A method for vehicle motion planning in low-precision navigation environments

Abstract

This invention provides a vehicle motion planning method in a low-precision navigation environment. The method includes: Step 1, constructing a semantic lane guidance tree; Step 2, sampling, dividing the sampling points into two branches: the vehicle's own part and the surrounding vehicles' part; Step 3, obtaining the vehicle's forward extended trajectory from the sampling points of the vehicle's own part; Step 4, obtaining the surrounding vehicles' forward predicted trajectory based on the sampling points of the surrounding vehicles' part; Step 5, determining whether there is a passable safe trajectory. If not, proceed to Step 6; if so, proceed to Step 7; Step 6, deleting the sampling point and returning to Step 2; Step 7, performing node matching on the vehicle's forward extended trajectory to obtain new data information, adding it to the semantic lane guidance tree, and then returning to Step 2, until the semantic lane guidance tree reaches the termination condition, ending the loop and extracting the trajectory set; Step 8, selecting the optimal path from the trajectory set for the vehicle to use.

Description

Technical Field

[0001] This application relates to the field of intelligent driving technology, and in particular to a vehicle motion planning method in a low-precision navigation environment. Background Technology

[0002] In today's rapidly developing autonomous driving technology, motion planning, as a core component, directly impacts the reliability and practicality of autonomous driving systems. However, the development of current autonomous driving motion planning technology still faces a series of challenges, particularly its reliance on high-precision maps and positioning. In the field of autonomous driving applications, positioning schemes relying on high-precision maps have long held the mainstream position in the industry, and the accuracy and efficiency of high-precision maps directly affect the performance of autonomous vehicles.

[0003] While high-precision maps and positioning systems provide strong support for autonomous driving, their practical applications still face limitations such as high data acquisition costs, difficulty in maintaining data freshness, and inability to expand to smaller cities. Obtaining high-precision maps of the entire urban landscape requires extensive data acquisition and preliminary data processing, which is costly and difficult to widely implement. Furthermore, autonomous vehicle controller platforms have limited computing power and storage resources. As their operating range expands, the environments they need to deal with become increasingly complex and variable. The memory consumption of high-precision map storage and the computing power consumed during positioning retrieval are also significant factors, undoubtedly increasing the maintenance difficulty of autonomous driving systems.

[0004] While real-time semantic mapping technology can reduce the reliance on high-precision maps for autonomous driving, enabling lightweight and real-time mapping, its absolute coordinate positioning accuracy is lower than that of high-precision maps, and it also contains less road element attribute information. To achieve widespread application and commercialization of autonomous driving in urban areas, it is essential to consider autonomous driving strategies in navigation environments with lower-precision real-time semantic maps. Therefore, developing autonomous driving motion planning technology based on real-time semantic maps is particularly crucial. This requires fully leveraging the advantages of lightweight and real-time capabilities of real-time semantic maps while overcoming their limitations in accuracy and richness compared to high-precision maps at the autonomous driving planning level, in order to ensure the safety and stability of the autonomous driving system. Summary of the Invention

[0005] In view of this, the present invention provides a vehicle motion planning method in a low-precision navigation environment to solve the technical problem that the existing technology cannot adapt to the low-precision navigation environment of autonomous driving.

[0006] This invention provides a vehicle motion planning method in a low-precision navigation environment, applicable to autonomous vehicles. The method includes: Step 1, constructing a semantic lane guidance tree based on data information of the vehicle itself and other vehicles; Step 2, sampling based on the data information to obtain sampling points, dividing the sampling points into two branches: the vehicle's own part and the surrounding vehicles' part; Step 3, generating semantic pre-aiming points from the sampling point positions of the vehicle's own part, selecting the optimal parent node, and calculating the vehicle's forward extension trajectory using a fast curve expansion algorithm; Step 4, filtering semantically associated vehicles based on the sampling of the surrounding vehicles' part, and calculating the surrounding vehicles' forward predicted trajectory using a fast curve expansion algorithm; Step 5, determining whether there is a passable safe trajectory based on the vehicle's forward extension trajectory and the surrounding vehicles' forward predicted trajectory. If not, proceed to Step 6; if yes, proceed to Step 7; Step 6, deleting the sampling point, returning to Step 2, and resampling; Step 7, performing node matching on the vehicle's forward extension trajectory to obtain new data information, adding it to the semantic lane guidance tree, and then returning to Step 2, until the semantic lane guidance tree reaches the termination condition, ending the loop, and extracting the trajectory set; Step 8, selecting the optimal trajectory based on the optimal trajectory selection function: The optimal path is selected from the trajectory set for the vehicle to use, where t1 is any passable trajectory extracted from the semantic lane guidance tree, cost_t(t1) represents the cost of trajectory t1 under this evaluation system, and efficiency(t1) represents the calculated efficiency of trajectory t1 under the current road conditions, δ e It is the weight of the traffic efficiency cost; the safety cost safety(t1) represents the degree of safety of trajectory t1 relative to the obstacle, δ s The weights represent the safety costs of the trajectory; navigation(t1) represents the navigation cost of trajectory t1, δ n It is the weight of curvature cost.

