An autonomous vehicle and a decision-making method thereof

Abstract

The application relates to an automatic driving vehicle and a decision-making method thereof, and belongs to the technical field of automatic driving vehicles. The method comprises the following steps: when there is an obstacle in front, planning a lane keeping track and a lane borrowing track; determining an optimal driving track according to the track length and the average vehicle speed; when the optimal driving track is the lane borrowing track, if there is a track point with a vehicle speed of 0 in the lane borrowing track or the maximum lateral offset of the vehicle exceeds an offset threshold, a passive lane changing track is generated, and the final optimal driving track is selected from the lane borrowing track and the passive lane changing track; when the optimal driving track is the lane keeping track, a passive lane changing track is generated, and the final optimal driving track is selected from the passive lane changing track and the lane keeping track. The decision-making method is accurate and efficient, and can meet the real-time requirements of the automatic driving vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to an automatic driving vehicle and a decision-making method thereof, and belongs to the technical field of automatic driving vehicles. BACKGROUND

[0002] In a complex urban road scene, lane-level path planning and navigation guidance directly affect the intelligence, safety and comfort of the automatic driving system. Based on the requirement that the automatic driving system must rely on clear, accurate and continuously changing navigation information to complete the automatic driving task, the automatic driving vehicle needs to arbitrate multiple source intentions such as navigation lane changing based on global tasks and overtaking lane changing based on local scene perception in real time, so as to make more reasonable and intelligent automatic driving behavior decisions.

[0003] Therefore, some people have proposed a decision-making method suitable for an automatic driving vehicle, for example: a Chinese patent application for invention with the application publication number CN110673602A, which discloses a reinforcement learning model, a method for vehicle automatic driving decision-making and a vehicle-mounted device. The decision-making method of automatic driving adopts a reinforcement learning model for decision-making. The reinforcement learning model is obtained by training a data set, including an evaluation network, a control network and a policy network. After the reinforcement learning model is trained, the environment information obtained during the vehicle motion process is input to output the decision result of the driving behavior. The vehicle drives according to the decision result.

[0004] However, the method of decision-making based on the reinforcement learning model has the following defects:

[0005] 1. The reinforcement learning model needs to obtain a large amount of data of different scenes for training. The data collection is complex and tedious, and the training data is difficult to cover all working conditions, resulting in low decision-making accuracy of the model;

[0006] 2. Each network in the model is a black box and is difficult to modify. Therefore, based on the low decision-making accuracy of the model, once the decision-making error occurs, the model needs to be retrained, which takes a long time and results in poor real-time decision-making.

[0007] In summary, the existing decision-making method has low accuracy and poor real-time performance. SUMMARY

[0008] The purpose of the present application is to provide an automatic driving vehicle and a decision-making method thereof to solve the problem of low accuracy of the existing decision-making method.

[0009] To achieve the above-mentioned purpose, the present application provides a technical scheme of a decision-making method of an automatic driving vehicle, including the following steps:

[0010] 1) When there is an obstacle in front of the vehicle during driving, a driving track is planned according to the size and position of the obstacle in combination with a map; the driving track includes the position of each track point and the vehicle speed; the planned obstacle-avoiding track includes a lane-keeping track and a lane-changing driving track;

[0011] 2) The optimal driving track is determined according to the track length and the average vehicle speed:

[0012] If the optimal driving track is the lane-changing driving track, it is determined whether the lane-changing driving track meets the driving condition; if there is no track point with a speed of 0 in the lane-changing driving track, and the maximum lateral deviation of the vehicle does not exceed the deviation threshold, the lane-changing driving track is determined as the final optimal driving track; if there is a track point with a speed of 0 in the lane-changing driving track, or the maximum lateral deviation of the vehicle exceeds the deviation threshold, it is determined whether a passive lane-changing driving track in the same lane-changing direction can be generated; if it can be generated, the final optimal driving track is determined according to the track length and the average vehicle speed of the passive lane-changing driving track and the lane-changing driving track; if it cannot be generated, the lane-changing driving track is determined as the final optimal driving track; and driving is performed according to the final optimal driving track.

[0013] If the optimal driving track is the lane-keeping track, it is determined whether a passive lane-changing driving track can be generated; if it can be generated, the final optimal driving track is determined according to the track length and the average vehicle speed of the passive lane-changing driving track and the lane-keeping track; if it cannot be generated, the lane-keeping track is the final optimal driving track; and driving is performed according to the final optimal driving track.

