lane changing distance cost can be determined based on the preset lane changing distance threshold and the trajectory lane changing distance. The trajectory lane changing distance refers to the distance between the current vehicle position and the solid line of the lane, which can be directly obtained through the map navigation information. The preset lane changing distance threshold refers to the preset maximum critical value used to determine whether the lane changing distance meets the requirements. It can be a preset fixed value (such as 200 meters) or a value related to the vehicle speed (such as the vehicle speed of the current frame multiplied by the converted distance of 10 seconds). The lane changing distance cost is calculated by multiplying the difference between the preset lane changing distance threshold and the trajectory lane changing distance by the weight corresponding to the lane changing distance cost. The lane changing distance cost is a cost value used to characterize the degree of influence of the lane changing distance on the quality of the candidate path trajectory itself, so as to avoid going into the wrong lane or not having enough time to change lanes.

[0031] The lane-changing time cost can be determined based on a preset lane-changing time threshold and the trajectory lane-changing time. The trajectory lane-changing time refers to the time corresponding to when the center point of the rear axle of the vehicle enters the lane to be changed in the candidate path trajectory. The preset lane-changing time threshold refers to the maximum critical value preset for determining whether the lane-changing duration meets the requirements. For example, the default value can be set to 8 seconds. The difference between the preset lane-changing time threshold and the trajectory lane-changing time is calculated and then multiplied by the weight corresponding to the lane-changing time cost as the lane-changing time cost. The lane-changing time cost is a cost value used to characterize the influence degree of the trajectory lane-changing time on the quality of the candidate path trajectory itself, which can ensure that the vehicle enters the lane to be changed as early as possible during lane-changing.

[0032] This embodiment can accurately calculate the trajectory cost parameters by using the trajectory parameters and obstacle parameters, comprehensively considering multiple factors such as speed, collision safety, trajectory duration, lane-changing distance, and lane-changing time, which helps to comprehensively evaluate the quality of the candidate path trajectory in autonomous driving.

[0033] In one embodiment, the trajectory parameters further include the trajectory point acceleration and the trajectory point rotation angle; as Figure 5 shown, in step S20, that is, after determining the lane-changing time cost according to the preset lane-changing time threshold and the trajectory lane-changing time, it further includes: S207. Determine the trajectory point acceleration cost according to the trajectory point acceleration; S208. Obtain the trajectory point acceleration change rate according to the trajectory point acceleration, and determine the trajectory point acceleration change cost according to the trajectory point acceleration change rate; S209. Determine the trajectory point curvature according to the trajectory point rotation angle, and determine the trajectory point curvature change cost according to the trajectory point curvature; S210. Obtain the trajectory point centripetal acceleration according to the trajectory point speed and the trajectory point curvature, and determine the trajectory point centripetal acceleration cost according to the trajectory point centripetal acceleration; As Figure 6 shown, in step S206, that is, determining the trajectory cost parameter according to the trajectory point speed cost, trajectory point safety cost, trajectory duration cost, lane-changing distance cost, and lane-changing time cost, includes: S2061. Determine the trajectory cost parameter according to the trajectory point speed cost, trajectory point safety cost, trajectory duration cost, lane-changing distance cost, lane-changing time cost, trajectory point acceleration cost, trajectory point acceleration change cost, trajectory point curvature change cost, and trajectory point centripetal acceleration cost.

[0034] Understandably, the candidate path trajectory may not be traveling at a constant speed and may also involve turning. For the influencing factors related to the trajectory points, the trajectory parameters also include the trajectory point acceleration and the trajectory point turning angle. The trajectory point acceleration refers to the predicted acceleration value of each trajectory point in the candidate path trajectory, and the trajectory point turning angle refers to the front-wheel steering angle value of each trajectory point in the candidate path trajectory. In vehicle dynamics, when the vehicle needs to change its driving direction, the vehicle will generate a lateral acceleration and accordingly change the curvature of the path.

[0035] When determining the trajectory point acceleration cost, the absolute value of the trajectory point acceleration is multiplied by the weight corresponding to the trajectory point acceleration cost to obtain the trajectory point acceleration cost. The trajectory point acceleration cost is a cost value used to characterize the influence degree of the trajectory point acceleration on the quality of the candidate path trajectory itself, and its function is to avoid sudden acceleration and sudden deceleration that affect the driving experience.

