Multi-panel collaborative planning method in complex scene based on dynamic priority

Abstract

The invention relates to the technical field of road traffic networks and multi-vehicle collaboration, and provides a multi-panel collaborative planning method in a complex scene based on dynamic priority, which comprises the following steps: A, constructing an undirected topological graph of a dynamic complex scene and a list of tasks to be executed, and updating the topological graph of the dynamic complex scene in real time based on environmental information; b, performing double-layer dynamic evaluation on a to-be-executed task; c, constructing a self-adaptive auction algorithm, inputting the input variable into the self-adaptive auction algorithm, and generating a task allocation result of the flat car cluster; d, the flat car cluster generates a collaborative planning path based on the dynamic complex scene topological graph updated in real time and the task allocation result, the collaborative planning path is converted into a control instruction sequence, and the control instruction sequence is used for conducting collaborative control on the flat car cluster. According to the invention, the real-time performance, the task completion rate and the driving safety of collaborative operation of the flat car cluster can be improved.

Description

Technical Field

[0001] This invention relates to the fields of road traffic networks and multi-vehicle cooperative technology, specifically to a multi-platform cooperative planning method for complex scenarios based on dynamic priority. Background Technology

[0002] With the rapid development of autonomous driving technology, autonomous driving test platforms are playing an increasingly crucial role in simulating real-world traffic scenarios and verifying system functions and performance. To ensure the safety and reliability of autonomous driving systems, extensive and thorough testing is essential. This testing not only requires coverage of diverse typical traffic scenarios but also the ability to reproduce various extreme or unexpected environmental conditions (such as severe weather, sudden obstacles, and complex interactions). Against this backdrop, flatbed trucks (i.e., autonomous driving test platforms) are widely used as a core testing tool to construct controllable, repeatable, and high-fidelity road and traffic environments to systematically evaluate the perception, decision-making, and control capabilities of autonomous driving systems under different operating conditions.

[0003] With the continuous upgrading of testing requirements and the ongoing expansion of testing scale, the testing scenarios provided by a single flatbed truck are no longer sufficient to meet the requirements for comprehensiveness, robustness, and edge case coverage of autonomous driving systems. Therefore, multi-flatbed truck collaborative operation has gradually become a necessary means to improve testing efficiency and scenario complexity. However, how to efficiently plan and schedule the collaborative behavior among multiple flatbed trucks to achieve integrated control of task allocation, path coordination, and dynamic collision avoidance has become a key technical problem that urgently needs to be solved in the current field of autonomous driving testing.

[0004] In existing technologies, some progress has been made in multi-vehicle cooperative planning. For example, a multi-vehicle cooperative planning method based on distributed swarm intelligence learning, patent application number CN202111563958.8, introduces an evolutionary game mechanism to model the interaction and competition between vehicles in the path selection process. When the system reaches game equilibrium, each vehicle can obtain a routing decision that balances individual interests and group efficiency, thereby improving the adaptability and robustness of the overall path planning. Another example is a multi-agent path planning method, patent application number CN202111602040.X, which uses a composite neural network architecture (integrating CNN, GraphSAGE, and MLP) to extract and fuse multi-level features from environmental perception information, thereby generating a globally optimized driving path.

[0005] While the aforementioned agent-based collaborative strategies demonstrate good performance in static or semi-structured scenarios, they still have significant limitations when facing real-world testing environments: First, insufficient dynamic adaptability—most methods only consider changes in static objects in the environment and their own behavior, which can easily lead to restricted movement or reduced safety. Second, low computational efficiency; current deep learning or game theory-driven methods experience significantly increased computational overhead when dealing with large-scale agent clusters or high-dimensional complex environments, resulting in high decision-making latency and failing to meet the demands of real-world scenarios. Summary of the Invention

[0006] The present invention aims to provide a multi-flatbed collaborative planning method and system based on dynamic priority in complex scenarios, which effectively solves the problems of poor environmental adaptability, unreasonable task scheduling, resource waste and path conflict of multi-flatbed vehicles in complex dynamic scenarios, and improves the real-time performance, task completion rate and driving safety of flatbed vehicle cluster collaborative operation.

[0007] The first approach is a multi-flatbed collaborative planning method for complex scenarios based on dynamic priority, comprising: A: constructing an undirected topology graph of the dynamic complex scenario and a list of tasks to be executed for the flatbed truck cluster, obtaining environmental and location information for each flatbed truck, and updating the dynamic complex scenario topology graph in real time based on the environmental information; B: performing a two-layer dynamic evaluation on the tasks to be executed, the two-layer dynamic evaluation including dynamic scenario complexity evaluation and dynamic priority evaluation, to obtain the dynamic scenario complexity and dynamic priority of the tasks to be executed; C: constructing an adaptive auction algorithm using the dynamic scenario complexity, dynamic priority, and location information of the multiple flatbed trucks as input variables, inputting the input variables into the adaptive auction algorithm to generate the task allocation result of the flatbed truck cluster; D: generating a collaborative planning path for the flatbed truck cluster based on the real-time updated dynamic complex scenario topology graph and the task allocation result, and converting the collaborative planning path into a control command sequence, which is used to perform collaborative control of the flatbed truck cluster.