[0014] In addition, the present application also provides a technical scheme of an autonomous vehicle, the autonomous vehicle comprising a vehicle body, a positioning module and a perception module arranged on the vehicle body, and a controller, the controller comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, the processor implementing the decision-making method of the autonomous vehicle described above when executing the computer program.

[0015] The automatic driving vehicle and the decision-making method thereof have the advantages that: in the case that there is an obstacle in front of the vehicle, a lane keeping track and a lane borrowing track are planned, and the two tracks are sorted according to the track length and the average speed to select the optimal driving track; when the lane borrowing track is the optimal driving track, it is further judged whether the optimal driving track meets the driving condition, if not, a passive lane changing driving track is generated, and then the selection of the optimal driving track is performed; when the lane keeping track is the optimal driving track, a passive lane changing driving track is generated, and then the selection of the optimal driving track is performed, and finally the vehicle drives according to the final optimal driving track. The application has high flexibility and strong adaptability, and the driving scene can be added or deleted according to the actual situation; the decision is accurate and efficient, and can meet the real-time requirements of the automatic driving vehicle.

[0016] Further, in the automatic driving vehicle and the decision-making method thereof, when the vehicle drives according to the lane borrowing track, after passing the obstacle, a return original lane track is generated, the final optimal driving track is determined according to the track length and the average speed of the return original lane track and the lane borrowing track, and the vehicle drives according to the final optimal driving track.

[0017] Further, in the automatic driving vehicle and the decision-making method thereof, when the vehicle drives according to the lane borrowing track, if the time length of the lane borrowing exceeds the time threshold, a passive lane changing driving track is generated, the final optimal driving track is determined according to the track length and the average speed of the passive lane changing driving track and the lane borrowing track, and the vehicle drives according to the final optimal driving track.

[0018] Further, in the automatic driving vehicle and the decision-making method thereof, when the vehicle drives according to the passive lane changing driving track, after the lane changing is completed, a return original lane track is generated, the final optimal driving track is determined according to the track length and the average speed of the return original lane track and the passive lane changing driving track, and the vehicle drives according to the final optimal driving track.

[0019] Further, in the automatic driving vehicle and the decision-making method thereof, the step of determining the optimal driving track according to the track length and the average speed comprises:

[0020] The track with longer track length is the optimal driving track, and if the track lengths are the same, the track with larger average speed is the optimal driving track.

[0021] Further, in the automatic driving vehicle and the decision-making method thereof, the calculation process of the average speed is:

[0022]

[0023] Wherein, vel is the average speed; s i is the position of the i th track point; vi is the speed of the vehicle for the i th trajectory point; len is the length of the trajectory, and n is the total number of trajectory points.

[0024] Further, in the automatic driving vehicle and the decision method thereof, when there is no obstacle in front, the planning of the active driving trajectory is performed according to the position of the destination in combination with the map, and the active driving trajectory includes a lane keeping trajectory and an active lane changing trajectory. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is a flowchart of the decision method of the automatic driving vehicle of the present application;

[0026] Figure 2 is a state switching diagram of the automatic driving vehicle of the present application;

[0027] Figure 3 is a trajectory planning schematic diagram when there is an obstacle in front of the vehicle. DETAILED DESCRIPTION

[0028] Automatic driving vehicle embodiment:

[0029] The main idea of the present application is that, based on the problem of low accuracy and poor real-time performance of the existing method, the present application plans several driving trajectories according to the real-time conditions of vehicle driving, determines the optimal driving trajectory according to the trajectory length and average speed of each driving trajectory, and then judges whether the optimal driving trajectory meets the driving condition, and in the case of meeting the driving condition, drives according to the optimal driving trajectory, and in the case of not meeting the driving condition, generates other trajectories, and then selects the optimal driving trajectory in combination with other estimates, and drives according to the optimal driving trajectory. The present application makes decisions according to real-time road conditions, and has high accuracy and good real-time performance.

[0030] Specifically, the automatic driving vehicle as shown in Figure 1 includes a vehicle body, a positioning module and a perception module arranged on the vehicle body, and a controller, the controller is composed of a prediction module, a high-precision map module, and a global planning module, and a decision module. The positioning module is used for positioning the automatic driving vehicle, the perception module is used for perceiving the obstacle information around the vehicle, the prediction module, the high-precision map module, and the global planning module are used for planning the driving path according to the information of the positioning module and the perception module, the decision module is used for selecting the optimal driving path from the planned path, and outputting the optimal driving path, and then controlling other control and execution modules in the automatic driving vehicle to drive according to the optimal driving path, so that the vehicle can efficiently reach the destination.

