A Decision-making Planning Method for Wrong-way Meeting Vehicle Based on Game Planning

Abstract

The present invention discloses a wrong-way passing meeting vehicle decision-making and planning method based on game planning, which includes the following steps: output the current frame state based on the perception information; update the current self-vehicle state based on the current frame state and the self-vehicle state at the previous moment, determine whether the self-vehicle needs to enter the wrong-way passing meeting vehicle state, and update the interactive obstacle information stored inside the self-vehicle; determine whether the self-vehicle needs to re-plan currently, and execute the following steps according to the judgment result: if the judgment result is yes, then plan the optimal rough trajectory of the self-vehicle through interactive game and then select a differential flatness converter to access the optimal rough trajectory of the self-vehicle for backend refinement; if the judgment result is no, then directly select a differential flatness converter to access the optimal rough trajectory of the self-vehicle for backend refinement. The beneficial effects of the present invention are: it can improve the cleaning effect of the self-vehicle and the road passing efficiency under the extreme situation of a narrow road, and can fully ensure the overall optimality of the driving trajectories of the self-vehicle and other vehicles in the current scenario.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent driving, and particularly to a wrong-way meeting vehicle decision-making and planning method based on game planning. Background Art

[0002] In recent years, with the development of driverless technology, autonomous driving has also been widely applied in the field of unmanned cleaning technology, effectively alleviating the pressure of road cleaning. However, when an autonomous driving cleaning vehicle travels on a public road, how to achieve efficient overall road traffic while complying with traffic rules has become a key challenge. Traditional autonomous driving prediction and decision-making planning often forces the unmanned cleaning vehicle to stop suddenly, or blocks the road, resulting in reduced traffic efficiency of other vehicles. This not only limits its cleaning efficiency in specific scenarios but also brings an additional burden to public transportation. Summary of the Invention

[0003] The purpose of the present invention is to provide a wrong-way meeting vehicle decision-making and planning method based on game planning to solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solution: A wrong-way meeting vehicle decision-making and planning method based on game planning, including the following steps:

[0005] Output the current frame state based on the perception information;

[0006] Update the current ego-vehicle state based on the current frame state and the ego-vehicle state at the previous moment, determine whether the ego-vehicle needs to enter the wrong-way meeting vehicle state, and update the interactive obstacle information stored inside the ego-vehicle.

[0007] Determine whether the ego-vehicle currently needs to re-plan, and perform the following steps according to the judgment result:

[0008] If the judgment result is yes, then plan the optimal rough trajectory of the ego-vehicle through interactive game, and then select a differential flatness controller to access the optimal rough trajectory of the ego-vehicle for backend refinement;

[0009] If the judgment result is no, then directly select a differential flatness controller to access the optimal rough trajectory of the ego-vehicle for backend refinement.

[0010] Preferably, the outputting the current frame state based on the perception information includes:

[0011] Update the longitudinal interactive obstacle history queue;

[0012] Select candidate interactive obstacles;

[0013] Judge the current frame state based on the passable space of the interactive obstacles;

[0014] Output the final current frame.

[0015] Preferably, the updating of the longitudinal interaction obstacle history queue includes:

[0016] Updating the status information of the obstacles in the planner;

[0017] Classifying the obstacles into offset meeting obstacles and yielding to rear obstacles based on whether they are in front of or behind the host vehicle.

[0018] Preferably, the step of selecting candidate interaction obstacles includes:

[0019] Selecting one obstacle with the closest distance to the current mileage of the host vehicle from all the offset meeting obstacles and yielding to rear obstacles respectively;

[0020] Selecting the obstacle with the closest distance to the current mileage of the host vehicle;

[0021] Judging whether to enter the offset meeting and yielding state based on its status.

[0022] Preferably, the judging of the current frame state based on the passable space of the interaction obstacle includes:

[0023] Judging whether the current remaining maximum driving space is less than the minimum passable width. If it is greater and the previous frame was not in the offset meeting and yielding state, then it will not enter the offset meeting and yielding state;

[0024] Otherwise, update the offset meeting and yielding target information inside the program and update the offset meeting and yielding state.

[0025] Preferably, the output of the current frame finally includes the offset meeting state, the yielding to rear state or the normal driving state, and the relevant information of the interaction obstacle will be output.