[0007] Furthermore, the data information includes vehicle attitude information, topology information, semantic information, and trajectory information.

[0008] Furthermore, the forward extension trajectory of the vehicle is (V parent V sample ,E), where V parent V represents the optimal mother node of the selected sampling points. sample denoted by , and E represents the extended trajectory.

[0009] Furthermore, the predicted forward trajectory of the vehicle is (Agent) id ,E), where Agent id The symbol represents the number of the semantically associated vehicle, and E represents the forward predicted trajectory of the semantically associated vehicle.

[0010] Furthermore, the method for performing node matching on the vehicle's forward extended trajectory to obtain new data information includes: Step 71, matching the vehicle's forward extended trajectory (V... parent V sample Perform node matching on E) to obtain the matched node (V) parent V child Step 72, based on the matched node (V) parent V child E), we get edge (V) child Step 73, extend the new node V generated by the forward extension of the vehicle. new With edge (V) child E) combined into new data information (V) new E new ).

[0011] Furthermore, the termination condition is: termination_condition = (Num Trajectory >Num threshold )||(Time calculation Time threshold ), where Num Trajectory Num represents the number of trajectories. threshold Indicates the end of the track count, Time calculation Indicates the time of a single calculation. threshold This indicates the maximum time limit for a single calculation.

[0012] Furthermore, the optimal trajectory is the trajectories with the lowest cost in the trajectory set.

[0013] Furthermore, the traffic efficiency cost efficiency(t1) = δ e1 *abs(velocity ego -velocity desire )+δ e2 *abs(velocity ego -velocity leadingVehicle ), where velocity ego For average vehicle speed, velocity desire Velocity is the target cruising speed for the current road. leadingVehicle δ represents the speed of the car following ahead. e1 and δ e2 There are two weights.

[0014] Furthermore, the safety cost (t1) is evaluated using a responsibility-sensitive safety model.

[0015] Furthermore, the navigation cost navigation(t1) = ∑δ reward *Trajectory recommend +∑δ punish *Trajectory non-recommend , among which, Trajectory recommend This indicates the recommended portion of the vehicle's current trajectory. non-recommend Indicates the portion of the vehicle's current trajectory that is not recommended, δ reward δ represents the reward coefficient. punish This represents the penalty coefficient.

[0016] This invention provides a vehicle motion planning method for low-precision navigation environments. The proposed solution is a Semantic Lane-Guided Rapidly-exploring RandomTree (SLG-RRT) algorithm, which includes a sampling method based on vehicle behavior semantic guidance, a fast pruning method based on SLG feature trees, and a trajectory optimization method that incorporates weighted fusion of multiple constraints related to traffic efficiency, safety, and semantic guidance. It primarily addresses the low sampling efficiency of motion planning algorithms in existing semantic map environments, the exponentially increasing computational difficulty of interaction with surrounding vehicles in urban road environments, and the problem of optimal trajectory selection after autonomous driving motion planning algorithms have obtained multiple solutions. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a vehicle motion planning method in a low-precision navigation environment provided by the present invention;

[0018] Figure 2 This is a diagram of a fast expanding random tree algorithm architecture for semantic lane guidance provided by the present invention;

[0019] Figure 3 This is a schematic diagram of the method for obtaining new data information by performing node matching on the forward extension trajectory of a vehicle, as provided by the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1

[0022] This invention provides a vehicle motion planning method for low-precision navigation environments, applicable to autonomous vehicles, such as... Figure 1 As shown, the method includes the following steps.