[0031] The controller includes a processor, a memory, and a computer program stored in the memory and executable on the processor, and the processor implements the decision method of the automatic driving vehicle when executing the computer program.

[0032] The decision-making method of the autonomous driving vehicle proposed in the present invention switches the driving state of the vehicle according to the driving path determined by the decision module. The driving state of the vehicle is as follows: Figure 2 As shown, it includes lane keeping state, borrowing lane state (also known as passive borrowing lane state), passive lane changing state, returning to original lane state, and active lane changing state. In the lane keeping state, the vehicle drives according to the lane keeping trajectory; in the borrowing lane state, the vehicle drives according to the borrowing lane driving trajectory; in the passive lane changing state, the vehicle drives according to the passive lane changing driving trajectory; in the returning to original lane state, the vehicle drives according to the returning to original lane trajectory; in the active lane changing state, the vehicle drives according to the active lane changing driving trajectory.

[0033] The decision-making method of autonomous driving vehicles mainly includes two situations: there are obstacles in front of the vehicle and there are no obstacles during the vehicle's driving.

[0034] The decision-making process for an obstacle ahead includes the following steps:

[0035] 1) The obstacle avoidance trajectory is planned based on the size and location of the obstacle and the map. The planned obstacle avoidance trajectory includes lane keeping trajectory and lane-taking trajectory.

[0036] Lane keeping trajectory and lane-taking trajectory Figure 3 As shown, the driving trajectory of the borrowing lane is planned according to the position and size of the obstacle in combination with the map. When the obstacle is on the right side of the lane, the driving trajectory of borrowing lane to the left is planned. When the obstacle is on the left side of the lane, the driving trajectory of borrowing lane to the right is planned. When the obstacle is in the middle of the lane line, the driving trajectory of borrowing lane to the left or right is planned.

[0037] The driving trajectory includes the position and speed of each trajectory point. The driving trajectory is expressed as:

[0038] Path:(s i ,v i ),i=0,…,n;

[0039] Among them, s i is the position of the i-th trajectory point; v i is the vehicle speed of the ith trajectory point; n is the total number of trajectory points.

[0040] from Figure 3 It can be seen that the maximum lateral offset lat of the planned borrowing lane trajectory can be determined, and the maximum lateral offset lat is the maximum distance that a trajectory point on the trajectory deviates from the center line of the lane where the obstacle is located.

[0041] At the same time, the trajectory length len and average vehicle speed vel of the planned lane-borrowing driving trajectory can also be calculated.

[0042] The length of the trajectory len is the length of the entire trajectory. If there is a trajectory point that slows down to 0, the length of the trajectory is from the starting position s0 to the position of the trajectory point where the speed planning slows down to 0. If there is no trajectory point that slows down to 0, the length of the trajectory is from the starting position to the position s n of the nth trajectory point, that is, the length of the trajectory is

[0043]

[0044] The calculation method of the average vehicle speed vel is as follows:

[0045]

[0046] When there is a trajectory point that slows down to 0 in the planned trajectory, the average vehicle speed is considered to be 0. When the average vehicle speed is not 0, it is generally considered that the vehicle can safely pass.

[0047] 2) According to the trajectory length and the average vehicle speed of the lane keeping trajectory and the lane borrowing trajectory, the two trajectories are sorted, and the first one is the optimal driving trajectory.

[0048] The specific sorting process is: the trajectory with a longer length is the optimal driving trajectory. If the lengths of the trajectories are the same, the trajectory with a larger average vehicle speed is the optimal driving trajectory.

[0049] The lane keeping state is the default state of the vehicle. After comparing the trajectory length and the average vehicle speed of the lane keeping trajectory and the lane borrowing trajectory, two results can occur. The first result is that the optimal driving trajectory is the lane borrowing trajectory. The second result is that the optimal driving trajectory is the lane keeping trajectory, and then the state of the vehicle is switched according to the optimal driving trajectory.

[0050] When the optimal driving trajectory is the lane borrowing trajectory, it is determined whether the lane borrowing trajectory meets the driving condition:

[0051] If there is no trajectory point with a speed of 0 (that is, a trajectory point that slows down to 0) in the lane borrowing trajectory, and the maximum lateral deviation of the vehicle does not exceed the deviation threshold, the lane borrowing trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the lane borrowing state, and the vehicle drives according to the lane borrowing trajectory.