[0008] Beneficial Effects: By implementing a complete process of "dynamic topology graph construction and updating - two-layer dynamic evaluation - adaptive auction algorithm task allocation - collaborative path generation and control," the core problem of multi-flatbed truck collaborative planning in complex dynamic scenarios is systematically solved. It achieves real-time perception and response to the dynamic environment, and through a multi-dimensional factor-integrated task evaluation and optimization allocation mechanism, it balances the timeliness of task execution, the rationality of resource matching, and the safety of multi-truck collaboration. This effectively compensates for the shortcomings of existing technologies, such as insufficient dynamic adaptability and low computational efficiency, and significantly improves the collaborative operation efficiency and robustness of multi-flatbed trucks in complex scenarios.

[0009] Preferably, the first key node of the dynamic complex scene, the node paths between passable key nodes, and the additional attributes of the node paths are obtained to construct an undirected topology graph of the dynamic complex scene; the types of the first key node of the dynamic complex scene include intersections, corners, task locations, static objects, and dynamic objects; the first key node includes static nodes and dynamic nodes; the types of static nodes include intersections, corners, task locations, and static objects; the environmental information of the dynamic complex scene is perceived by the flatbed vehicle, and the dynamic nodes are updated in real time; the type of dynamic node is a dynamic object, and the dynamic object is represented as an object that performs irregular movements in the dynamic complex scene; the additional attributes of the node paths include path length and path width; the types of passable key nodes include intersections, corners, and task locations.

[0010] Beneficial Effects: By clarifying the construction elements of an undirected topology graph for dynamic and complex scenarios, and refining key node types (static / dynamic), walkable node paths, and additional attributes, a structured and precise representation of the scene is achieved. The real-time update mechanism for dynamic nodes accurately captures the state changes of irregularly moving objects, and additional attributes such as path width provide direct evidence for path feasibility judgments. This enables the topology graph to truly reflect dynamic changes in the scene, providing highly reliable environmental foundation data for subsequent task evaluation, task allocation, and path planning, significantly improving the targeting and accuracy of collaborative planning.

[0011] Preferably, by installing multi-source sensors on a single flatbed truck, the environmental and positional information of the flatbed truck is perceived in real time by the multi-source sensors, and the dynamic complex scene topology map is updated in real time based on the environmental information. This includes: the multi-source sensors include a lidar, a pose sensor, a positioning device, and a communication module; the lidar emits laser pulses around the flatbed truck, and environmental point cloud data of spatial points around the flatbed truck is generated based on the reflected laser signals, the environmental point cloud data including several spatial points; the RANSAC method is used to fit the ground plane, spatial points belonging to the ground are removed, and Euclidean clustering is used to cluster the remaining spatial points to obtain multiple spatial point clusters. Spatial point clusters represent environmental objects surrounding the flatbed vehicle; a positioning device acquires the flatbed vehicle's position information, and uses a communication module to analyze the spatial point clusters and the flatbed vehicle's position information to calculate the first geometric center coordinates of the spatial point clusters; a pose sensor senses the Euler angles of the flatbed vehicle, calculates the flatbed vehicle's pose matrix using the Euler angles, and constructs a pose estimation transformation matrix by combining the flatbed vehicle's position coordinates; the pose estimation transformation matrix is ​​used to convert the first geometric center coordinates into second geometric center coordinates, which are the coordinates in the world coordinate system where the first key node is located, and the minimum distance between the second geometric center coordinates and the static node position coordinates is calculated.

[0012] Beneficial Effects: Leveraging multi-source sensor fusion sensing technology and algorithms such as RANSAC ground fitting and Euclidean clustering, high-precision extraction and coordinate transformation of environmental information were achieved. Point cloud data was acquired using LiDAR and environmental objects were filtered. A transformation matrix was constructed using pose sensors and positioning devices to map local coordinates to the world coordinate system, ensuring the accuracy of key node location information. Simultaneously, distance threshold judgment enabled precise differentiation between static and dynamic nodes, guaranteeing the real-time and effective updates of the topology map. This provided accurate and real-time environmental perception support for multi-flatbed collaborative planning, enhancing the safety and reliability of the planning process.

[0013] Preferably, tasks to be executed, along with their location, generation time, and expected completion time, are extracted from the list of tasks to be executed. Calculate the scene complexity of the location where the task to be performed is to be executed, and the scene complexity of the flatbed truck closest to the location: ; ; in, Indicates the location of the task to be performed. The complexity of the scene, Representing the topology of a dynamic and complex scene by task location Centered on, with radius The number of static nodes in the circular region. Representing the topology of a dynamic and complex scene by task location Centered on, with radius The number of dynamic nodes not on the flatbed vehicle within the circular region. Indicates the dynamic node weight factor; This indicates the location of the flatbed truck closest to the task's destination. Representing the topology of a dynamic and complex scene using the flatbed truck's position Centered on, with radius The number of static nodes in the circular region; Representing the topology of a dynamic and complex scene using the flatbed truck's position Centered on, with radius The number of dynamic nodes in the circular region; Indicates distance from mission location The dynamic scene complexity of recent flatbed trucks; The dynamic scene complexity of the task location is obtained by weighting the scene complexity of the location to be executed and the scene complexity of the flatbed truck closest to the task location.