[0036] Furthermore, the trajectory point acceleration change rate can be obtained by using the accelerations of two adjacent trajectory points, and the absolute value of the trajectory point acceleration change rate is multiplied by the weight corresponding to the trajectory point acceleration change cost to obtain the trajectory point acceleration change cost. The trajectory point acceleration change cost is a cost value used to characterize the influence degree of the trajectory point acceleration change rate on the quality of the candidate path trajectory itself, and its function is to avoid jerks caused by frequent switching of acceleration and deceleration.

[0037] The trajectory point turning angle can be converted into the trajectory point curvature by using the conversion relationship. Based on the curvatures of two adjacent trajectory points, the trajectory point curvature change rate can be obtained. The absolute value of the trajectory point curvature change rate is multiplied by the weight corresponding to the trajectory point curvature change cost to obtain the trajectory point curvature change cost. The trajectory point curvature change cost is a cost value used to characterize the influence degree of the trajectory point curvature change rate on the quality of the candidate path trajectory itself, and its function is to avoid overly rapid turning of the direction.

[0038] The trajectory point centripetal acceleration can be obtained based on the trajectory point speed and the trajectory point curvature. Specifically, the trajectory point centripetal acceleration is calculated by multiplying the square of the trajectory point speed by the trajectory point curvature. Similarly, the absolute value of the trajectory point centripetal acceleration is multiplied by the weight corresponding to the trajectory point centripetal acceleration cost to obtain the trajectory point centripetal acceleration cost. The trajectory point centripetal acceleration cost is a cost value used to characterize the influence degree of the trajectory point centripetal acceleration on the quality of the candidate path trajectory itself, and its function is to avoid overly rapid acceleration and a large curvature, which may cause the passengers to feel thrown out laterally.

[0039] This embodiment further considers the influence of the trajectory point acceleration, the trajectory point acceleration change rate, the trajectory point curvature change rate, and the trajectory point centripetal acceleration on the driving experience, which helps to improve the flexibility and comfort experience of autonomous driving during decision-making.

[0040] In one embodiment, the trajectory parameters include the lane stability state and the candidate trajectory behavior type, and the obstacle parameters include the relative position information of the obstacle and the following information of the obstacle; as Figure 7 shown, in step S20, that is, determining the behavior cost parameter of each candidate path trajectory according to the trajectory parameter and the obstacle parameter includes: S211. Obtain the historical trajectory behavior type corresponding to the previous moment of the current moment, and determine the behavior jump cost according to the lane stability state, the candidate trajectory behavior type and the historical trajectory behavior type; S212. Determine the overtaking reward cost according to the relative position information of the obstacle; S213. Determine the following stop cost according to the following information of the obstacle; S214. Determine the behavior cost parameter according to the behavior jump cost, the overtaking reward cost and the following stop cost.

[0041] Understandably, the trajectory parameters include the lane stability state and the candidate trajectory behavior type. Among them, the lane stability state is the state information used to characterize whether the motion type needs to be kept unchanged at the current moment compared with the previous moment. The candidate trajectory behavior type is the type of vehicle driving behavior corresponding to the candidate path trajectory, such as the lane keeping type, the left lane change driving type, the right lane borrowing and overtaking type, etc. Different candidate trajectory behavior types can be grouped according to the lane and the detour direction. For example, the left lane change driving type, the left lane borrowing and overtaking type, and the left avoidance type are divided into the same group. The obstacle parameters include the relative position information of the obstacle and the following information of the obstacle. Among them, the relative position information of the obstacle is the relative position relationship information between the candidate path trajectory at the current moment and the obstacle that needs to be detoured compared with the previous moment. For example, the end point of the candidate path trajectory exceeds the obstacle or does not exceed the obstacle. The following information of the obstacle is the motion state information of the obstacle that needs to be followed in the candidate path trajectory at the current moment.

[0042] Before determining the behavior cost parameter, it is necessary to calculate the behavior jump cost, the overtaking reward cost and the following stop cost respectively. First, obtain the historical trajectory behavior type corresponding to the previous moment of the current moment, and determine the behavior jump cost according to the lane stability state, the candidate trajectory behavior type and the historical trajectory behavior type. The current moment refers to the path planning and path decision-making process of this frame, and the previous moment refers to the path planning and path decision-making process of the previous frame. The historical trajectory behavior type is the type of vehicle driving behavior corresponding to the target decision trajectory selected at the previous moment. Under different lane stability states, different relationships between the candidate trajectory behavior type and the historical trajectory behavior type correspond to different behavior jump costs. The behavior jump cost is the cost value used to characterize the influence degree of the behavior change at the current moment compared with the previous moment on the rationality of the candidate path trajectory driving behavior.

