Knowledge and data collaboratively driven trajectory planning conflict avoidance method, medium and product

Abstract

The invention discloses a knowledge and data collaboratively-driven trajectory planning conflict avoidance method, medium and product, and the method comprises the steps: generating a plurality of groups of combined trajectories through a Cartesian product based on a Pareto optimal planning trajectory of a plurality of moving targets; storing root nodes initialized for each group of joint tracks by using an OPEN list, and creating a binary tree; repeating the following steps: selecting nodes by using a neural network model for node sorting; adding the selected nodes into a feasible solution set when track conflicts do not exist in the joint track; when track conflicts exist in the joint tracks, conflict avoidance is carried out based on the selected nodes and at least two moving targets generating conflicts, and at least two child nodes containing new joint tracks are generated and added into an OPEN list; and selecting a node with the minimum total cost from the feasible solution set, and distributing a conflict-free planning track for each moving target. According to the method, the binary tree model is effectively constructed, and the nodes are reasonably selected for trajectory planning conflict avoidance.

Description

Technical Field

[0001] The present invention relates to a trajectory planning conflict avoidance method, medium and product driven by the collaboration of knowledge and data, belonging to the cross-technical field of transportation and artificial intelligence. Background Art

[0002] The conflict avoidance algorithm based on tree search has shown significant advantages in mobile target path planning. Its core lies in the ability to effectively identify and eliminate possible collisions between different mobile target planning paths using a pre-designed computational framework. This type of algorithm usually explores possible conflict elimination solutions by constructing a tree data structure, and continuously improves the conflict elimination solution through iterative expansion search until it is feasible. In each step of the search, the algorithm generates new nodes, representing the path planning solution after eliminating some conflicts, and gradually builds a conflict-free path planning result by expanding these nodes. When a potential collision conflict is detected, the algorithm will adopt corresponding avoidance strategies by adding new constraints, such as adjusting the path, slowing down, or waiting, to ensure that all mobile targets can reach their destination safely and without conflict. Tree search-based methods have been widely used in many fields such as autonomous driving, robot navigation, and air traffic management due to their flexibility and scalability.

[0003] Although the conflict avoidance algorithm based on tree search has achieved remarkable results in the field of path planning, its limitations gradually emerge when facing more complex multi-objective trajectory planning conflict avoidance. First, how to effectively construct a binary tree model suitable for multi-objective trajectory planning problems is a challenge. Traditional tree search algorithms often focus on the optimization of a single goal such as the shortest path or minimum time. In practical applications, trajectory planning not only involves the problem of coordinated optimization of paths and trajectories, but also often needs to consider multiple goals at the same time, such as the safety, energy consumption, and comfort of the trajectory. These goals may conflict with each other, making it difficult for the existing binary tree model to fully reflect all optimization requirements. Secondly, the reasonable selection of nodes for expansion is also a major problem. In the scenario where multiple mobile targets compete for road resources, the evaluation criteria for node selection during the expansion search become complex and diverse. Not only the direct benefits of the current node need to be considered, but also its impact on the subsequent conflict avoidance process and the overall objective function needs to be predicted, which greatly increases the complexity and computational cost of the search. Therefore, there are still deficiencies in how to effectively construct a binary tree model for multi-objective trajectory planning and how to efficiently select expansion nodes, which limits the further application and development of the conflict avoidance algorithm based on tree search in the field of multi-objective trajectory planning.

[0004] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgement or any form of suggestion that the information constitutes the prior art already known to ordinary technicians in this field. Summary of the invention

[0005] The technical problem to be solved by the present invention is: how to effectively construct a binary tree model and reasonably select nodes to avoid trajectory planning conflicts.

[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions.