[0014] Beneficial Effects: By weighted and fused with the complexity of the task location scenario and the surrounding environment of the nearest flatbed truck, a more comprehensive dynamic scenario complexity assessment system is constructed. This system considers not only the obstacle density (static + dynamic) at the task execution location itself, but also the complexity of the environment in which the flatbed truck is currently performing the task. This ensures that the complexity assessment reflects both the difficulty of the task itself and the difficulty of resource mobilization. This dual-point real-time perception assessment method adapts to the uncertainty of dynamic scenarios, providing a scientific basis for subsequent task priority calculation and task allocation, and avoiding the planning irrationality caused by a single assessment dimension.

[0015] Preferably, the urgency index of the task to be executed is calculated based on the current time, the expected completion time of the task, and the task generation time. Based on the current time and the time the task to be executed was generated, the waiting time index of the task to be executed is calculated. The urgency index of the task to be performed is obtained by using a nonlinear function. and waiting time index The mapping process is performed to obtain the urgency factor and waiting penalty term of the task to be executed, and combined with the dynamic scene complexity of the task location, the dynamic priority of the task to be executed is generated: ; ; ;in, Indicates the dynamic priority of tasks to be executed. Factors indicating the urgency of tasks to be performed. This indicates the pending penalty items for tasks to be performed. Indicates the location of the task to be performed. Indicates the location of the mission The complexity of dynamic and complex scenarios, This represents the Sigmoid function, which... As a scenario factor for the task to be performed, Indicates the urgency coefficient. This represents the waiting penalty coefficient. Represents the scene coefficient. This represents the nonlinear control coefficient.

[0016] Beneficial Effects: By introducing a triple-indicator system of urgency, waiting penalty, and scenario factors, a multi-dimensional dynamic priority evaluation mechanism was constructed. The nonlinear function mapping process rapidly amplifies the weight of the urgency factor when a task approaches its timeout boundary, the waiting penalty effectively suppresses task starvation, and the scenario factor ensures that complex tasks receive appropriate resource allocation. This mechanism overcomes the limitations of existing single-dimensional ranking technologies, achieving a three-dimensional balance of timeliness, fairness, and environmental resources. It improves the response speed of time-sensitive tasks while ensuring the integrity and rationality of the task queue, significantly optimizing the responsiveness and completion rate of multi-cart task scheduling.

[0017] Preferably, a task revenue function for the tasks to be executed is constructed. The task revenue function uses the dynamic scene complexity and dynamic priority of the tasks to be executed as fixed parameters and the real-time position of the flatbed truck as a variable. The output of the task revenue function is the revenue generated by different flatbed trucks executing tasks. The revenue generated by the flatbed trucks executing any task to be executed is calculated using the task revenue function, and the task to be executed with the highest current revenue is selected. The flatbed truck submits a bidding request to the selected task to be executed. The task to be executed selects the flatbed truck with the highest revenue from all the flatbed trucks that have submitted bidding requests as the executor, and removes the task to be executed with an executor from the task to be executed list, and proceeds to the next round of bidding until the task to be executed list is empty.

[0018] Beneficial Effects: The constructed adaptive auction algorithm achieves optimal matching between tasks and flatbed trucks, with task revenue as the core. Through a two-way matching mechanism—flatbed trucks actively bidding for the task with the highest revenue and tasks selecting the optimal executor—combined with conflict resolution rules, task allocation is completed efficiently, avoiding resource waste and allocation conflicts. Its dynamic adaptation to the real-time status of tasks and flatbed trucks ensures the flexibility and optimization of task allocation. Compared to traditional allocation methods, it significantly improves the resource utilization and task execution efficiency of multi-flatbed truck clusters, making it particularly suitable for dynamic and complex operational scenarios.

[0019] Preferably, the expression for the task reward function is: ; ; ;in, This represents the revenue generated by the nth flatbed truck performing the task to be performed. This indicates the normalized dynamic priority of the tasks to be executed. This represents the normalized dynamic scenario complexity of the task to be executed. This represents the interactive penalty items for the task to be performed. This represents the normalized scheduling cost of dispatching the nth flatbed truck to the location of the task to be performed. N represents the total number of flatbed trucks; Indicates the position of the nth flatbed truck With tasks to be performed Mission location The shortest path distance between them This indicates the maximum preset path distance.

[0020] Beneficial Effects: A clearly defined task reward function, through quantifying dynamic priority, dynamic scenario complexity, scheduling costs, and interaction penalties, enables precise calculation of task rewards. The interaction penalty effectively characterizes the scheduling risk of "high complexity, low priority" tasks, and normalization processing allows for direct fusion of factors from different dimensions, ensuring the scientific rigor and comparability of reward assessment. This function provides a core quantitative basis for the adaptive auction algorithm, making the matching of flatbed trucks and tasks more closely aligned with actual needs, avoiding inefficiencies or task failures caused by unreasonable scheduling, and further improving the overall optimization level of multi-flatbed truck collaborative planning.

[0021] Preferably, a heuristic search algorithm is used to generate multiple candidate planning paths for each flatbed truck in a dynamic and complex scene topology map. The candidate planning paths start from the current position of the flatbed truck and end at the task location of the task to be performed by the flatbed truck. Calculate the conflict factor between candidate planned paths for different flatbed trucks; Enumerate candidate planning path combinations consisting of at least one flatbed truck, and calculate the global path conflict score for each candidate planning path combination. The candidate planning path combination consists of a group of candidate planning paths consisting of at least one flatbed truck. The global path conflict score is the sum of the conflict factors of any two candidate planning path combinations in the candidate planning path combination. The candidate path combination with the lowest global path conflict score is selected as the collaborative path combination for the N flatbed trucks: ,in This represents the collaborative planning path for the nth flatbed truck. N is the number of flatbed trucks.