[0043] Secondly, it is determined whether the relative position information of the obstacle is that the end point of the candidate path trajectory exceeds the obstacle. If the end point of the candidate path trajectory exceeds the obstacle, the overtaking reward cost is counted. That is, if based on the projection of the longest reference line of the candidate path trajectory, the trajectory end point is greater than the predicted trajectory end point of the obstacle, that is, the vehicle can reach the front of the road direction of the obstacle after the execution of the candidate path trajectory is completed, it indicates that the candidate path trajectory can effectively overtake the obstacle that needed to be bypassed at the previous moment. At this time, the overtaking reward cost is counted. The overtaking reward cost is a value used to characterize the influence degree of the candidate path trajectory's driving behavior rationality when overtaking the obstacle that needed to be bypassed at the previous moment. The overtaking reward cost is a preset fixed value. The overtaking reward cost is of a reward nature, while other costs are of a penalty nature. The greater the overtaking reward cost, the better the performance of the candidate path trajectory, and the total cost value can be reduced. Among them, the judgment based on the projection of the longest reference line of the candidate path trajectory is to avoid misjudgment of the end point of the obstacle and affect the accuracy of the overtaking determination due to special reasons (such as there being an obstacle in the target lane) in the U-turn scenario, resulting in choosing the straight lane reference line instead of the U-turn lane reference line. As Figure 8 shown, "obs" represents the obstacle, "ego" represents the current vehicle, the dashed line represents the U-turn lane reference line, and the solid line represents the straight lane reference line. At this time, based on the U-turn lane reference line, the obstacle appears in front of the vehicle; based on the straight lane reference line, the obstacle appears behind the vehicle. In addition, if the candidate path trajectory passes through the solid line entrance of the intersection traffic light, the overtaking reward cost will be blocked (the overtaking reward cost is counted as 0) to avoid misjudgment in the automatic driving decision at the solid line entrance.

[0044] Finally, the following-following stop cost can be determined based on the obstacle following information. When the candidate path trajectory is lane keeping and following an obstacle, it is necessary to judge whether the motion state of the obstacle is a stationary state according to the obstacle following information. If the motion state of the obstacle is a stationary state, the following-following stop cost is counted. The following-following stop cost is a value used to characterize the influence degree of the candidate path trajectory's driving behavior rationality when following a stationary obstacle. The following-following stop cost is a preset fixed value. The following-following stop cost is of a penalty nature to prevent the vehicle from being forced to stop due to inability to move.

[0045] In this embodiment, not only the behavior jump cost is considered at the behavior cost level, but also the overtaking reward cost and the following-following stop cost are introduced, which helps to effectively improve the flexibility and rationality of the automatic driving decision.

[0046] In one embodiment, as Figure 9 shown, in step S211, that is, determining the behavior jump cost according to the lane stable state, the candidate trajectory behavior type, and the historical trajectory behavior type includes: S2111. Compare the candidate trajectory behavior type with the historical trajectory behavior type, and determine a behavior jump coefficient according to the behavior comparison result; S2112. Determine whether the lane stability state is a forced stability state; S2113. If the lane stability state is a forced stability state, determine the behavior jump cost according to a preset jump base cost, a preset stability coefficient, and the behavior jump coefficient; S2114. If the lane stability state is a non-forced stability state, determine the behavior jump cost according to the preset jump base cost and the behavior jump coefficient.

[0047] Understandably, the lane stability state includes a forced stability state and a non-forced stability state. The forced stability state is a state requirement used to represent that the movement type needs to remain unchanged at the current moment compared with the previous moment, and the non-forced stability state is a state requirement used to represent that the movement type does not need to remain unchanged at the current moment compared with the previous moment. For example, when the target decision trajectory at the previous moment is a type of borrowing a lane to overtake, changing lanes to drive, or withdrawing from changing lanes and the end condition has not been reached, the lane stability state is a forced stability state. Among them, for the type of borrowing a lane to overtake, the end condition is whether the center of the rear axle of the vehicle returns to the original lane after borrowing a lane; for the type of changing lanes to drive, the end condition is whether the center of the rear axle of the vehicle reaches the lane to be changed; for the type of withdrawing from changing lanes, the end condition is that it has been 3 seconds since the vehicle returned to the original lane during the lane-changing process.