[0007] In a first aspect, the present invention provides a trajectory planning conflict avoidance method driven by knowledge and data collaboration, comprising:

[0008] Obtaining operation plans for multiple moving targets;

[0009] According to the operation plan, based on the Pareto optimal planning trajectories of multiple moving targets, multiple sets of joint trajectories are generated through Cartesian products;

[0010] Initialize a root node for each set of joint trajectories and create a binary tree based on each root node;

[0011] Use OPEN list to store the root node of each binary tree;

[0012] Repeat the following steps until the OPEN list is empty or the calculation time threshold is reached, and then output a feasible solution set:

[0013] Using the pre-trained node sorting neural network model, select nodes from the OPEN list to expand:

[0014] When the joint trajectory included in the selected node does not have trajectory conflict: adding the selected node to the feasible solution set and then removing it from the OPEN list;

[0015] When the joint trajectory included in the selected node has a trajectory conflict: remove the selected node from the OPEN list, and based on the selected node and at least two mobile targets that generate the conflict, perform conflict avoidance, generate at least two child nodes including new joint trajectories, and add them to the OPEN list;

[0016] Select the node with the minimum total cost from the feasible solution set, and assign conflict-free planning trajectories to each mobile target according to the joint trajectory contained in the node with the minimum total cost;

[0017] The joint trajectory contained in the selected node is a trajectory set, including a planned trajectory for each mobile target.

[0018] Further, according to the operation plan, based on the Pareto optimal planning trajectories of multiple moving targets, multiple sets of joint trajectories are generated through Cartesian products, including:

[0019] The operation plan includes a starting position and a target position of the mobile object;

[0020] Generating a Pareto optimal planning trajectory for each mobile target by a multi-objective optimization method according to the starting position and the target position;

[0021] The Cartesian product is calculated based on the Pareto optimal planning trajectory of each mobile target to generate multiple sets of joint trajectories.

[0022] Further, each node of the binary tree includes a triplet, and the triplet includes a joint trajectory of the moving target, a total cost, and a constraint set;

[0023] The joint trajectory is a set of trajectories, including a planned trajectory for each mobile target;

[0024] The total cost is a numerical representation of the sum of the costs of each planned trajectory in each set of joint trajectories, and is used to evaluate the overall performance of each set of joint trajectories;

[0025] The constraint set is a set of constraint conditions for each moving target when planning a trajectory, and is initially set to an empty set.

[0026] Furthermore, the node sorting neural network model includes:

[0027] The first deep neural network is constructed based on the gated recurrent unit and is used to perform feature encoding on each planned trajectory in the joint trajectory contained in the nodes in the OPEN list to obtain multiple trajectory feature vectors;

[0028] The second deep neural network is constructed based on the fully connected layer and the ReLU activation function, and is used to concatenate the multiple trajectory feature vectors, obtain the scores of the nodes in the OPEN list through softmax operation, and sort them according to the scores, wherein the nodes with larger scores have higher sorting priority.

[0029] Furthermore, the method for acquiring training data of the node-sorted neural network model comprises:

[0030] Obtain historical operation plans for multiple mobile targets;

[0031] According to the historical operation plan, based on the Pareto optimal planning trajectories of multiple moving targets, multiple sets of joint trajectories are generated through Cartesian products;

[0032] Initialize a root node for each set of joint trajectories and create a binary tree based on each root node;

[0033] Use OPEN list to store the root node of each binary tree;

[0034] Repeat the following steps until the OPEN list is empty:

[0035] Selecting a node from the OPEN list using any heuristic strategy;

[0036] When the joint trajectory included in the selected node does not have trajectory conflict: the selected node is taken as a leaf node of the binary tree and added to the feasible solution set, and then the selected node is removed from the OPEN list;

[0037] When the joint trajectory included in the selected node has a trajectory conflict: remove the selected node from the OPEN list, and based on the selected node and at least two mobile targets that generate the conflict, perform conflict avoidance, generate at least two child nodes including new joint trajectories, and add them to the OPEN list;

[0038] For all leaf nodes included in the feasible solution set, trace back from the leaf nodes to the root node in turn, record the information of each intermediate node, and generate training data with sample labels.

[0039] Furthermore, the sample label is the shortest distance from each intermediate node to a leaf node in its sub-binary tree.