[0052] If there is a trajectory point with a speed of 0 in the trajectory of the lane borrowing, or the maximum lateral deviation of the vehicle exceeds the deviation threshold, it is determined whether a passive lane changing trajectory in the same direction as the lane borrowing can be generated. If it can be generated, the final optimal driving trajectory is determined according to the trajectory length and average speed of the passive lane changing trajectory and the lane borrowing trajectory. When the lane borrowing trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the lane borrowing state and drives according to the lane borrowing trajectory. When the passive lane changing trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the passive lane changing state and drives according to the passive lane changing trajectory. If it cannot be generated, the lane borrowing trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the lane borrowing state, and drives according to the lane borrowing trajectory.

[0053] When the optimal driving trajectory is the lane keeping trajectory, it is determined whether a passive lane changing trajectory can be generated. If it can be generated, the final optimal driving trajectory is determined according to the trajectory length and average speed of the passive lane changing trajectory and the lane keeping trajectory. When the lane keeping trajectory is determined as the final optimal driving trajectory, the vehicle is in the lane keeping state and drives according to the lane keeping trajectory. When the passive lane changing trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the passive lane changing state and drives according to the passive lane changing trajectory. If it cannot be generated, the lane keeping trajectory is the final optimal driving trajectory, the vehicle is in the lane keeping state, and drives according to the lane keeping trajectory.

[0054] 3) When the vehicle is in the lane borrowing state and drives according to the lane borrowing trajectory, after passing around the obstacle, a return to the original lane trajectory is generated. The final optimal driving trajectory is determined according to the trajectory length and average speed of the return to the original lane trajectory and the lane borrowing trajectory. When the return to the original lane trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the return to the original lane state and drives according to the return to the original lane trajectory. After returning to the original lane, the vehicle is switched to the lane keeping state. When the lane borrowing trajectory is determined as the final optimal driving trajectory, the vehicle remains in the lane borrowing state and drives according to the lane borrowing trajectory.

[0055] When the vehicle is in the lane borrowing state and drives according to the lane borrowing trajectory, another situation may also occur: the duration of lane borrowing is too long, i.e. the duration of lane borrowing exceeds the time threshold. This situation generally occurs due to the obstacle being too long, causing the vehicle to be unable to return to the original lane. Therefore, in this case, a passive lane changing trajectory is generated. The final optimal driving trajectory is determined according to the trajectory length and average speed of the passive lane changing trajectory and the lane borrowing trajectory. When the lane borrowing trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the lane borrowing state and drives according to the lane borrowing trajectory. When the passive lane changing trajectory is determined as the final optimal driving trajectory, the vehicle is switched to the passive lane changing state and drives according to the passive lane changing trajectory.

[0056] When the vehicle is in the passive lane-changing state and travels according to the passive lane-changing driving track, after lane changing is completed, a track for returning to the original lane is generated, the final optimal driving track is determined according to the track length and average speed of the track for returning to the original lane and the passive lane-changing driving track, when the track for returning to the original lane is determined as the final optimal driving track, the vehicle is switched to the state of returning to the original lane and travels according to the track for returning to the original lane, and after returning to the original lane, the vehicle is switched to the lane-keeping state; when the passive lane-changing driving track is determined as the final optimal driving track, the vehicle remains in the passive lane-changing state and travels according to the passive lane-changing driving track.

[0057] The decision-making process for the case that there is no obstacle in front of the vehicle includes the following steps:

[0058] The active driving track is planned according to the position of the terminal point in combination with the map, and the active driving track includes the lane-keeping track and the active lane-changing driving track.

[0059] The lane-keeping state is the default state of the vehicle, and if only the lane-keeping track is generated without other driving tracks, the vehicle remains in the lane-keeping state.

[0060] If active lane changing is required in the global planning, and there is no track point with a speed of 0 on the active lane-changing driving track, the vehicle is switched to the active lane-changing state and travels according to the active lane-changing driving track, when the active lane changing is completed, the lane after the active lane changing becomes the default lane, and the vehicle is switched to the lane-keeping state.

[0061] In the above embodiment, in the process of selecting the optimal driving track, the track length is the first choice, and in the case of the same track length, the average speed is the basis for selection, as other implementation manners, the average speed can also be the first choice, and in the case of the same average speed, the track length is the basis for selection, and the present application does not limit this.