[0026] Preferably, updating the current host vehicle state based on the current frame state and the host vehicle state at the previous moment, judging whether the host vehicle needs to enter the offset meeting state, and updating the interaction obstacle information stored inside the host vehicle, including:

[0027] If the current host vehicle is not in the offset meeting state and the current frame outputs the offset meeting state, then record the initial starting position of the host vehicle, update the offset meeting obstacle information stored inside the host vehicle, and set that the host vehicle needs to re-plan;

[0028] If the current host vehicle is in the offset meeting state and the current frame outputs a non-offset meeting state, then the host vehicle will exit the offset meeting state and clear the information related to the offset meeting in the host vehicle;

[0029] If the current host vehicle is in the offset meeting state and the current frame outputs the offset meeting state, then update the offset meeting obstacle information stored inside the host vehicle and judge whether re-planning is needed.

[0030] Preferably, the judgment of whether the vehicle needs to re-plan currently includes:

[0031] Re-plan for the first entry into the scene state;

[0032] If the vehicle has reached near the last point of the previous plan, and if the current state machine is the same as the previous state, re-planning is required;

[0033] If the vehicle deviates from the previous planned trajectory by a certain distance, angle, and turning angle threshold, count is performed, and re-planning is required after the count meets the threshold;

[0034] If the vehicle's previous planned trajectory collides with other non-interactive obstacles, count is performed, and re-planning is required after the count meets the threshold;

[0035] If another vehicle deviates from the predicted trajectory of the previous plan by a certain distance and angle threshold and count is performed, re-planning is required after the count meets the threshold.

[0036] Preferably, the planning of the optimal rough trajectory of the vehicle through interactive game includes:

[0037] For other non-interactive obstacles, a distance map is constructed according to the prediction and the fixed planning of static obstacles;

[0038] Perform interactive rough planning on the vehicle and the other vehicle interacting with it. The vehicle samples the search tree in the control space;

[0039] The other vehicle further samples the search based on the state space;

[0040] Based on the sampled trajectories of the vehicle and the other vehicle and the corresponding scenarios, further obtain the optimal scenario based on the game cost function.

[0041] Preferably, the differential flatness controller accesses the optimal rough trajectory of the vehicle for backend refinement, including:

[0042] For static obstacles, the vehicle's rough trajectory constructs a passable corridor for each point;

[0043] For dynamic obstacles, the interactive obstacles use the trajectory obtained by rough planning, and other non-interactive obstacles use the predicted trajectory.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] First, the present invention fully considers the actual environment of the current road, and combines dynamic scene recognition and finite state machine to ensure that the vehicle can flexibly switch states during passing-by meeting and yielding to the following vehicle, and can improve the cleaning effect of the vehicle and the road traffic efficiency in extreme narrow road conditions;

[0046] Second, in the scenario of strong interaction for passing vehicles with misaligned positions, the present invention uses a strategy of simultaneous search and sampling of two vehicles to obtain feasible trajectories. Therefore, various future situations of the two vehicles are considered, and the optimal rough trajectories of the host vehicle and the other vehicle are selected based on the game cost function, which can fully ensure the overall optimality of the driving trajectories of the host vehicle and the other vehicle in the current scenario. Description of the Drawings

[0047] Figure 1 It is a flowchart of the method according to an embodiment of the present invention;

[0048] Figure 2 It is a specific flowchart of step S100 according to an embodiment of the present invention;

[0049] Figure 3 It is a specific flowchart of step S200 according to an embodiment of the present invention;

[0050] Figure 4 It is a flowchart of planning the optimal rough trajectory of the host vehicle through interactive game in step S300 according to an embodiment of the present invention;

[0051] Figure 5 It is a flowchart of determining whether the host vehicle needs to re-plan currently in step S300 according to an embodiment of the present invention;

[0052] Figure 6 It is a simulation trajectory diagram of the host vehicle passing by the oncoming other vehicle in the embodiment of the present invention;

[0053] Figure 7 It is a simulation trajectory diagram of the host vehicle reversing to give way to the other vehicle in the embodiment of the present invention;

[0054] Figure 8 It is a simulation trajectory diagram of the host vehicle giving way to the following other vehicle in the embodiment of the present invention. Detailed Embodiment

[0055] 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 only a part of the embodiments of the present invention, rather than all 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.