[0040] Furthermore, the conflict avoidance is performed based on the selected node and the at least two mobile targets that generate the conflict, and at least two child nodes containing new joint trajectories are generated, including:

[0041] The selected node is used as a parent node, and at least two child nodes are created based on at least two conflicting moving targets, and the joint trajectory and constraint set of the at least two created child nodes are initialized to be the same as the joint trajectory and constraint set of the parent node, wherein one moving target corresponds to one child node;

[0042] Adding new conflict avoidance constraints to the at least two mobile targets that generate conflict, respectively, and adding the new conflict avoidance constraints to the constraint set included in the corresponding child node;

[0043] Replacing the planned trajectories of at least two moving targets with the Pareto optimal planned trajectories of at least two moving targets to generate a new joint trajectory;

[0044] When a new joint trajectory is not dominated by an existing feasible solution, the child node containing the new joint trajectory is added to the OPEN list, and the total cost contained in the child node is updated according to the new joint trajectory.

[0045] Furthermore, the Pareto optimal trajectories of the at least two moving targets are determined according to a multi-objective optimization method and a constraint set contained in the nodes corresponding to the at least two moving targets.

[0046] In a second aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, wherein the computer program / instruction, when executed by a processor, implements the steps of the method described in the first aspect.

[0047] In a third aspect, the present invention provides: a computer program product, comprising a computer program / instruction, characterized in that when the computer program / instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention obtains the operation plans of multiple mobile targets, uses multi-objective optimization to generate Pareto optimal planning trajectories, constructs multiple groups of joint trajectories through Cartesian products, and establishes a binary tree structure to perform extended search for conflict avoidance solutions. It not only takes into account the coordination and balance among multiple optimization objectives, but also constructs an effective search strategy for trajectory planning conflict avoidance problems, solves the mutual influence of conflicts between different mobile targets, and enables the conflict avoidance algorithm based on tree search to fully explore potential solutions.

[0050] The present invention also trains a node sorting neural network model to intelligently select nodes of a binary tree structure during the search process, thereby improving the efficiency of node evaluation and expansion. When there is a conflict in the joint trajectory represented by the selected node, the conflict is avoided by adding constraints, a new joint trajectory is generated, and corresponding child nodes are created, which not only effectively solves the conflict problem, but also ensures the diversity of solutions, so that the final output planning trajectory maintains Pareto optimality on multiple optimization objectives, and provides an effective technical solution for trajectory planning conflict avoidance. The present invention realizes the effective construction of a binary tree model and the reasonable selection of nodes for trajectory planning conflict avoidance. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of a trajectory planning conflict avoidance method driven by knowledge and data collaboration provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0053] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0054] Example 1

[0055] This embodiment introduces a trajectory planning conflict avoidance method driven by knowledge and data collaboration, including:

[0056] Step 1: Obtain operation plans of multiple mobile targets, and generate multiple sets of joint trajectories through Cartesian products based on the Pareto optimal planning trajectories of the multiple mobile targets according to the operation plans.

[0057] The present invention collects the operation plan information of all mobile targets in the current environment, such as vehicles, aircraft, robots, etc. The mobile target operation plan information can ensure the accuracy of trajectory planning and conflict avoidance.

[0058] Based on the Pareto optimality theory, the present invention comprehensively considers multiple optimization objectives, such as time, energy consumption, safety, etc., to generate the optimal planning trajectory for the mobile target. By calculating the Cartesian product of the planned trajectories of different mobile targets, the present invention generates all possible trajectory combinations, providing a rich and diverse selection space for the selection and expansion of subsequent nodes, thereby ensuring the feasibility and efficiency of trajectory planning and conflict avoidance.

[0059] Step 2: Initialize a root node for each set of joint trajectories and create a binary tree based on each root node.

[0060] The present invention regards each set of joint trajectories as the starting point of the search space, and further constructs a binary tree structure with this point as the root node. The establishment of this structure provides a clear and explicit framework for subsequent search and node expansion, making it easier and more efficient to manage and track different conflict avoidance schemes.

[0061] Step 3: Use OPEN list to store the root node of each binary tree.