[0062] The present application divides the automatic driving state into several typical states, selects the optimal driving track according to the path length and average speed of each planned driving track, and then completes the switching of the vehicle state. The present application has high flexibility and strong adaptability, and can increase or reduce the driving scene according to the actual situation; the decision-making is efficient, and can meet the real-time requirements of the automatic driving vehicle.

[0063] The present application divides the driving state into lane keeping, active lane changing, lane changing, passive lane changing, returning to the original lane, etc. The driving state can also be increased or reduced according to the different habits of human driving, application industries or running environments, and is also included in the scope of the present application.

[0064] The decision-making method of the automatic driving vehicle embodiment:

[0065] The implementation process and effects of the decision method of the autonomous vehicle have been described in the above autonomous vehicle embodiment, and will not be repeated here.

Claims

1. A decision-making method of an autonomous vehicle, characterized by, The method comprises the following steps: 1) When there is an obstacle in front of the vehicle during driving, a driving track is planned according to the size and position of the obstacle in combination with a map; the driving track comprises the position of each track point and the vehicle speed; the planned track comprises a lane keeping track and a lane borrowing track; 2) The optimal driving track is determined according to the track length and the average vehicle speed: If the optimal driving track is the lane borrowing track, it is determined whether the lane borrowing track meets the driving condition; if there is no track point with a speed of 0 in the lane borrowing track and the maximum lateral deviation of the vehicle does not exceed the deviation threshold, the lane borrowing track is determined as the final optimal driving track; if there is a track point with a speed of 0 in the lane borrowing track or the maximum lateral deviation of the vehicle exceeds the deviation threshold, it is determined whether a passive lane changing track in the same direction as the lane borrowing track can be generated; if yes, the final optimal driving track is determined according to the track length and the average vehicle speed of the passive lane changing track and the lane borrowing track; if no, the lane borrowing track is determined as the final optimal driving track; the vehicle drives according to the final optimal driving track; If the optimal driving track is the lane keeping track, it is determined whether a passive lane changing track can be generated; if yes, the final optimal driving track is determined according to the track length and the average vehicle speed of the passive lane changing track and the lane keeping track; if no, the lane keeping track is the final optimal driving track; the vehicle drives according to the final optimal driving track.

2. The decision method of an autonomous vehicle according to claim 1, wherein, When the vehicle drives according to the lane borrowing track, after passing the obstacle, a track returning to the original lane is generated; the final optimal driving track is determined according to the track length and the average vehicle speed of the track returning to the original lane and the lane borrowing track; the vehicle drives according to the final optimal driving track.

3. The decision method of an autonomous vehicle according to claim 1 or 2, characterized in that, When the vehicle drives according to the lane borrowing track, if the time length of the lane borrowing exceeds a time threshold, a passive lane changing track is generated; the final optimal driving track is determined according to the track length and the average vehicle speed of the passive lane changing track and the lane borrowing track; the vehicle drives according to the final optimal driving track.

4. The decision method of an autonomous vehicle according to claim 1, wherein, When the vehicle drives according to the passive lane changing track, after completing the lane changing, a track returning to the original lane is generated; the final optimal driving track is determined according to the track length and the average vehicle speed of the track returning to the original lane and the passive lane changing track; the vehicle drives according to the final optimal driving track.

5. The decision method of an autonomous vehicle according to claim 1 or 2 or 4, characterized in that, The step of determining the optimal driving track according to the track length and the average vehicle speed comprises: The track with a longer track length is the optimal driving track; if the track lengths are the same, the track with a larger average vehicle speed is the optimal driving track.

6. The decision method of an autonomous vehicle according to claim 1, wherein, The calculation process of the average vehicle speed is: where vel is the average vehicle speed; s i is the position of the i-th trajectory point; v i is the vehicle speed of the i-th trajectory point; len is the trajectory length, and n is the total number of trajectory points.

7. The decision-making method for an autonomous driving vehicle according to claim 1, wherein: When there is no obstacle in front of the vehicle, an active driving track is planned according to the position of the terminal point in combination with a map; the active driving track comprises a lane keeping track and an active lane changing track.

8. An autonomous vehicle comprising a vehicle body, a positioning module and a perception module disposed on the vehicle body, and a controller, wherein, The controller comprises a processor, a memory, and a computer program stored in the memory and executable on the processor; the processor realizes the decision-making method of the autonomous vehicle according to any one of claims 1-7 when executing the computer program.

Images