[0023] Step 1: Construct a semantic lane guidance tree based on data information of your own vehicle and other vehicles;

[0024] In the semantic lane environment, we define autonomous vehicles as "self-vehicles" and other traffic participants as "circle vehicles," collectively referred to as intelligent agents. The Semantic Lane Guidance (SLG) tree is a tree structure containing a node set V and an edge set E. The node set V contains the vehicle's attitude information (coordinates x, y, heading, velocity v, acceleration a) and topological information (parent node, child node set). The edge set contains semantic information connecting the parent node to the child nodes (lateral behavioral semantics, longitudinal behavioral semantics) and trajectory information (a set of points containing vehicle attitude information).

[0025] Step 2: Sample the data to obtain sampling points, and divide the sampling points into two branches: the self-car part and the surrounding car part;

[0026] Step 3: Generate semantic preview points from the sampling points of the vehicle part, select the optimal mother node and calculate the forward extension trajectory of the vehicle using the fast curve extension algorithm;

[0027] Step 4: Based on the sampling of the surrounding vehicles, semantically related vehicles are selected, and the forward predicted trajectory of the surrounding vehicles is calculated using the fast curve expansion algorithm.

[0028] Step 5: Based on the forward extension trajectory of the vehicle and the forward prediction trajectory of the surrounding vehicles, determine whether there is a passable safe trajectory. If not, proceed to step 6; if yes, proceed to step 7.

[0029] Step 6: Delete the sampling point, go back to step 2, and resample;

[0030] Step 7: Perform node matching on the forward extended trajectory of the vehicle to obtain new data information, add it to the semantic lane guidance tree, then go back to step 2 until the semantic lane guidance tree reaches the termination condition, end the loop, and extract the trajectory set.

[0031] Step 8: Select a function based on the optimal trajectory: The optimal path is selected from the trajectory set for the vehicle to use, where t1 is any passable trajectory extracted from the semantic lane guidance tree, cost_t(t1) represents the cost of trajectory t1 under this evaluation system, and efficiency(t1) represents the calculated efficiency of trajectory t1 under the current road conditions, δ eIt is the weight of the traffic efficiency cost; the safety cost safety(t1) represents the degree of safety of trajectory t1 relative to the obstacle, δ s The weights represent the safety costs of the trajectory; navigation(t1) represents the navigation cost of trajectory t1, δ n It is the weight of curvature cost.

[0032] This invention provides a vehicle motion planning method for low-precision navigation environments. The proposed solution is a Semantic Lane-Guided Rapidly-exploring RandomTree (SLG-RRT) algorithm, which includes a sampling method based on vehicle behavior semantic guidance, a fast pruning method based on SLG feature trees, and a trajectory optimization method that incorporates weighted fusion of multiple constraints related to traffic efficiency, safety, and semantic guidance. It primarily addresses the low sampling efficiency of motion planning algorithms in existing semantic map environments, the exponentially increasing computational difficulty of interaction with surrounding vehicles in urban road environments, and the problem of optimal trajectory selection after autonomous driving motion planning algorithms have obtained multiple solutions.

[0033] Example 2

[0034] This invention provides a vehicle motion planning method for low-precision navigation environments, applicable to autonomous vehicles, such as... Figure 1 As shown, the method includes the following steps.

[0035] Step 1: Construct a semantic lane guidance tree based on data information of your own vehicle and other vehicles;

[0036] Step 2: Sample the data to obtain sampling points, and divide the sampling points into two branches: the self-car part and the surrounding car part;