[0022] Beneficial Effects: By generating multiple candidate paths through heuristic search and selecting the optimal combination based on global path conflict scores, efficient planning of collaborative paths for multiple flatbed trucks is achieved. With the goal of minimizing conflict factors, it ensures that the selected path combination minimizes driving conflicts between multiple vehicles by enumerating path combinations and calculating the total conflict. This path planning method not only guarantees the rationality of individual vehicle paths but also highlights the collision avoidance requirements of multi-vehicle collaboration, effectively improving the safety and traffic efficiency of multi-flatbed truck operation and adapting to the path coordination needs of multi-vehicle parallel operations in complex scenarios.

[0023] Preferably, the conflict factor includes: ; in, This represents the m-th candidate planning path for the nth flatbed truck. The g-th candidate path planning group for the q-th flatbed truck Conflict factors between them , , , N represents the total number of flatbed trucks. Indicate candidate planning paths The set of static nodes visited Indicate candidate planning paths The set of static nodes visited This indicates the number of nodes in the set. Indicate candidate planning paths Path length, Indicate candidate planning paths The path length.

[0024] Beneficial Effects: The conflict factor calculation method accurately quantifies the degree of conflict between different flatbed truck paths by integrating the overlap of static node sets traversed by candidate paths with the difference in path length. Node set overlap reflects the risk of spatial conflict along the path, while the difference in path length reflects the possibility of temporal conflict. Combining these two factors makes conflict assessment more comprehensive and accurate. This calculation method provides a reliable quantitative standard for selecting cooperative path combinations, ensuring that the selected path combinations have the lowest conflict risk, and providing key technical support for the safety of multi-flatbed truck cooperative driving.

[0025] Preferably, It is 0.4. It is 0.3. It is 0.2.

[0026] Beneficial effects: The specific ratios of urgency coefficient, waiting penalty coefficient, and scenario coefficient in dynamic priority calculation were clarified. This parameter combination, after optimization and verification in actual scenarios, can achieve a dynamic balance of the three evaluation indicators. A reasonable coefficient allocation highlights the core position of task urgency while also taking into account the impact of waiting fairness and scenario complexity, avoiding planning imbalances caused by excessive weighting of a single indicator. These clarified parameters make dynamic priority calculation directly implementable, improving the practicality and stability of the entire collaborative planning method.

[0027] Beneficial effects of the first option First, this solution utilizes a dynamic topology map update method based on LiDAR point cloud clustering to achieve real-time perception and structured representation of obstacle information in dynamic and complex scenes. Compared to existing static topology maps that rely on pre-set maps and cannot reflect real-time environmental changes, this solution deploys LiDAR on a flatbed truck to acquire high-precision 3D point cloud data and uses Euclidean clustering to extract effective obstacle targets. Furthermore, it maps the spatial location and size features of dynamic objects onto the topology map, achieving dynamic modeling of traffic risk states. This effectively improves the adaptability and expressive power of the topology map in dynamic and complex scenes to the real environment. This method possesses good real-time performance and versatility, supports synchronous updates of perception results from multiple flatbed trucks, and provides highly reliable environmental information support for subsequent task scheduling and path generation, enhancing safety in the multi-flatbed collaborative planning process.

[0028] Meanwhile, in the dynamic priority calculation process of tasks to be executed, this solution introduces three indicators: urgency factor, waiting penalty term, and scenario factor, to construct a multi-dimensional fusion and dynamically adaptive priority evaluation mechanism. This effectively solves the problems of single task sorting basis, response delay, or unreasonable resource matching in existing scheduling strategies. The urgency factor is used to quantify the urgency of the current task to be executed from its expected completion time. It is defined as the proportion of the remaining available time of the task to the total execution window. This factor adopts a non-linear function modeling method, which rapidly amplifies its weight when the task is close to the timeout boundary, thereby prompting the system to prioritize scheduling those time-sensitive tasks and improve the timeliness guarantee capability of the overall scheduling system.

[0029] Secondly, the waiting penalty term is used to dynamically suppress task starvation. When a task has not been executed for a long time, its accumulated waiting time will significantly increase its priority. Even if its urgency is low, it can still receive reasonable resource allocation, thus ensuring the fairness and integrity of the task queue. This penalty term uses a logarithmic function model to effectively control the pace of priority inflation, making task scheduling more stable. The scenario factor is scored based on the dynamic scenario complexity. Tasks with higher complexity have lower success rates or higher completion costs, so they receive a moderate increase in priority ranking, ensuring that complex tasks are not ignored for a long time and avoiding the local optimal scheduling trap of favoring simple tasks. Compared with the existing scheduling methods that only consider single factors such as task type or execution time for priority scoring, this scheme achieves three-dimensional fusion of timeliness, task history waiting time, and environmental resource dimensions. This mechanism can not only more realistically reflect the urgency of task scheduling, but also achieve dynamic balance and reasonable competition of task priorities in the resource scheduling strategy, greatly improving the scheduling responsiveness and task completion rate of multi-flatbed trucks in dynamic and complex scenarios.