[0048] Under different lane stability states, different relationships between the candidate trajectory behavior type and the historical trajectory behavior type correspond to different behavior jump costs. First, compare the candidate trajectory behavior type with the historical trajectory behavior type, and determine a behavior jump coefficient according to the behavior comparison result. The behavior comparison result between the candidate trajectory behavior type and the historical trajectory behavior type is divided into the same group and different groups. For example, when the candidate trajectory behavior type and the historical trajectory behavior type are respectively one or two of the types of changing lanes to the left to drive, borrowing a lane to overtake to the left, and avoiding to the left, the candidate trajectory behavior type and the historical trajectory behavior type are in the same group. The behavior jump coefficient is a coefficient that reflects different behavior comparison results in the calculation of the behavior jump cost. The behavior jump coefficient is a fixed coefficient preset corresponding to different behavior comparison results. Specifically, when the behavior comparison result is the same group, the behavior jump coefficient is 0.5; when the behavior comparison result is different groups, the behavior jump coefficient is 1.

[0049] Then, it is determined whether the lane stability state is a forced stability state. When the lane stability state is a forced stability state, the behavior jump cost is determined according to a preset jump base cost, a preset stability coefficient, and a behavior jump coefficient. The preset jump base cost is a preset fixed value used as the basis for calculating the behavior jump cost. For example, the preset jump base cost is default set to 0.5. The preset stability coefficient refers to a coefficient preset to reflect the forced stability state in the calculation process of the behavior jump cost. For example, the preset stability coefficient is default set to 1.5. When the lane stability state is a non-forced stability state, the behavior jump cost is determined according to the preset jump base cost and the behavior jump coefficient. Specifically, when the lane stability state is a forced stability state, the behavior jump cost with the same behavior comparison result for grouping is 0.5×0.5×1.5, and the behavior jump cost with different behavior comparison results for grouping is 0.5×1×1.5. When the lane stability state is a non-forced stability state, the behavior jump cost with the same behavior comparison result for grouping is 0.5×0.5, and the behavior jump cost with different behavior comparison results for grouping is 0.5×1.

[0050] In this embodiment, the behavior jump cost is larger in the strong stability state and smaller in the non-strong stability state, which can effectively suppress unreasonable frequent left and right lane-changing jumps and improve the decision-making stability and rationality of autonomous driving.

[0051] In one embodiment, as Figure 10 shown, in step S30, that is, determining the comprehensive cost value of each of the candidate path trajectories according to the trajectory cost parameter and the behavior cost parameter, and determining the target decision trajectory from all the candidate path trajectories includes: S301. Perform a weighted calculation process on the trajectory cost parameter and the behavior cost parameter of each of the candidate path trajectories, and determine the weighted calculation result as the comprehensive cost value; S302. Sort the comprehensive cost values of all the candidate path trajectories according to their magnitudes, and determine the candidate path trajectory with the smallest comprehensive cost value as the target driving trajectory.

[0052] Understandably, perform a weighted calculation process on the trajectory cost parameter and the behavior cost parameter of each candidate path trajectory according to a preset weight, and determine the weighted calculation result as the comprehensive cost value. Sort the comprehensive cost values of all the candidate path trajectories according to their magnitudes, and determine the candidate path trajectory with the smallest comprehensive cost value as the target driving trajectory.

[0053] In this embodiment, the comprehensive cost value is obtained by combining the two costs at the trajectory level and the behavior level in the same priority, which can realize the costization of all influencing factors, give full play to the ability of parallel computing, avoid the problem of particularly rigid decision-making in the prior art, and ensure the comprehensiveness, flexibility, and accuracy of the decision-making.

[0054] In one embodiment, as Figure 11 shown, in step S30, that is, after determining the target decision trajectory from all the candidate path trajectories according to the comprehensive cost value, the following steps are further included: S303. After confirming that the target decision trajectory corresponding to the current moment is executed, if a target obstacle is detected, obtain an obstacle bypass recommendation value corresponding to the target obstacle, update the obstacle bypass list corresponding to the current moment according to the obstacle bypass recommendation value, and determine the path planning information for the next moment based on the updated obstacle bypass list.