[0037] As mentioned earlier, the data information includes vehicle attitude information, topology information, semantic information, and trajectory information. Sampling based on vehicle behavior semantic guidance is first required in the semantic lane environment. Vehicle behavior semantic actions are high-level actions that can be directly interpreted using human senses. We first divide vehicle behavior semantic actions into lateral semantic actions and longitudinal semantic actions. Lateral semantic actions include lane keeping, left lane changing, and right lane changing, represented by (K, L, R) respectively; longitudinal semantic actions include speed maintenance, acceleration, and deceleration, represented by (M, A, D). For example, changing lanes to the left and decelerating on a public road directly constitute a set of semantic actions, which can be expressed by (L, D). Semantic actions actually involve multiple small motion planning steps. To implement a set of left lane changing and deceleration (L, D) actions, multiple lateral turning sequences and multiple longitudinal acceleration sequences are required. However, considering such local actions (lateral turning and longitudinal acceleration) in the sampling algorithm would result in an excessively deep SLG tree, leading to a computational dimensionality explosion and affecting the real-time performance of autonomous vehicles. After sampling, a sampling point will be obtained, and the obtained sampling point will be divided into two branches: the self-car part and the surrounding car part.

[0038] Step 3: Generate semantic preview points from the sampling points of the vehicle part, select the optimal mother node and calculate the forward extension trajectory of the vehicle using the fast curve extension algorithm;

[0039] As mentioned above, the sampling points are divided into the vehicle-centric part and the surrounding vehicle part. In the vehicle-centric part, semantic preview points are generated based on the sampling point positions. The optimal parent node is selected, and the forward extension trajectory (V) is calculated using the fast Clothoid extension algorithm. parent V sample (,E). Where V parent V represents the optimal mother node of the selected sampling points. sample denoted by , and E represents the extended trajectory.

[0040] Step 4: Based on the sampling of the surrounding vehicles, semantically related vehicles are selected, and the forward predicted trajectory of the surrounding vehicles is calculated using the fast curve expansion algorithm.

[0041] Similarly, in the surrounding vehicle section, semantically associated vehicles are filtered out through semantic lane determination at sampling points, and the forward predicted trajectory of the semantically associated vehicles is calculated using the fast Clothoid extension algorithm. (Agent) id ,E), where Agent id The symbol represents the number of the semantically associated vehicle, and E represents the forward predicted trajectory of the semantically associated vehicle.

[0042] Step 5: Based on the forward extension trajectory of the vehicle and the forward prediction trajectory of the surrounding vehicles, determine whether there is a passable safe trajectory. If not, proceed to step 6; if yes, proceed to step 7.

[0043] Step 6: Delete the sampling point, go back to step 2, and resample;

[0044] Step 7: Perform node matching on the forward extended trajectory of the vehicle to obtain new data information, add it to the semantic lane guidance tree, then go back to step 2 until the semantic lane guidance tree reaches the termination condition, end the loop, and extract the trajectory set.

[0045] Step 8: Select a function based on the optimal trajectory: The optimal path is selected from the trajectory set for the vehicle to use, where t1 is any passable trajectory extracted from the semantic lane guidance tree, cost_t(t1) represents the cost of trajectory t1 under this evaluation system, and efficiency(t1) represents the calculated efficiency of trajectory t1 under the current road conditions, δ e It is the weight of the traffic efficiency cost; the safety cost safety(t1) represents the degree of safety of trajectory t1 relative to the obstacle, δ s The weights represent the safety costs of the trajectory; navigation(t1) represents the navigation cost of trajectory t1, δ n It is the weight of curvature cost.

[0046] This invention provides a vehicle motion planning method for low-precision navigation environments. The proposed solution is a Semantic Lane-Guided Rapidly-exploring RandomTree (SLG-RRT) algorithm, which includes a sampling method based on vehicle behavior semantic guidance, a fast pruning method based on SLG feature trees, and a trajectory optimization method that incorporates weighted fusion of multiple constraints related to traffic efficiency, safety, and semantic guidance. It primarily addresses the low sampling efficiency of motion planning algorithms in existing semantic map environments, the exponentially increasing computational difficulty of interaction with surrounding vehicles in urban road environments, and the problem of optimal trajectory selection after autonomous driving motion planning algorithms have obtained multiple solutions.

[0047] Example 3

[0048] This invention provides a vehicle motion planning method for low-precision navigation environments, applicable to autonomous vehicles, such as... Figure 2 As shown, the method includes the following steps.