[0062] The present invention stores the root nodes of all binary trees in an OPEN list as a set of nodes to be expanded. The OPEN list provides an effective way to manage and select nodes to be processed, ensuring the orderliness and efficiency of the search process.

[0063] Step 4: Repeat the following steps until the OPEN list is empty or the calculation time threshold is reached, and output a feasible solution set:

[0064] Using the pre-trained node sorting neural network model, select nodes from the OPEN list to expand:

[0065] When the joint trajectory included in the selected node does not have a trajectory conflict: the selected node is added to the feasible solution set and then removed from the OPEN list.

[0066] When the joint trajectory contained in the selected node has a trajectory conflict: the selected node is removed from the OPEN list, and based on the selected node and at least two mobile targets that cause the conflict, conflict avoidance is performed, at least two child nodes containing new joint trajectories are generated, and added to the OPEN list.

[0067] The present invention gradually approaches the optimal solution by continuously expanding nodes and evaluating the feasibility of the joint trajectory contained in the current node. In the process of iterative search, the search space is continuously narrowed until a satisfactory solution is found or the limitation of computing resources is reached. At the same time, the present invention uses neural networks for prediction and intelligently selects the most potential nodes from the OPEN list for expansion, which improves search efficiency and reduces unnecessary computing overhead.

[0068] When the joint trajectory represented by the selected node does not have trajectory conflict, it is directly added to the feasible solution set; when the joint trajectory represented by the selected node has trajectory conflict, a new joint trajectory is generated by creating child nodes and performing conflict avoidance, which effectively solves the conflict problem and ensures the diversity of solutions.

[0069] The joint trajectory contained in the selected node is a trajectory set, including a planned trajectory for each mobile target.

[0070] Step 5: Select the node with the minimum total cost from the feasible solution set, and assign conflict-free planning trajectories to each mobile target based on the joint trajectory contained in the node with the minimum total cost.

[0071] The present invention selects the node with the minimum total cost in the feasible solution set based on the comprehensive evaluation of the total cost, including time, energy consumption, safety and other indicators to determine the optimal trajectory planning scheme. This scheme assigns conflict-free planning trajectories to each mobile target. Subsequently, the joint trajectory contained in the node with the minimum total cost is converted into the actual motion instructions of each mobile target, ensuring that each mobile target can reach the destination safely and efficiently according to the planned trajectory.

[0072] The entire method provided by the present invention realizes efficient and intelligent trajectory planning conflict avoidance through the collaborative drive of knowledge and data, and provides strong support for the collaborative operation of multiple mobile targets in complex environments.

[0073] Example 2

[0074] Based on the same inventive concept as in Example 1, Figure 1 As shown, this embodiment introduces a trajectory planning conflict avoidance method driven by knowledge and data collaboration, which specifically includes the following steps:

[0075] Step 1: Get The invention provides an operation plan for a plurality of mobile targets, and according to the operation plan, generates a plurality of sets of joint trajectories through Cartesian products based on the Pareto optimal planning trajectories of the plurality of mobile targets.

[0076] The operation plan includes a starting position and a target position of the mobile object;

[0077] According to the operation plan, a plurality of Pareto optimal trajectories are planned for each moving target by a multi-objective optimization method;

[0078] According to the multiple Pareto optimal trajectories, the Cartesian product between all the moving target planning trajectories is calculated to generate Group joint trajectory ,in, represents the first set of joint trajectories, represents the second set of joint trajectories, Indicates Group joint trajectory.

[0079] Step 2: Initialize a root node by grouping joint trajectories , a binary tree is created based on each root node; each node of the binary tree includes a triple, and the triple includes the joint trajectory of the mobile target Total cost and constraint set ,in, .

[0080] The joint trajectory is a set of trajectories, including one planned trajectory for each moving target;

[0081] The total consideration For each set of joint trajectories The numerical representation of the sum of the costs of each planned trajectory in is used to evaluate each set of joint trajectories Overall performance;

[0082] The constraint set is a set of constraints for each moving target when planning the trajectory, which is initially set to an empty set.