[0055] Understandably, after confirming that the target decision trajectory corresponding to the current moment is executed and before starting the path planning and path decision process for the next moment, it is necessary to update the obstacle bypass list corresponding to the current moment and determine the path planning information for the next moment based on the updated obstacle bypass list. First, determine whether there is a target obstacle in the target decision trajectory at the current moment. The target obstacle refers to an obstacle within the driving range of the target decision trajectory, such as a range of 200 meters in front of the current lane and the adjacent left and right lanes of the target decision trajectory. When a target obstacle is detected, obtain the obstacle bypass recommendation value corresponding to the target obstacle and update the obstacle bypass list corresponding to the current moment according to the obstacle bypass recommendation value. The obstacle bypass recommendation value is used to represent the recommended degree of whether the obstacle needs to be bypassed in the path planning for the next moment. For example, the obstacle bypass recommendation value is a value from 0 to 1. When the obstacle bypass recommendation value is 1, it indicates that the obstacle needs to be bypassed in the path planning for the next moment. At the first moment when the vehicle turns on the automatic driving function, the obstacle bypass list is empty, and the path planning information starts to be determined using the obstacle bypass list updated in the previous moment from the second moment.

[0056] This embodiment introduces an obstacle bypass list. After confirming that the target decision trajectory corresponding to the current moment is executed, updating the obstacle bypass list and determining the path planning information for the next moment based on the updated obstacle bypass list helps the automatic driving system consider the factor of obstacle bypass in the path planning process, improve the rationality of obstacle bypass, effectively avoid obstacles, and ensure the safety of automatic driving.

[0057] In one embodiment, as Figure 12 shown, in step S303, that is, obtaining the obstacle bypass recommendation value corresponding to the target obstacle and updating the obstacle bypass list corresponding to the current moment according to the obstacle bypass recommendation value includes: S3031. Obtain the obstacle speed information and obstacle distance information of the target obstacle, and the vehicle speed information when the target decision trajectory is executed. S3032. Determine whether the target obstacle meets a preset bypass condition according to the vehicle speed information, obstacle speed information, and obstacle distance information; S3033. If the target obstacle meets the preset bypass condition, obtain the obstacle motion state and obstacle cumulative frame number of the target obstacle, and determine the obstacle bypass recommendation value of the target obstacle according to the obstacle motion state and obstacle cumulative frame number; S3034. If the target obstacle does not exist in the obstacle bypass list and the obstacle bypass recommendation value of the target obstacle is greater than or equal to a first preset bypass threshold, add the target obstacle to the obstacle bypass list to obtain an updated obstacle bypass list; S3035. If the target obstacle exists in the obstacle bypass list and the obstacle bypass recommendation value of the target obstacle is less than or equal to a second preset bypass threshold, remove the target obstacle from the obstacle bypass list to obtain an updated obstacle bypass list; the first preset bypass threshold is greater than the second preset bypass threshold.

[0058] Understandably, in the process of updating the obstacle bypass list corresponding to the current moment using the obstacle bypass recommendation value, it is necessary to first determine the obstacle bypass recommendation value through conditional judgment. First, obtain the obstacle speed information and obstacle distance information of the target obstacle, as well as the vehicle speed information when the target decision trajectory is executed. The obstacle speed information refers to the speed at which the obstacle moves, and the obstacle distance information refers to the distance between the obstacle and the vehicle when the target decision trajectory corresponding to the current moment is executed.

[0059] Then, based on the vehicle speed information, obstacle speed information, and obstacle distance information, it is determined whether the target obstacle meets the preset bypass condition. The preset bypass condition refers to the condition information preset for determining whether the target obstacle needs to start calculating the obstacle bypass recommendation value. The preset bypass condition includes a speed condition and a time-distance condition. When both the speed condition and the time-distance condition are met, it is confirmed that the target obstacle meets the preset bypass condition. Specifically, on the one hand, it is judged whether the obstacle speed information is less than the vehicle speed information. If the obstacle speed information is less than the vehicle speed information, the speed condition is met; if the obstacle speed information is greater than or equal to the vehicle speed information, the speed condition is not met. On the other hand, based on the vehicle speed information and the obstacle distance information, the time-distance information between the obstacle and the vehicle is calculated, and it is judged whether the time-distance information is less than the preset time-distance threshold. If the time-distance information is less than the preset time-distance threshold, the time-distance condition is met; if the time-distance information is greater than or equal to the preset time-distance threshold, the time-distance condition is not met. The preset time-distance threshold is the maximum time-distance critical value preset, for example, the preset time-distance threshold is default set to 4 seconds. When the target obstacle does not meet the preset bypass condition, the obstacle bypass recommendation value will not be calculated and will not be added to the obstacle bypass list.