[0049] Step 1: Construct a semantic lane guidance tree based on data information of your own vehicle and other vehicles;

[0050] Step 2: Sample the data to obtain sampling points, and divide the sampling points into two branches: the self-car part and the surrounding car part;

[0051] Step 3: Generate semantic preview points from the sampling points of the vehicle part, select the optimal mother node and calculate the forward extension trajectory of the vehicle using the fast curve extension algorithm;

[0052] Step 4: Based on the sampling of the surrounding vehicles, semantically related vehicles are selected, and the forward predicted trajectory of the surrounding vehicles is calculated using the fast curve expansion algorithm.

[0053] Step 5: Based on the forward extension trajectory of the vehicle and the forward prediction trajectory of the surrounding vehicles, determine whether there is a passable safe trajectory. If not, proceed to step 6; if yes, proceed to step 7.

[0054] We conduct a decision-making scenario game between the forward extended trajectory of the autonomous vehicle and the forward predicted trajectory of semantically associated vehicles. If the decision-making scenario game fails, the sampling point is discarded and we return to the behavior semantic guidance sampling; if it succeeds, we then perform a decision-making scenario game between (V) and ... parent V sample Perform node matching on E) to obtain (V) parent V child ,E), and connect the nodes and edges (V) child E) is transformed into a new combination of nodes and edges (V) new E new Add to the SLG tree.

[0055] Step 6: Delete the sampling point, go back to step 2, and resample;

[0056] Step 7: Perform node matching on the forward extended trajectory of the vehicle to obtain new data information, add it to the semantic lane guidance tree, then go back to step 2 until the semantic lane guidance tree reaches the termination condition, end the loop, and extract the trajectory set.

[0057] like Figure 3 As shown, the method for performing node matching on the forward extension trajectory of the vehicle to obtain new data information is as follows.

[0058] Step 71, extend the vehicle's forward trajectory (V) parent V sample Perform node matching on E) to obtain the matched node (V) parent V child ,E);

[0059] Step 72, based on the matched node (V) parent V child E), we get edge (V) child ,E);

[0060] Step 73, extend the new node V generated by the forward extension of the vehicle. new With edge (V) child E) combined into new data information (V) new E new ).

[0061] The termination condition is: termination_condition = (Num Trajectory >

[0062] Num threshold )||(Time calculation Time threshold ), where Num Trajectory Num represents the number of trajectories. threshold Indicates the end of the track count, Time calculation Indicates the time of a single calculation. threshold This indicates the maximum time limit for a single calculation.

[0063] After the above steps, update the SLG tree and continue behavioral semantic guided sampling. Repeat steps 2-7 to update the SLG tree in this loop until the SLG tree reaches the termination condition, at which point the loop ends.

[0064] Step 8: Select a function based on the optimal trajectory: The optimal path is selected from the trajectory set for the vehicle to use, where t1 is any passable trajectory extracted from the semantic lane guidance tree, cost_t(t1) represents the cost of trajectory t1 under this evaluation system, and efficiency(t1) represents the calculated efficiency of trajectory t1 under the current road conditions, δ e It is the weight of the traffic efficiency cost; the safety cost safety(t1) represents the degree of safety of trajectory t1 relative to the obstacle, δ s The weights represent the safety costs of the trajectory; navigation(t1) represents the navigation cost of trajectory t1, δ n It is the weight of curvature cost.

[0065] This invention provides a vehicle motion planning method for low-precision navigation environments. The proposed solution is a Semantic Lane-Guided Rapidly-exploring RandomTree (SLG-RRT) algorithm, which includes a sampling method based on vehicle behavior semantic guidance, a fast pruning method based on SLG feature trees, and a trajectory optimization method that incorporates weighted fusion of multiple constraints related to traffic efficiency, safety, and semantic guidance. It primarily addresses the low sampling efficiency of motion planning algorithms in existing semantic map environments, the exponentially increasing computational difficulty of interaction with surrounding vehicles in urban road environments, and the problem of optimal trajectory selection after autonomous driving motion planning algorithms have obtained multiple solutions.

[0066] Example 4

[0067] This invention provides a vehicle motion planning method for low-precision navigation environments, applicable to autonomous vehicles, such as... Figure 1 As shown, the method includes the following steps.