[0056] Please refer to Figure 1, the present invention provides a technical solution: a wrong-way passing meeting vehicle decision-making and planning method based on game planning. This method enables a driverless cleaning vehicle to flexibly switch the state of the vehicle behind after wrong-way passing meeting according to the actual environment of the current road by combining dynamic scene recognition and a finite state machine, so that the driverless cleaning vehicle can improve the cleaning effect of the vehicle itself and the road traffic efficiency in the case of extremely narrow roads. The implementation solution of the technical solution of this application is that the driverless cleaning vehicle detects the current frame based on sensing information, determines whether the current frame needs to enter the wrong-way passing meeting state and outputs the obstacle information that needs to be interacted, then updates the current vehicle state based on the state output by the current frame and the vehicle state at the previous moment, and then determines whether the vehicle needs to enter the wrong-way passing meeting state and updates the interacted obstacle information stored inside the vehicle; then determines whether the vehicle needs to re-plan currently;

[0057] Then, through an interactive game planning scheme, in this process, multiple trajectories of the vehicle itself and the other vehicle are obtained through two-vehicle search sampling, the optimal rough trajectories of the vehicle itself and the other vehicle are selected through a game cost function, and finally the selected optimal rough trajectory of the vehicle itself is used as the reference trajectory of the differential flatness optimizer for backend fine optimization and planning.

[0058] In an embodiment of the present invention, the implementation of this method includes the following steps:

[0059] S100: Output the current frame state based on sensing information;

[0060] S200: Update the current vehicle state based on the current frame state and the vehicle state at the previous moment, determine whether the vehicle needs to enter the wrong-way passing meeting state, and update the interacted obstacle information stored inside the vehicle;

[0061] S300: Determine whether the vehicle needs to re-plan currently, and perform the following steps according to the judgment result:

[0062] If the judgment result is yes, then select the differential flatness optimizer to access the optimal rough trajectory of the vehicle itself for backend refinement after planning the optimal rough trajectory of the vehicle itself through interactive game;

[0063] If the judgment result is no, then directly select the differential flatness optimizer to access the optimal rough trajectory of the vehicle itself for backend refinement.

[0064] In an embodiment of the present invention, the driverless cleaning vehicle detects the current frame based on sensing information, determines whether the current frame needs to enter the wrong-way passing meeting state and outputs the obstacle information that needs to be interacted. Therefore, step S100 specifically includes:

[0065] S110: Update the longitudinal interacted obstacle history queue.

[0066] In this embodiment, the specific steps of the first step include:

[0067] Update the status information of the obstacles in the planner. That is, eliminate the historical non-displaced passing and yielding rear obstacles, and add the displaced passing and yielding rear obstacles within a certain range from the current ego vehicle (only calculate vehicles and non-motor vehicles).

[0068] Classify the obstacles into displaced passing obstacles and yielding rear obstacles based on whether they are in front of or behind the ego vehicle.

[0069] S120: Select candidate interaction obstacles.

[0070] In this embodiment, specifically, step S120 includes:

[0071] S121: Select one obstacle with the closest distance to the current mileage of the ego vehicle from all the displaced passing obstacles and yielding rear obstacles respectively.

[0072] S122: Select the obstacle with the closest distance to the current mileage of the ego vehicle.

[0073] S123: Judge whether to enter the passing and yielding state based on its status.

[0074] S130: Judge the current frame state based on the passable space of the interaction obstacles.

[0075] In this embodiment, step S130 includes:

[0076] S131: Judge whether the current remaining maximum driving space is less than the minimum passable width. If it is greater and the previous frame was not in the passing and yielding state, then it will not enter the passing and yielding state.

[0077] S132: Otherwise, update the passing and yielding target information inside the program and update the passing and yielding state.

[0078] In this embodiment, specifically, the counter updates its value according to the situation of the current frame. If the previous frame did not enter the passing and yielding state in the current scenario and the current counter is greater than a certain threshold, then it will enter the passing and yielding state; otherwise, if the previous frame was in the wrong passing and yielding state and the current counter is less than or equal to 0, then it will exit the passing and yielding state.

[0079] S140: Output the current frame finally.

[0080] In this embodiment, the finally output states include the displaced passing state, the yielding rear state or the normal driving state, and the relevant information of the interaction obstacles will be output.