[0083] Step 3: Use the OPEN list to store the root node of each binary tree;

[0084] Step 4: Repeat the following steps until the OPEN list is empty or the calculation time threshold is reached, and output a feasible solution set:

[0085] Using the pre-trained node sorting neural network model, select nodes from the OPEN list to expand:

[0086] When the joint trajectory included in the selected node does not have trajectory conflict: adding the selected node to the feasible solution set and then removing it from the OPEN list;

[0087] When the joint trajectory included in the selected node has a trajectory conflict: removing the selected node from the OPEN list, and creating at least two child nodes based on the selected node and at least two mobile targets generating the conflict, including:

[0088] The selected node is used as a parent node, and at least two child nodes are created based on at least two conflicting moving targets, and the joint trajectory and constraint set of the at least two created child nodes are initialized to be the same as the joint trajectory and constraint set of the parent node, wherein one moving target corresponds to one child node;

[0089] Adding new conflict avoidance constraints to the at least two mobile targets that generate conflict, respectively, and adding the new conflict avoidance constraints to the constraint set included in the corresponding child node;

[0090] Replacing the planned trajectories of at least two moving targets with the Pareto optimal planned trajectories of at least two moving targets to generate a new joint trajectory;

[0091] When a new joint trajectory is not dominated by an existing feasible solution, the child node containing the new joint trajectory is added to the OPEN list, and the total cost contained in the child node is updated according to the new joint trajectory.

[0092] For example, select a node from the OPEN list To expand, when the selected node The joint trajectory If there is no conflict, a new feasible solution is found. , the selected node Add feasible solution set, where Represents the index of a node in a binary tree.

[0093] When the selected node contains two planned trajectories in the joint trajectory and There are track conflicts: Based on the selected nodes Create two child nodes and ,in, , is the index of the created child node in the binary tree. Corresponding to a moving target x in the conflict, Corresponding to the other moving target y in the conflict, initialize the child nodes respectively and The joint trajectory and constraint set of are the same as those of the parent node.

[0094] Add a new collision avoidance constraint for the moving target x , and join the selected node The constraint set , and then determine the Pareto optimal trajectory of the moving target x after adding constraints ,in, Indicates that the new conflict avoidance constraint added in the constraint set The index in .

[0095] Pareto optimal trajectory of moving target x after adding constraints Replace the original planned trajectory of the moving target x , generating a new joint trajectory .

[0096] When the new joint trajectory Not dominated by existing feasible solutions, the child nodes Add to the OPEN list.

[0097] Similarly, the trajectory of the moving target y is also recalculated for conflict avoidance.

[0098] The node sorting neural network model includes:

[0099] The first deep neural network, built on a gated recurrent unit, is used to train the joint trajectory of each node in the OPEN list. The planned trajectories are feature encoded to obtain multiple trajectory feature vectors ,in, represents the first trajectory feature vector, represents the second trajectory eigenvector, represents the last trajectory feature vector.

[0100] The second deep neural network is constructed based on a fully connected layer and a ReLU activation function, and is used to transform the plurality of trajectory feature vectors After concatenation, the scores of the corresponding nodes are obtained through softmax operation, and the nodes are sorted according to the scores. The nodes with larger scores have higher sorting priority.

[0101] The node sorting neural network model is trained using training data with sample labels to obtain a pre-trained node sorting neural network model.

[0102] The method for acquiring training data of the node-sorted neural network model comprises:

[0103] Obtain historical operation plans for multiple mobile targets;

[0104] According to the historical operation plan, generating multiple sets of joint trajectories through Cartesian products based on Pareto optimal planning trajectories of multiple mobile targets according to the operation plan;

[0105] Initialize a root node for each set of joint trajectories and create a binary tree based on each root node;

[0106] Use OPEN list to store the root node of each binary tree;

[0107] Repeat the following steps until the OPEN list is empty:

[0108] Selecting a node from the OPEN list using any heuristic strategy;

[0109] When the joint trajectory included in the selected node does not have trajectory conflict: the selected node is taken as a leaf node of the binary tree and added to the feasible solution set, and then the selected node is removed from the OPEN list;

[0110] When the joint trajectory included in the selected node has a trajectory conflict: remove the selected node from the OPEN list, and based on the selected node and at least two mobile targets that generate the conflict, perform conflict avoidance, generate at least two child nodes including new joint trajectories, and add them to the OPEN list;

[0111] For all leaf nodes included in the feasible solution set, trace back from the leaf nodes to the root node in turn, record the information of each intermediate node, and generate training data with sample labels.