[0030] The second solution is a multi-panel collaborative planning system for dynamic complex scenarios based on dynamic priority, including a data analysis module. The data analysis module includes an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it enables the electronic device to implement a multi-panel collaborative planning method for complex scenarios based on dynamic priority, as described in the first solution.

[0031] Beneficial Effects: By constructing a collaborative planning system that includes a data analysis module and corresponding electronic devices (memory, processor, computer program), the multi-flatbed collaborative planning method of the first scheme is implemented as a technically feasible engineering tool. Through the organic combination of hardware (sensors, processors, etc.) and software (computer program), the system ensures the efficient operation of core steps such as dynamic topology updates, task evaluation, task allocation, and path planning, realizing the transformation from method to application. It not only guarantees the stable execution and real-time response of the planning method but also possesses good scalability and adaptability, meeting the collaborative operation needs of flatbed truck clusters of different sizes in complex dynamic scenarios, providing reliable technical support for practical applications. Attached Figure Description

[0032] Figure 1 This is a schematic diagram illustrating an implementation method for a multi-flat collaborative planning method in complex scenarios based on dynamic priorities. Figure 2 This is a schematic diagram of a multi-panel collaborative planning system for dynamic complex scenarios based on dynamic priorities, as described in Embodiment 3. Figure 3 This is a schematic diagram of an electronic device structure for a multi-tablet collaborative planning system in a dynamic and complex scenario based on dynamic priority, as described in Embodiment 3.

[0033] The reference numerals in the accompanying drawings include: Processor 101, input device 102, output device 103, memory 104, bus 105, computer program 1041. Detailed Implementation

[0034] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0035] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0036] Example 1 This embodiment provides a multi-plank collaborative planning method for complex scenarios based on dynamic priority, including: A: Construct an undirected topology graph of a dynamic and complex scene, and plan a list of tasks to be executed for the flatbed truck cluster. Obtain the environmental and location information of each flatbed truck, and update the dynamic and complex scene topology graph in real time based on the environmental information.

[0037] Specifically, the first key node of the dynamic complex scene, the node paths between passable key nodes, and the additional attributes of the node paths are obtained to construct an undirected topology graph of the dynamic complex scene. The types of the first key node in the dynamic complex scene include intersections, corners, task locations, static objects, and dynamic objects. The first key node includes both static and dynamic nodes. Static nodes include intersections, corners, task locations, and static objects. The flatbed vehicle senses the environmental information of the dynamic complex scene and updates the dynamic nodes in real time. Dynamic nodes are classified as dynamic objects, which are characterized as objects undergoing irregular movement in the dynamic complex scene. The additional attributes of the node paths include path length and path width. Passable key nodes include intersections, corners, and task locations. Furthermore, if the path width is less than the width of the flatbed vehicle, it indicates that the node path cannot be used as a driving path for the flatbed vehicle. In this embodiment, static objects are obstacles in the dynamic complex scene, and dynamic objects include obstacles moving in real time and the flatbed vehicle itself.

[0038] Furthermore, specifically, the representation of a dynamic complex scene topology graph is as follows: ; ; in, This represents the dynamic complex scene topology graph updated at time t. This represents the set of key nodes in a dynamic and complex scenario at time t. Represents a static set of nodes. This represents the dynamic set of nodes at time t. This represents the c-th static node in the set of static nodes. This represents the number of static nodes in a dynamic and complex scenario. This represents the i-th dynamic node updated at time t. This represents the number of dynamic nodes updated at time t.

[0039] Set of key nodes The set of node paths formed by the node paths between any two passable key nodes. In this embodiment, the node paths are straight-line paths. For the set of node paths The set of additional attributes formed by the additional attributes of any node path in the middle.

[0040] By installing multi-source sensors on a single flatbed truck, the environmental and location information of the flatbed truck can be perceived in real time, and the dynamic and complex scene topology map can be updated in real time based on the environmental information, including: The multi-source sensors include a lidar, an IMU pose sensor, a GPS positioning device, and a communication module. The environmental information of the flatbed truck is perceived in real time using these sensors. The perception process is as follows: The lidar emits laser pulses around the flatbed truck, and the reflected laser signals generate environmental point cloud data of spatial points around the truck. This point cloud data consists of several spatial points, each with three-dimensional coordinates. The RANSAC method is used to fit the ground plane, removing points belonging to the ground, and Euclidean clustering is used to cluster the remaining points, resulting in multiple point clusters. These clusters are considered as environmental objects around the flatbed truck. The communication module uploads the point clusters and the flatbed truck's GPS position information to the central control system, which calculates the geometric center coordinates of the point clusters. The IMU pose sensor perceives the Euler angles of the flatbed truck and uploads them to the central control system. The central control system uses the Euler angles to calculate the flatbed truck's attitude matrix, and combines this with the GPS positioning device's position coordinates in the world coordinate system to construct a GPS / IMU pose estimation transformation matrix.