[0081] In an embodiment of the present invention, step S200 includes:

[0082] S201: If the current host vehicle is not in the state of passing by a vehicle out of position, and the current frame outputs the state of passing by a vehicle out of position, then record the starting position of the initial host vehicle, update the information of the out-of-position passing vehicle obstacle stored inside the host vehicle, and set the host vehicle to require replanning;

[0083] S202: If the current host vehicle is in the state of passing by a vehicle out of position, and the current frame outputs a non-out-of-position passing vehicle state, then the host vehicle will exit the state of passing by a vehicle out of position, and clear the information related to passing by a vehicle out of position in the host vehicle;

[0084] S203: If the current host vehicle is in the state of passing by a vehicle out of position, and the current frame outputs the state of passing by a vehicle out of position, then update the information of the out-of-position passing vehicle obstacle stored inside the host vehicle, and determine whether replanning is required.

[0085] In this embodiment, step S200 is the step for the host vehicle to yield to the vehicle in front. The same applies to the state of the host vehicle yielding to the vehicle behind, which will not be elaborated here.

[0086] In an embodiment of the present invention, the specific process of determining whether the host vehicle currently needs to replan in step S300 is as follows:

[0087] S301: Perform replanning for the first entry into the scene state;

[0088] S302: If the host vehicle has reached near the last point of the previous plan, and if the current state machine is the same as the previous one, replanning is required;

[0089] S303: If the host vehicle deviates from the previous planned trajectory by a certain distance, angle, and turning angle threshold and then counts, replanning is required after the count meets the threshold;

[0090] S304: If the host vehicle counts after colliding with other non-interactive obstacles in the previous planned trajectory, replanning is required after the count meets the threshold;

[0091] S305: If another vehicle deviates from the predicted trajectory of the other vehicle in our previous plan by a certain distance and angle threshold and then counts, replanning is required after the count meets the threshold.

[0092] In a specific embodiment of the present invention, the specific process of planning the optimal rough trajectory of the host vehicle through interactive game theory in step S300 includes:

[0093] S310: For other non-interactive obstacles, plan according to prediction and static obstacle fixation, and construct a distance map.

[0094] S320: Perform interactive rough planning for the host vehicle and the other vehicle interacting with it. The host vehicle samples the search tree in the control space.

[0095] In this embodiment, the ego vehicle performs search tree sampling in the control space: using the steering angle and step size as the expansion for each step, and for each step of expansion, the distance map is used to eliminate the collision points with non-interactive obstacles. The expansion is set with a maximum distance and a maximum angle change to stop the expansion. Finally, M paths are obtained, and a trajectory with a fixed desired speed is assigned to each path.

[0096] S330: The other vehicle further performs search sampling based on the state space.

[0097] In this embodiment, the last point on the predicted multi-modal path is selected as the intention end point k of the other vehicle, and then K end points are sampled laterally with a Gaussian distribution based on the intention end point k. The sampling range is the lane width of each predicted end point. For the K end points, StateLatticePlanner planning based on model prediction is performed to sample N paths. The distance map is used to check for collisions for each path, and the collided points are eliminated. Finally, different speed allocations are made for these N paths, and ultimately M trajectories are also obtained.

[0098] S340: Based on the sampled trajectories of the ego vehicle and the other vehicle and the corresponding scenarios, the optimal scenario is further obtained based on the game cost function.

[0099] In this embodiment, each of the M trajectories of the ego vehicle and the other vehicle corresponds to a total of scenarios. The simulation results are as Figure 5 shown. The process of further obtaining the optimal scenario based on the game cost function is as follows:

[0100] S3401: The ego vehicle and the other vehicle first need to calculate the current trajectory costs and , which mainly includes two parts: First is the kinematic constraint cost of each, that is, whether it meets its own physical constraints; then it is judged whether the end point of the current trajectory is close enough to the set end point. For the other vehicle, the game cost function is designed as the weighted sum of the distances to multiple intention end points, and the weight coefficient is the probability of each predicted intention end point, that is, the intention uncertainty of the other vehicle's movement is calculated.