[0112] The sample label is the shortest distance from each intermediate node to a leaf node in its sub-binary tree.

[0113] Step 5: Select the node with the minimum total cost from the feasible solution set, and assign conflict-free planning trajectories to each mobile target based on the joint trajectory contained in the node with the minimum total cost.

[0114] Example 3

[0115] With the same inventive concept as other embodiments, this embodiment introduces a computer-readable storage medium on which computer instructions are stored, characterized in that when the computer instructions are executed by a processor, the steps of the method of the above-mentioned embodiment 1 or 2 are implemented.

[0116] Example 4

[0117] With the same inventive concept as other embodiments, the present invention further provides a computer program product, including computer instructions, characterized in that when the computer instructions are executed by a processor, the steps of the method in the above-mentioned embodiment 1 or 2 are implemented.

[0118] In summary, the present invention obtains the operation plans of multiple mobile targets, uses multi-objective optimization to generate Pareto optimal planning trajectories, constructs multiple groups of joint trajectories through Cartesian products, and establishes a binary tree structure to perform extended search for conflict avoidance solutions. It not only takes into account the coordination and balance among multiple optimization objectives, but also constructs an effective search strategy for trajectory planning conflict avoidance problems, solves the mutual influence of conflicts between different mobile targets, and enables the conflict avoidance algorithm based on tree search to fully explore potential solutions.

[0119] The present invention also trains a node sorting neural network model to intelligently select nodes of a binary tree structure during the search process, thereby improving the efficiency of node evaluation and expansion. When there is a conflict in the joint trajectory represented by the selected node, the conflict is avoided by adding constraints, a new joint trajectory is generated, and corresponding child nodes are created, which not only effectively solves the conflict problem, but also ensures the diversity of solutions, so that the final output planning trajectory maintains Pareto optimality on multiple optimization objectives, and provides an effective technical solution for trajectory planning conflict avoidance. The present invention realizes the effective construction of a binary tree model and the reasonable selection of nodes for trajectory planning conflict avoidance.

[0120] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0122] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0123] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0124] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A knowledge and data-cooperatively driven trajectory planning conflict avoidance method, characterized in that: include: Obtaining operation plans for multiple moving targets; According to the operation plan, based on the Pareto optimal planning trajectories of multiple moving targets, multiple sets of joint trajectories are generated through Cartesian products; Initialize a root node for each set of joint trajectories and create a binary tree based on each root node; Use OPEN list to store the root node of each binary tree; Repeat the following steps until the OPEN list is empty or the calculation time threshold is reached, and then output a feasible solution set: Using the pre-trained node sorting neural network model, select nodes from the OPEN list to expand: When the joint trajectory included in the selected node does not have trajectory conflict: adding the selected node to the feasible solution set and then removing it from the OPEN list; When the joint trajectory included in the selected node has a trajectory conflict: remove the selected node from the OPEN list, and based on the selected node and at least two mobile targets that generate the conflict, perform conflict avoidance, generate at least two child nodes including new joint trajectories, and add them to the OPEN list; Select the node with the minimum total cost from the feasible solution set, and assign conflict-free planning trajectories to each mobile target according to the joint trajectory contained in the node with the minimum total cost; The joint trajectory contained in the selected node is a trajectory set, including a planned trajectory for each mobile target.

2. The knowledge and data collaboratively driven trajectory planning conflict avoidance method according to claim 1, characterized in that: According to the operation plan, based on the Pareto optimal planning trajectories of multiple moving targets, multiple sets of joint trajectories are generated through Cartesian products, including: The operation plan includes a starting position and a target position of the mobile object; Generating a Pareto optimal planning trajectory for each mobile target by a multi-objective optimization method according to the starting position and the target position; The Cartesian product is calculated based on the Pareto optimal planning trajectory of each mobile target to generate multiple sets of joint trajectories.