[0041] If the minimum value is less than the preset distance threshold, it means that the environmental object corresponding to the spatial point cluster is a static node in the dynamic complex scene topology graph, and the spatial point cluster is deleted; otherwise, it means that the environmental object corresponding to the spatial point cluster is a dynamic node in the dynamic complex scene topology graph, and the minimum bounding box of the spatial point cluster is calculated, the length, width and height of the minimum bounding box are extracted, and the key nodes, the node paths between passable key nodes and the additional attributes of the node paths in the dynamic complex scene topology graph are updated.

[0042] The first geometric center coordinates are converted to the second geometric center coordinates using the pose estimation transformation matrix. The second geometric center coordinates are the coordinates in the world coordinate system where the first key node is located. The minimum distance between the second geometric center coordinates and the static node position coordinates is calculated.

[0043] Specifically, the GPS / IMU pose estimation transformation matrix is ​​represented as follows: : ; in, , This represents a 1x3 vector of zeros. , This represents a 3x3 attitude matrix. This represents the position coordinates of the flatbed vehicle in the world coordinate system provided by the GPS positioning device, where T represents transpose; The formula for transforming the coordinates of the geometric center to the coordinates of the second geometric center in the world coordinate system is: ; in, Represents the coordinates of the geometric center. This represents the coordinates of the geometric center in the world coordinate system.

[0044] B: Perform a two-layer dynamic evaluation on the task to be executed. The two-layer dynamic evaluation includes dynamic scenario complexity evaluation and dynamic priority evaluation to obtain the dynamic scenario complexity and dynamic priority of the task to be executed.

[0045] Specifically, the list of tasks to be executed records the task location, task generation time, and expected completion time of each task. Based on the dynamic complex scene topology map and the list of tasks to be executed, the complexity of the complex dynamic scene is evaluated for the tasks in the list, including: extracting the tasks to be executed from the list, and obtaining the location, generation time, and expected completion time of each task. The scene complexity of the location of each task and the scene complexity of the flatbed truck closest to that location are calculated. ; .in, Indicates the location of the task to be performed. The complexity of the scene, Representing the topology of a dynamic and complex scene by task location Centered on, with radius The number of static nodes in the circular region. Representing the topology of a dynamic and complex scene by task location Centered on, with radius The number of dynamic nodes not on the flatbed vehicle within the circular region. This represents the dynamic node weight factor. This indicates the location of the flatbed truck closest to the task's destination. Representing the topology of a dynamic and complex scene using the flatbed truck's position Centered on, with radius The number of static nodes in the circular region. Representing the topology of a dynamic and complex scene using the flatbed truck's position Centered on, with radius The number of dynamic nodes in the circular region; Indicates distance from mission location The dynamic scene complexity of the nearest flatbed truck is calculated by weighting the scene complexity of the location where the task is to be performed and the scene complexity of the flatbed truck closest to the task location.

[0046] Specifically, the mission location The dynamic scene complexity is ,in Represents the scene complexity weights. .

[0047] Unlike existing methods that assess environmental complexity solely based on static features of the task location (such as path length and map labels), this method fuses the obstacle density (including both static and dynamic obstacles) around the task location with the obstacle density around the nearest flatbed truck. This design enables the system to not only evaluate the difficulty of the task itself but also dynamically determine the complexity of resource mobilization required to execute the task, achieving real-time dual-point perception of task space complexity. By introducing dynamic obstacle density, the scene complexity becomes adaptable to changes in highly dynamic traffic environments, improving the model's robustness to unexpected scenarios.

[0048] Specifically, the tasks in the task list are dynamically prioritized, including: calculating the urgency index of the tasks based on the current time, the expected completion time of the tasks, and the task generation time. Based on the current time and the time the task to be executed was generated, the waiting time index of the task to be executed is calculated. The urgency index of the task to be performed is obtained by using a nonlinear function. and waiting time index The mapping process is performed to obtain the urgency factor and waiting penalty term of the task to be executed, and combined with the dynamic scene complexity of the task location, the dynamic priority of the task to be executed is generated: ; ; ;in, Indicates the dynamic priority of tasks to be executed. Factors indicating the urgency of tasks to be performed. This indicates the pending penalty items for tasks to be performed. Indicates the location of the task to be performed. Indicates the location of the mission The complexity of dynamic and complex scenarios, This represents the Sigmoid function, which... As a scenario factor for the task to be performed, Indicates the urgency coefficient. This represents the waiting penalty coefficient. Represents the scene coefficient. This represents the nonlinear control coefficient. In this embodiment, it is set... It is 0.4. It is 0.3. It is 0.2.

[0049] Specifically, ; in, Indicates tasks to be performed The urgency index, Indicates tasks to be performed Task generation time, Indicates tasks to be performed Expected completion time for the task This indicates the current time. The closer the current time is to the expected completion time of the task, the lower the urgency index, and the more urgent the task to be performed. Based on the current time and the task generation time, the tasks to be executed are calculated. Waiting time index , specifically: The task to be performed is carried out using a nonlinear function. Urgency Index and waiting time index Perform mapping processing to obtain the task to be executed. The urgency factor and waiting penalty items, combined with the tasks to be performed. The dynamic scene complexity of the task location is used to generate tasks to be executed. Dynamic priority.

[0050] C: Using the dynamic scenario complexity, dynamic priority, and location information of multiple flatbed trucks as input variables, construct an adaptive auction algorithm, input the input variables into the adaptive auction algorithm, and generate the task allocation result of the flatbed truck cluster.