[0101] S3402: Further, the interactive game cost between the ego vehicle's current trajectory and the other vehicle's current trajectory needs to be calculated . Specifically, calculate whether these two trajectories will collide and the severity of the collision: For the state of the ego vehicle trajectory at a certain moment t, calculate the number of collisions and the distance with all states within a certain range near the moment t of the other vehicle trajectory. The calculation result is the interaction game cost of the ego vehicle's current trajectory at the moment t. For the entire trajectory, the game costs at all moments are added up; for the state of the other vehicle trajectory at a certain moment t, it is the same. The number of collisions and the distance will be calculated with the trajectory states within a certain range of the ego vehicle trajectory at the moment t to obtain the game cost; finally, the respective interaction game costs of the ego vehicle and the other vehicle trajectories are added up to obtain the interaction game cost in the current scenario; among them, the calculation of the number of collision judgments and the distance is implemented using GPU acceleration to improve the overall calculation efficiency.

[0102] S3403: Calculate the final cost, that is, the weighted sum of the above three costs, that is

[0103] Among them, is the ego vehicle cost weight, is the other vehicle cost weight, is the interaction cost weight, is the ego vehicle cost, is the other vehicle cost, is the interaction cost.

[0104] S3404: Select the ego vehicle's trajectory in the scenario corresponding to the minimum game cost as the optimal ego vehicle rough trajectory, and the other vehicle's trajectory as the optimal other vehicle trajectory for the current prediction and planning. The simulation results are as Figure 6 , Figure 7 and Figure 8 shown, respectively showing the game planning results of the wrong meeting and yielding vehicle decision-making method based on game planning provided by the present invention when passing by, giving way when meeting and reversing, and giving way to the vehicle behind. Further, by continuously adjusting the weights, the selection priority of different scenarios can be adjusted, and the optimal other vehicle trajectory is stored for the replanning judgment at the next moment.

[0105] In an embodiment of the present invention, in order to obtain a smoother and more accurate ego vehicle trajectory, a differential flatness optimizer is used to access the optimal ego vehicle rough trajectory for backend refinement. In the fine optimization part, the kinematic cost of the ego vehicle is used, including the longitudinal speed limit cost, the acceleration limit cost, and the front wheel steering angle limit cost. The environmental cost includes the following two parts:

[0106] For avoiding static obstacles, a passable corridor is constructed for each point based on the ego vehicle rough trajectory. Each passable corridor is expressed by a convex polygon, and the vertex set of the convex polygon is:

[0107] Among them is each end point of the passable corridor, is the trajectory point of the vehicle itself, is the coordinate of the vertex of the passable corridor in the vehicle's own coordinate system, and R is the rotation matrix.

[0108] Then calculate each pose point to be optimized in each passable corridor The constraint cost inside:

[0109] For avoiding m dynamic obstacles, the trajectory obtained by using game planning for the strong interaction obstacles of misjudgment and passing vehicles is used as the predicted trajectory , and the predicted trajectories of other non-strong interaction obstacles are output by using the LTP prediction learning model based on GNN , and this model can output a predicted trajectory with higher accuracy in the case of non-strong interaction. Calculate the minimum distance between each time trajectory point and the convex polygon of the predicted trajectory of each dynamic obstacle at the corresponding moment. When the minimum distance threshold is less than the safety threshold, a penalty constraint cost is added. Here, the distance is expressed by the signed distance: Finally, construct the constraint cost with dynamic obstacles:

[0110] where is the distance between the vehicle itself and each dynamic obstacle at time t.

[0111] The trajectory of the vehicle itself obtained through the above differential flatness fine optimization is smoother. It can not only have good interaction with misjudgment and passing vehicles, but also avoid collisions with other static and dynamic obstacles.