[0068] Step 1: Construct a semantic lane guidance tree based on data information of your own vehicle and other vehicles;

[0069] Step 2: Sample the data to obtain sampling points, and divide the sampling points into two branches: the self-car part and the surrounding car part;

[0070] Step 3: Generate semantic preview points from the sampling points of the vehicle part, select the optimal mother node and calculate the forward extension trajectory of the vehicle using the fast curve extension algorithm;

[0071] Step 4: Based on the sampling of the surrounding vehicles, semantically related vehicles are selected, and the forward predicted trajectory of the surrounding vehicles is calculated using the fast curve expansion algorithm.

[0072] Step 5: Based on the forward extension trajectory of the vehicle and the forward prediction trajectory of the surrounding vehicles, determine whether there is a passable safe trajectory. If not, proceed to step 6; if yes, proceed to step 7.

[0073] Step 6: Delete the sampling point, go back to step 2, and resample;

[0074] Step 7: Perform node matching on the forward extended trajectory of the vehicle to obtain new data information, add it to the semantic lane guidance tree, then go back to step 2 until the semantic lane guidance tree reaches the termination condition, end the loop, and extract the trajectory set.

[0075] Step 8: Select a function based on the optimal trajectory: The optimal path is selected from the trajectory set for the vehicle to use, where t1 is any passable trajectory extracted from the semantic lane guidance tree, cost_t(t1) represents the cost of trajectory t1 under this evaluation system, and efficiency(t1) represents the calculated efficiency of trajectory t1 under the current road conditions, δ e It is the weight of the traffic efficiency cost; the safety cost safety(t1) represents the degree of safety of trajectory t1 relative to the obstacle, δ s The weights representing the cost of trajectory safety;

[0076] navigation(t1) represents the navigation cost of trajectory t1, δ n It is the weight of curvature cost.

[0077] The optimal trajectory is the walkable trajectory with the lowest cost in the trajectory set. Within the trajectory set, the optimal trajectory T is calculated using a weighted fusion method that integrates multiple constraints such as traffic efficiency, safety, and semantic guidance. optimal .

[0078] The efficiency cost (t1) is δ e1 *abs(velocity ego -velocity desire )+δ e2 *abs(velocity ego -velocity leadingVehicle ), where velocity ego For average vehicle speed, velocity desire Velocity is the target cruising speed for the current road. leadingVehicle δ represents the speed of the car following ahead. e1 and δ e2 There are two weights. The calculation of efficiency cost (t1) mainly includes two parts: the average vehicle speed (velocity) on trajectory t1. ego Velocity of the current road target cruise speed desire The difference, and the average vehicle speed velocity of trajectory t1 ego Velocity of the car following ahead leadingVehicle The speed difference (0 if there is no vehicle in front).

[0079] The safety cost (t1) mainly calculates the safety level of the current trajectory, and the safety cost (t1) is evaluated using a responsibility-sensitive safety model.

[0080] The calculation of navigation cost (t1) mainly consists of two parts: the current trajectory within the recommended lane and the trajectory itself. recommend The reward, and the current trajectory is not in the recommended lane part of the trajectory. non-recommend The penalty is the navigation cost navigation(t1) = ∑δ. reward *Trajectory recommend +∑δ punish *Trajectory non-recommend , among which, Trajectory recommend This indicates the recommended portion of the vehicle's current trajectory. non-recommend Indicates the portion of the vehicle's current trajectory that is not recommended, δ reward δ represents the reward coefficient. punish This represents the penalty coefficient. The recommended and non-recommended lanes described are data obtained from semantic lane navigation. For example, if the vehicle navigation indicates a left turn at the next intersection, then the lane for the left turn is the recommended lane, and all other lanes are non-recommended lanes.

[0081] All three costs mentioned above need to be normalized before calculation.

[0082] This invention provides a vehicle motion planning method for low-precision navigation environments. The proposed solution is a Semantic Lane-Guided Rapidly-exploring RandomTree (SLG-RRT) algorithm, which includes a sampling method based on vehicle behavior semantic guidance, a fast pruning method based on SLG feature trees, and a trajectory optimization method that incorporates weighted fusion of multiple constraints related to traffic efficiency, safety, and semantic guidance. It primarily addresses the low sampling efficiency of motion planning algorithms in existing semantic map environments, the exponentially increasing computational difficulty of interaction with surrounding vehicles in urban road environments, and the problem of optimal trajectory selection after autonomous driving motion planning algorithms have obtained multiple solutions.