[0051] An adaptive auction algorithm is used to generate task allocation results for multiple flatbed trucks. This includes: constructing a task reward function for each task to be executed. The task reward function uses the dynamic scenario complexity and dynamic priority of the task as fixed parameters, and the real-time position of the flatbed truck as a variable. The output of the task reward function is the reward generated by different flatbed trucks executing tasks. The reward generated by a flatbed truck executing any task is calculated using the task reward function. The task with the highest current reward is selected, and the flatbed truck submits a bidding request to the selected task. The task selects the flatbed truck with the highest reward from all submitted bidding requests as the executor, and removes the task with the executor from the task list, proceeding to the next round of bidding until the task list is empty.

[0052] Specifically, if a task is successfully bid on by multiple flatbed trucks, or if a flatbed truck is selected as the executor by multiple tasks, the following mechanisms resolve conflicts: First, flatbed trucks prioritize bidding on the task with the highest potential profit, and the rest are released. If tasks with the same profit exist, the flatbed truck bids on the task with the closest location. Second, the released tasks re-enter the bidding round. Third, the task selects the flatbed truck with the highest potential profit from all submitted bids. If the highest potential profit is the same for both flatbed trucks, the task selects the flatbed truck with the highest battery level.

[0053] Specifically, the expression for the task reward function includes: ; ; ;in, This represents the revenue generated by the nth flatbed truck performing the task to be performed. This indicates the normalized dynamic priority of the tasks to be executed. This represents the normalized dynamic scenario complexity of the task to be executed. This represents the interactive penalty items for the task to be performed. This represents the normalized scheduling cost of dispatching the nth flatbed truck to the location of the task to be performed. N represents the total number of flatbed trucks, and in this embodiment, the total number of flatbed trucks is greater than the number of tasks to be executed. Indicates the position of the nth flatbed truck With tasks to be performed Mission location The shortest path distance between them This indicates the maximum preset path distance.

[0054] Specifically, Indicates the position of the nth flatbed truck With tasks to be performed Mission location The shortest path distance between them This represents the maximum preset path distance; specifically, the shortest path distance is obtained by searching the topology graph of a dynamic and complex scene using the A* search algorithm. , Indicates the preset maximum dynamic priority. , This indicates the maximum dynamic scene complexity preset.

[0055] The adaptive auction algorithm, by improving upon traditional auction mechanisms, addresses the challenge of collaborative task allocation across multiple flatbed trucks in dynamic and complex scenarios. It systematically optimizes the traditional auction mechanism by introducing a reward function model that integrates multiple factors such as dynamic priority, scenario complexity, and task distance. This model, combined with normalization and an interactive penalty mechanism, makes task allocation more intelligent, stable, and environmentally adaptable. The interactive penalty mechanism addresses the scheduling inefficiency of "high-complexity, low-priority" tasks. By establishing a coupling relationship between complexity and priority, it dynamically suppresses the risk of resource waste. This mechanism effectively solves the problems of "repeated failures of difficult tasks" and "resource consumption by low-value tasks" that may occur in traditional auction mechanisms. The improved algorithm not only possesses higher scheduling rationality and task matching accuracy but also enhances robustness and flatbed truck scheduling efficiency in dynamic and complex environments.

[0056] D: The flatbed truck cluster generates a collaborative planning path based on the real-time updated dynamic complex scene topology map and task allocation results, and converts the collaborative planning path into a control command sequence, which is used to collaboratively control the flatbed truck cluster.

[0057] Specifically, the flatbed truck cluster generates collaborative planning paths based on a dynamic and complex scene topology map and task allocation results. This includes: using a heuristic search algorithm to generate M candidate planning paths for each flatbed truck in the dynamic and complex scene topology map. Each candidate planning path starts at the current position of the flatbed truck and ends at the location of the task to be executed by the flatbed truck. The conflict factor between candidate planning paths of different flatbed trucks is calculated. The candidate planning path combinations of N flatbed trucks are enumerated, and the global path conflict score of each combination is calculated. Each candidate planning path combination consists of one set of candidate planning paths from N flatbed trucks, and the global path conflict score is the sum of the conflict factors of any two sets of candidate planning path combinations. The candidate planning path combination with the lowest global path conflict score is selected as the collaborative planning path combination for the N flatbed trucks. ,in This represents the collaborative planning path for the nth flatbed truck. N and M are positive integers.

[0058] Furthermore, a heuristic search algorithm is used to generate M candidate planning paths for each flatbed truck in a dynamic and complex scene topology map. Each candidate planning path starts at the current position of the flatbed truck and ends at the location of the task to be performed by the flatbed truck. The heuristic search algorithm is the A* algorithm. The conflict factor between candidate planning paths of different flatbed trucks is calculated, where the conflict factor between the m-th candidate planning path of the n-th flatbed truck and the g-th candidate planning path of the q-th flatbed truck is: ; in, This represents the m-th candidate planning path for the nth flatbed truck. The g-th candidate path planning group for the q-th flatbed truck Conflict factors between them , , , N represents the total number of flatbed trucks. Indicate candidate planning paths The set of static nodes visited Indicate candidate planning paths The set of static nodes visited This indicates the number of nodes in the set. Indicate candidate planning paths Path length, Indicate candidate planning paths The path length.