[0060] Next, when the target obstacle meets the preset bypass condition, the obstacle motion state and the obstacle cumulative frame number of the target obstacle are obtained, and the obstacle bypass recommendation value of the target obstacle is determined according to the obstacle motion state and the obstacle cumulative frame number. The obstacle motion state refers to the motion mode and direction information of the obstacle, and the obstacle cumulative frame number refers to the number of frames accumulated from the moment when the obstacle bypass recommendation value starts to be calculated to the current moment, that is, the obstacle cumulative frame number is updated at each moment after the obstacle bypass recommendation value starts to be calculated. According to the preset accumulation rule, the obstacle motion state and the obstacle cumulative frame number are calculated to obtain the obstacle bypass recommendation value of the target obstacle. The preset accumulation rule is a calculation rule preset for calculating the obstacle bypass recommendation value by using the frame number accumulation method corresponding to different obstacle motion states. Specifically, the obstacle motion state includes a stationary state, a same-direction motion state, and a reverse-direction motion state. When the obstacle motion state is the same-direction motion state, the obstacle cumulative frame number is accumulated by 6 frames (each frame accumulates 0.17), that is, the obstacle and the vehicle move in the same direction, and the obstacle bypass recommendation value starts to accumulate 0.17 at each moment from 0, and exceeds 1 after six accumulations. When the obstacle motion state is the stationary and reverse-direction motion states, the obstacle cumulative frame number is accumulated by 3 frames (each frame accumulates 0.34), that is, the obstacle and the vehicle move in the reverse direction, and the obstacle bypass recommendation value starts to accumulate 0.34 at each moment from 0, and exceeds 1 after three accumulations. In addition, if the target obstacle accelerates during the accumulation period, the addition and accumulation will be abandoned, and instead, the obstacle bypass recommendation value will be updated in a way that decreases by 0.2 per frame.

[0061] Finally, it is determined whether there is a target obstacle in the obstacle bypass list, and how to update the obstacle bypass list is determined according to the obstacle bypass recommendation value. When there is no target obstacle in the obstacle bypass list, and the obstacle bypass recommendation value of the target obstacle is greater than or equal to the first preset bypass threshold, the target obstacle is added to the obstacle bypass list to obtain an updated obstacle bypass list. When there is a target obstacle in the obstacle bypass list, and the obstacle bypass recommendation value of the target obstacle is less than or equal to the second preset bypass threshold, the target obstacle is removed from the obstacle bypass list to obtain an updated obstacle bypass list. Among them, the first preset bypass threshold is greater than the second preset bypass threshold. The first preset bypass threshold is a preset upper limit of the obstacle bypass recommendation value indicating that the target obstacle needs to be bypassed, for example, the first preset bypass threshold is set to 1. The second preset bypass threshold is a preset lower limit of the obstacle bypass recommendation value indicating that the target obstacle does not need to be bypassed, for example, the second preset bypass threshold is set to 0. In addition, if it is confirmed that the target decision trajectory corresponding to the current moment has been executed, and it is detected according to the map navigation information that the vehicle is located in the solid line segment and the light ahead is red, the obstacle detour request list is directly cleared.

[0062] This embodiment introduces an obstacle bypass accumulation trigger mechanism when updating the obstacle bypass request list, which can avoid bypassing false obstacles that only flash for one frame, and decisively bypass static or reverse obstacles. At the same time, it is appropriately conservative when bypassing dynamic obstacles moving in the same direction, controls the trigger sensitivity of the bypass, and improves the rationality of the obstacle bypass.

[0063] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0064] In one embodiment, an automatic driving decision system is provided, which corresponds one-to-one to the automatic driving decision method in the above embodiment. The automatic driving decision system includes a candidate trajectory determination module, a cost calculation module and a target trajectory decision module. Each functional module is described in detail as follows: A candidate trajectory determination module, used to determine at least one candidate path trajectory based on the path planning information, and obtain trajectory parameters and obstacle parameters corresponding to each of the candidate path trajectories; A cost calculation module, used to determine the trajectory cost parameter and the behavior cost parameter of each candidate path trajectory according to the trajectory parameter and the obstacle parameter; The target trajectory decision module is used to determine the comprehensive cost value of each candidate path trajectory according to the trajectory cost parameter and the behavior cost parameter, and determine the target decision trajectory from all the candidate path trajectories according to the comprehensive cost value.

[0065] For the specific limitations of the autonomous driving decision-making system, reference can be made to the limitations of the autonomous driving decision-making method in the foregoing text, which will not be elaborated here. Each module in the above autonomous driving decision-making system can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to the above modules.