[0083] In summary, this invention provides a vehicle motion planning method for low-precision navigation environments. This solution reduces the sampling space of the RRT algorithm by using a driver semantic guidance branch in the action space, significantly improving sampling efficiency. Instead of directly considering all possible combinations of surrounding vehicle intentions to engage in a game with the vehicle, this solution filters potentially risky scenarios among surrounding vehicles based on the vehicle's semantic behavior conditions, reducing computational overhead during the game between the vehicle and surrounding vehicles, thus lowering computational complexity. Furthermore, this invention proposes an optimal trajectory selection method that integrates multiple constraints such as traffic efficiency, safety, and semantic guidance with weighted fusion, improving the comfort and safety of the optimal trajectory.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A vehicle motion planning method in a low-precision navigation environment, applied to autonomous vehicles, characterized in that, The method includes: Step 1: Construct a semantic lane guidance tree based on data information of your own vehicle and other vehicles; Step 2: Sample the data to obtain sampling points, and divide the sampling points into two branches: the self-car part and the surrounding car part; Step 3: Generate semantic preview points based on the sampling point positions of the vehicle part, select the optimal mother node and calculate the forward extension trajectory of the vehicle using the fast curve extension algorithm; Step 4: Based on the sampling points of the surrounding vehicle section, select semantically related vehicles and calculate the forward predicted trajectory of the surrounding vehicles using the fast curve expansion algorithm. Step 5: Based on the forward extension trajectory of the vehicle and the forward prediction trajectory of the surrounding vehicles, determine whether there is a passable safe trajectory. If not, proceed to step 6; if yes, proceed to step 7. Step 6: Delete the sampling point, go back to step 2, and resample; Step 7: Perform node matching on the forward extended trajectory of the vehicle to obtain new data information, add it to the semantic lane guidance tree, then go back to step 2 until the semantic lane guidance tree reaches the termination condition, end the loop, and extract the trajectory set. Step 8: Select a function based on the optimal trajectory: ， The optimal path is selected from the set of trajectories for the vehicle to use. It is any passable trajectory extracted from the semantic lane guidance tree. Representing the trajectory The cost of transportation efficiency; Representing the trajectory Calculated traffic efficiency under current road conditions It is the weight of the cost of traffic efficiency; the cost of security. Representing the trajectory Compared to the safety level of the obstacle, The weight representing the security cost; Representing the trajectory The cost of navigation It is the weight of the navigation cost.

2. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The data information includes vehicle attitude information, topology information, semantic information, and trajectory information.

3. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The vehicle's forward extension trajectory is as follows: ,in This represents the optimal mother node for the selected sampling points. Indicates the sampling point. This indicates the extended trajectory.

4. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The predicted forward trajectory of the circumferential vehicle is as follows: ,in The number representing the semantically associated vehicle. This represents the forward predicted trajectory of semantically associated vehicles.

5. The vehicle motion planning method in a low-precision navigation environment according to claim 3, characterized in that, The method for performing node matching on the vehicle's forward extended trajectory to obtain new data information includes: Step 71, extend the forward trajectory of the vehicle. Perform node matching to obtain the matched nodes. ; Step 72, based on the matched nodes , obtain the edge ( ); Step 73, extend the new node generated by the vehicle forward. With the edge ( ) combined into new data information ( ).

6. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The termination condition is: ， in, Indicates the number of trajectories. Indicates the end of the trajectory count. Indicates the time for a single calculation. This indicates the maximum time limit for a single calculation.

7. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The optimal path is the walkable path with the lowest cost in the trajectory set.

8. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The traffic efficiency cost ， in, Average vehicle speed The target cruising speed for the current road. This represents the speed of the car following ahead. and There are two weights.

9. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The security cost The evaluation was conducted using a responsibility-sensitive safety model.

10. The vehicle motion planning method in a low-precision navigation environment according to claim 1, characterized in that, The navigation cost = ， in, This indicates the recommended portion of the vehicle's current trajectory. This indicates the portion of the vehicle's current trajectory that is not recommended. Indicates the reward coefficient. This represents the penalty coefficient.

Images