3. The knowledge and data collaboratively driven trajectory planning conflict avoidance method according to claim 1, characterized in that: Each node of the binary tree includes a triplet, and the triplet includes a joint trajectory of the moving target, a total cost, and a constraint set; The joint trajectory is a set of trajectories, including a planned trajectory for each mobile target; The total cost is a numerical representation of the sum of the costs of each planned trajectory in each set of joint trajectories, and is used to evaluate the overall performance of each set of joint trajectories; The constraint set is a set of constraint conditions for each moving target when planning a trajectory, and is initially set to an empty set.

4. The knowledge and data collaboratively driven trajectory planning conflict avoidance method according to claim 1, characterized in that: The node sorting neural network model includes: The first deep neural network is constructed based on the gated recurrent unit and is used to perform feature encoding on each planned trajectory in the joint trajectory contained in the nodes in the OPEN list to obtain multiple trajectory feature vectors; The second deep neural network is constructed based on the fully connected layer and the ReLU activation function, and is used to concatenate the multiple trajectory feature vectors, obtain the scores of the nodes in the OPEN list through softmax operation, and sort them according to the scores, wherein the nodes with larger scores have higher sorting priority.

5. The knowledge and data collaboratively driven trajectory planning conflict avoidance method according to claim 1, characterized in that: The method for acquiring training data of the node-sorted neural network model comprises: Obtain historical operation plans for multiple mobile targets; According to the historical operation plan, based on the Pareto optimal planning trajectories of multiple moving targets, multiple sets of joint trajectories are generated through Cartesian products; Initialize a root node for each set of joint trajectories and create a binary tree based on each root node; Use OPEN list to store the root node of each binary tree; Repeat the following steps until the OPEN list is empty: Selecting a node from the OPEN list using any heuristic strategy; When the joint trajectory included in the selected node does not have trajectory conflict: the selected node is taken as a leaf node of the binary tree and added to the feasible solution set, and then the selected node is removed from the OPEN list; When the joint trajectory included in the selected node has a trajectory conflict: remove the selected node from the OPEN list, and based on the selected node and at least two mobile targets that generate the conflict, perform conflict avoidance, generate at least two child nodes including new joint trajectories, and add them to the OPEN list; For all leaf nodes included in the feasible solution set, trace back from the leaf nodes to the root node in turn, record the information of each intermediate node, and generate training data with sample labels.

6. The knowledge and data collaboratively driven trajectory planning conflict avoidance method according to claim 5, characterized in that: The sample label is the shortest distance from each intermediate node to its child binary tree leaf node.

7. The knowledge and data collaboratively driven trajectory planning conflict avoidance method according to claim 1, characterized in that: The step of performing conflict avoidance based on the selected node and at least two conflicting mobile targets to generate at least two sub-nodes containing new joint trajectories includes: The selected node is used as a parent node, and at least two child nodes are created based on at least two conflicting moving targets, and the joint trajectory and constraint set of the at least two created child nodes are initialized to be the same as the joint trajectory and constraint set of the parent node, wherein one moving target corresponds to one child node; Adding new conflict avoidance constraints to the at least two mobile targets that generate conflict, respectively, and adding the new conflict avoidance constraints to the constraint set included in the corresponding child node; Replacing the planned trajectories of at least two moving targets with the Pareto optimal planned trajectories of at least two moving targets to generate a new joint trajectory; When a new joint trajectory is not dominated by an existing feasible solution, the child node containing the new joint trajectory is added to the OPEN list, and the total cost contained in the child node is updated according to the new joint trajectory.

8. The knowledge and data collaboratively driven trajectory planning conflict avoidance method according to claim 7, characterized in that: The Pareto optimal trajectories of the at least two moving targets are determined according to a multi-objective optimization method and a constraint set contained in the nodes corresponding to the at least two moving targets.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images