[0066] In one embodiment, as Figure 13 shown, an autonomous driving decision-making device is provided. The autonomous driving decision-making device includes a memory 10 and a processor 20 that are communicatively connected. Among them, the memory 10 is used to store a computer program, and the processor 20 is used to execute the computer program stored on the memory to implement the above autonomous driving decision-making method. The autonomous driving decision-making device can be implemented in whole or in part by software, hardware, and their combination. The autonomous driving decision-making device can be a vehicle, a controller, or a vehicle head unit, etc. When the autonomous driving decision-making device is a controller, the controller can only include a vehicle controller, i.e., a vehicle control unit (VCU), or only include a vehicle driving component controller, or can also include a vehicle controller and a vehicle driving component controller that are communicatively connected. The controller can be a computer device, a microprocessor, or a microcontroller, and can implement various complex control strategies. For the specific limitations of the controller, reference can be made to the limitations of the autonomous driving decision-making method in the foregoing text, which will not be elaborated here.

[0067] In one embodiment, the processor of the autonomous driving decision-making device is used to implement the autonomous driving decision-making method introduced in any embodiment of the present application.

[0068] In one embodiment, one or more computer-readable storage media storing computer-readable instructions are provided. The readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage media. Computer-readable instructions are stored on the readable storage media, and when the computer-readable instructions are executed by one or more processors, the autonomous driving decision-making method introduced in any embodiment of the present application is implemented.

[0069] It can be completed by instructing relevant hardware through computer-readable instructions. The computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When the computer-readable instructions are executed, they can include the processes of the embodiments of the above-mentioned various methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0070] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0071] The above-mentioned embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. An autonomous driving decision-making method, characterized in that, Including: Determine at least one candidate path trajectory based on path planning information, and obtain trajectory parameters and obstacle parameters corresponding to each of the candidate path trajectories; Determine the trajectory cost parameter and behavior cost parameter of each of the candidate path trajectories according to the trajectory parameters and obstacle parameters; Determine the comprehensive cost value of each of the candidate path trajectories according to the trajectory cost parameter and behavior cost parameter, and determine the target decision trajectory from all the candidate path trajectories according to the comprehensive cost value.

2. The autonomous driving decision-making method according to claim 1, wherein The path planning information includes an obstacle bypass list and map navigation information; The determining at least one candidate path trajectory based on path planning information includes: Performing data analysis on the obstacle bypass list and map navigation information through a preset heuristic model to obtain at least one initial planned trajectory; Performing trajectory search processing on each of the initial planned trajectories through a preset search model to obtain the candidate path trajectories and the trajectory parameters and obstacle parameters corresponding to each of the candidate path trajectories.

3. The autonomous driving decision-making method according to claim 1, characterized in that The trajectory parameters include trajectory point speed, trajectory point potential field coordinates, trajectory duration, trajectory lane change distance, and trajectory lane change time, and the obstacle parameters include obstacle potential field coordinates; The determining the trajectory cost parameter of each of the candidate path trajectories according to the trajectory parameters and obstacle parameters includes: Obtain the speed limit of the trajectory point, and determine the trajectory point speed cost according to the speed limit of the trajectory point and the trajectory point speed; Determine the trajectory point safety cost according to the obstacle potential field coordinates and the trajectory point potential field coordinates; Determine the trajectory duration cost according to a preset trajectory duration threshold and the trajectory duration; Determine the lane change distance cost according to a preset lane change distance threshold and the trajectory lane change distance; Determine the lane change time cost according to a preset lane change time threshold and the trajectory lane change time; Determine the trajectory cost parameter according to the trajectory point speed cost, trajectory point safety cost, trajectory duration cost, lane change distance cost, and lane change time cost.

4. The autonomous driving decision-making method according to claim 3, wherein The trajectory parameters further include trajectory point acceleration and trajectory point turning angle; After determining the lane change time cost according to the preset lane change time threshold and the trajectory lane change time, it further includes: Determine the trajectory point acceleration cost according to the trajectory point acceleration; Obtain the trajectory point acceleration change rate according to the trajectory point acceleration, and determine the trajectory point acceleration change cost according to the trajectory point acceleration change rate; Determine the trajectory point curvature according to the trajectory point turning angle, and determine the trajectory point curvature change cost according to the trajectory point curvature; Obtain the trajectory point centripetal acceleration according to the trajectory point speed and the trajectory point curvature, and determine the trajectory point centripetal acceleration cost according to the trajectory point centripetal acceleration; The determining the trajectory cost parameter according to the trajectory point speed cost, trajectory point safety cost, trajectory duration cost, lane change distance cost, and lane change time cost includes: Determine the trajectory cost parameter according to the trajectory point speed cost, trajectory point safety cost, trajectory duration cost, lane change distance cost, lane change time cost, trajectory point acceleration cost, trajectory point acceleration change cost, trajectory point curvature change cost, and trajectory point centripetal acceleration cost.