[0112] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A wrong-way meeting vehicle decision-making and planning method based on game planning, characterized in that It includes the following steps: Output the current frame state based on the perception information; Update the current ego vehicle state based on the current frame state and the ego vehicle state at the previous moment, determine whether the ego vehicle needs to enter the staggered passing state, and update the interactive obstacle information stored inside the ego vehicle; Judge whether the ego vehicle currently needs replanning, and execute the following steps according to the judgment result: If the judgment result is yes, after planning the optimal rough trajectory of the ego vehicle through interactive game theory, select a differential flatness controller to access the optimal rough trajectory of the ego vehicle for backend refinement, including: for other non-interactive obstacles, construct a distance map according to the prediction and static obstacle fixed planning, perform interactive rough planning for the ego vehicle and the other vehicle interacting with it. The ego vehicle samples the search tree in the control space, and the other vehicle further samples based on the state space. Based on the sampled trajectories of the ego vehicle and the other vehicle and the corresponding scenarios, further based on the game cost function C total = W own * C own + W other * C other + W interact * C interact Obtain the optimal scenario, where W own is the ego vehicle cost weight, W other is the other vehicle cost weight, W interact is the interaction cost weight, C own is the ego vehicle cost, C other is the other vehicle cost, C interact is the interaction cost, C total is the weighted sum of the ego vehicle cost, the other vehicle cost and the interaction cost; If the judgment result is no, directly select the differential flatness controller to access the optimal rough ego vehicle trajectory for backend refinement, including: For static obstacles, construct the passable corridor for each point of the rough ego vehicle trajectory; For dynamic obstacles, the interactive obstacles use the trajectory obtained by rough planning, and other non-interactive obstacles use the predicted trajectory.

2. The method for wrong-passing meeting vehicle decision-making planning based on game planning according to claim 1, characterized in that: The output of the current frame state based on the perception information includes: Update the longitudinal interactive obstacle history queue; Select candidate interactive obstacles; Judge the current frame state based on the passable space of the interactive obstacles; Output the final current frame.

3. The method for making a wrong-way passing meeting decision plan based on game planning according to claim 2, characterized in that: The update of the longitudinal interactive obstacle history queue includes: Update the status information of the obstacles in the planner; Classify the obstacles into staggered passing obstacles and yielding to rear obstacles based on whether they are in front of or behind the ego vehicle.

4. A wrong-way passing and meeting vehicle decision-making and planning method based on game planning according to claim 3, characterized in that: The selection of candidate interactive obstacles includes: Select one obstacle with the closest distance to the current mileage of the ego vehicle from all the staggered passing obstacles and yielding to rear obstacles respectively; Select the obstacle with the closest distance to the current mileage of the ego vehicle; Judge whether to enter the staggered passing or yielding state based on its status.

5. The method for making a wrong-way passing meeting decision plan based on game planning according to claim 4, characterized in that: The judgment of the current frame state based on the passable space of the interactive obstacles includes: Judge whether the current remaining maximum driving space is less than the minimum passable width. If it is greater and the previous frame was not in the staggered passing or yielding state, then it will not enter the staggered passing or yielding state; Otherwise, update the internal staggered passing or yielding target information of the program and update the staggered passing or yielding state.

6. The method for making a wrong-way passing meeting decision plan based on game planning according to claim 5, wherein: The output of the final current frame includes the staggered passing state, yielding to rear state or normal driving state, and will output the relevant information of the interactive obstacles.

7. A wrong-way passing meeting decision-making planning method based on game planning according to claim 6, characterized in that: The update of the current ego vehicle state based on the current frame state and the ego vehicle state at the previous moment, determine whether the ego vehicle needs to enter the staggered passing state, and update the interactive obstacle information stored inside the ego vehicle includes: If the current ego vehicle is not in the staggered passing state and the current frame outputs the staggered passing state, then record the initial ego vehicle starting position, update the staggered passing obstacle information stored inside the ego vehicle, and set the ego vehicle to need replanning; If the current ego vehicle is in the staggered passing state and the current frame outputs a non-staggered passing state, then the ego vehicle will exit the staggered passing state and clear the information related to staggered passing in the ego vehicle; If the current ego vehicle is in the staggered passing state and the current frame outputs the staggered passing state, then update the staggered passing obstacle information stored inside the ego vehicle and judge whether replanning is needed.

8. The method for making a wrong-way passing meeting decision plan based on game planning according to claim 7, wherein: The judgment of whether the ego vehicle currently needs replanning includes: Perform replanning for the first entry into the scene state; If the ego vehicle has reached near the last point of the previous plan, and the current state machine is the same as the previous one, then replanning is needed; If the ego vehicle deviates from the previous planned trajectory by a certain distance, angle, and turning angle threshold and then counts, and after the count meets the threshold, replanning is needed; If counting starts after the vehicle's last planned trajectory collides with other non-interactive obstacles, replanning is required when the count meets the threshold. If another vehicle deviates from the previously planned predicted trajectory of the other vehicle by a certain distance and angle threshold and counting is performed, replanning is required when the count meets the threshold.

Images