5. The automatic driving decision-making method according to claim 1, wherein The trajectory parameters include the lane stability state and the candidate trajectory behavior type, and the obstacle parameters include the relative position information of the obstacle and the following information of the obstacle; Determining the behavior cost parameters of each of the candidate path trajectories according to the trajectory parameters and the obstacle parameters includes: Obtaining the historical trajectory behavior type corresponding to the previous moment of the current moment, and determining the behavior jump cost according to the lane stability state, the candidate trajectory behavior type, and the historical trajectory behavior type; Determining the overtaking reward cost according to the relative position information of the obstacle; Determining the following stop cost according to the following information of the obstacle; Determining the behavior cost parameter according to the behavior jump cost, the overtaking reward cost, and the following stop cost.

6. The autonomous driving decision-making method according to claim 5, wherein Determining the behavior jump cost according to the lane stability state, the candidate trajectory behavior type, and the historical trajectory behavior type includes: Comparing the candidate trajectory behavior type with the historical trajectory behavior type, and determining the behavior jump coefficient according to the behavior comparison result; Judging whether the lane stability state is a forced stability state; If the lane stability state is a forced stability state, determining the behavior jump cost according to a preset jump base cost, a preset stability coefficient, and the behavior jump coefficient; If the lane stability state is a non-forced stability state, determining the behavior jump cost according to a preset jump base cost and the behavior jump coefficient.

7. The automatic driving decision-making method according to claim 1, wherein Determining the comprehensive cost value of each of the candidate path trajectories according to the trajectory cost parameter and the behavior cost parameter, and determining the target decision trajectory from all the candidate path trajectories according to the comprehensive cost value includes: Performing a weighted calculation process on the trajectory cost parameter and the behavior cost parameter of each of the candidate path trajectories, and determining the weighted calculation result as the comprehensive cost value; Sorting the comprehensive cost values of all the candidate path trajectories according to their magnitudes, and determining the candidate path trajectory with the smallest comprehensive cost value as the target driving trajectory.

8. The autonomous driving decision-making method according to claim 1, wherein, After determining the target decision trajectory from all the candidate path trajectories according to the comprehensive cost value, it further includes: After confirming that the target decision trajectory corresponding to the current moment is executed, if a target obstacle is detected, obtaining the obstacle detour recommendation value corresponding to the target obstacle, updating the obstacle detour list corresponding to the current moment according to the obstacle detour recommendation value, and determining the path planning information for the next moment based on the updated obstacle detour list.

9. The autonomous driving decision-making method according to claim 8, wherein Obtaining the obstacle detour recommendation value corresponding to the target obstacle, and updating the obstacle detour list corresponding to the current moment according to the obstacle detour recommendation value includes: Obtaining the obstacle speed information and the obstacle distance information of the target obstacle, and the vehicle speed information when the target decision trajectory is executed; Judging whether the target obstacle meets the preset detour condition according to the vehicle speed information, the obstacle speed information, and the obstacle distance information; If the target obstacle meets the preset detour condition, obtaining the obstacle motion state and the obstacle cumulative frame number of the target obstacle, and determining the obstacle detour recommendation value of the target obstacle according to the obstacle motion state and the obstacle cumulative frame number; If the target obstacle does not exist in the obstacle detour list, and the obstacle detour recommendation value of the target obstacle is greater than or equal to the first preset detour threshold, then add the target obstacle to the obstacle detour list to obtain an updated obstacle detour list; If the target obstacle exists in the obstacle detour list, and the obstacle detour recommendation value of the target obstacle is less than or equal to the second preset detour threshold, then remove the target obstacle from the obstacle detour list to obtain an updated obstacle detour list; the first preset detour threshold is greater than the second preset detour threshold.

10. An automatic driving decision-making device, characterized in that, It includes a processor and a memory, wherein, The memory is used to store a computer program; The processor is used to execute the computer program stored on the memory to implement the autonomous driving decision-making method according to any one of claims 1-9.

Images