[0059] In summary, firstly, multi-source sensor collaborative perception: combining LiDAR (acquiring environmental point clouds), IMU pose sensor (acquiring Euler angles of the flatbed vehicle), GPS positioning device (acquiring world coordinates), and communication module (synchronizing data) to form an integrated "perception-positioning-data synchronization" system—LiDAR fits the ground through the RANSAC algorithm and extracts environmental objects (such as dynamic obstacles) through Euclidean clustering, while IMU and GPS construct a "pose estimation transformation matrix" to map local point cloud coordinates to the world coordinate system, solving the problems of "incomplete perception" (such as GPS being unable to identify obstacles) or "insufficient accuracy" (such as IMU drift when used alone) of a single sensor.

[0060] Second, real-time iteration of the topology map: Based on multi-source sensor data, the "dynamic nodes" (such as irregularly moving obstacles) and "node path attributes" (such as whether the path width meets the passage requirements) of the topology map are updated in real time. This enables the topology map to reflect the dynamic scene in real time, solving the core pain point of planning lag in dynamic scenes and providing real-time environmental data for subsequent task allocation and path planning.

[0061] Third, a triple evaluation factor is constructed: This integrates three independent dimensions—"urgency (remaining task time), waiting time (task starvation risk), and scenario complexity (task execution difficulty)"—and quantifies priority through three fixed weights, rather than a single indicator. For example, "sudden accident test tasks" (high urgency), "long-unexecuted severe weather tasks" (long waiting time), and "multi-vehicle convergence test tasks" (high complexity) can all obtain reasonable priorities through corresponding factors, preventing any dimension from being completely ignored. This breaks through the limitations of "single-dimensional ranking," making priority evaluation align with the three requirements of "timeliness, fairness, and difficulty" in dynamic scenarios. Specifically, it shortens the response time for urgent tasks, improves the task completion rate, and increases the probability of resource acquisition for high-complexity tasks, solving the core problem of traditional scheduling being inefficient in addressing all aspects.

[0062] Fourth, pre-defined conflict resolution rules include strategies for resolving three types of conflicts: First, when multiple vehicles bid for the same task, the vehicle with the highest benefit and closest proximity is selected. Second, when a vehicle is selected by multiple tasks, the vehicle with the highest benefit and highest task priority is selected. Then, the released task / vehicle re-enters the next round of bidding to avoid disorderly competition after a conflict. This achieves autonomous optimal matching of "task-vehicle": shortened scheduling distance, reduced vehicle energy consumption, reduced conflict resolution time, and 100% task allocation success rate, solving the core problems of "resource waste and conflict deadlock" in existing allocation methods.

[0063] Fifth, redundant options (such as multiple candidate solutions) are introduced into the system to address uncertainty and improve optimization space. The A* algorithm generates M candidate paths (e.g., M=3) for each flatbed truck, instead of a single path. Redundant paths ensure that if an obstacle exists on one path, other options remain. Simultaneously, a conflict factor E (integrating "node overlap" and "path length difference") is defined. All path combinations for all flatbed trucks are enumerated, and a "global conflict score" (the sum of all pairwise conflict factors) is calculated. The combination with the lowest score is selected, ensuring no spatial / temporal conflicts among multiple vehicles. This achieves global collaborative collision avoidance for multiple vehicles: the collision risk is reduced to zero when multiple vehicles are in parallel, improving traffic efficiency and solving the core problem of potential global congestion in existing path planning. It also meets the high requirements of "multi-vehicle interaction simulation" in autonomous driving testing.

[0064] As can be seen, this embodiment simultaneously solves the problems of dynamic environment adaptation, unreasonable task evaluation, resource mismatch and conflict, and global path congestion.

[0065] like Figure 1 As shown, in one implementation, a dynamic complex scene topology graph is provided, which can implement the aforementioned multi-board collaborative planning method for complex scenes based on dynamic priorities.

[0066] There are pathways between the first critical node and passable critical nodes. If there are static and dynamic objects along the path, causing the path width to be less than the flatbed truck's width, the path becomes impassable. There are static objects between the task location and intersection 2, and dynamic objects between corner 1 and corner 3 at time t. There are also two paths between intersection 1 and intersection 2: one through corner 1, but a dynamic object is detected at time t; the other involves corner 4 and the flatbed truck. A multi-flatbed collaborative planning method based on dynamic priority is adopted for complex scenarios. Due to the presence of static and dynamic objects along the path, the flatbed truck can choose to go from corner 4 to intersection 1 and then to the task location.

[0067] Beneficial effects of this embodiment First, by combining dynamic environmental perception and topology map updates, precise perception and real-time adaptation are achieved, solving the problem of static topology maps being unsuitable for complex scenarios. Firstly, high environmental perception accuracy avoids obstacle misjudgment. A multi-source perception scheme of "LiDAR, IMU pose sensor, and GPS positioning" is adopted: After LiDAR generates point cloud data, the ground plane is fitted using the RANSAC algorithm (precisely removing ground points to avoid misjudging road surfaces as obstacles). Then, Euclidean clustering is used to divide the remaining point cloud into spatial point clusters (such as static obstacles and dynamic pedestrians / vehicles), ensuring accurate environmental object recognition. Simultaneously, the attitude matrix is ​​calculated by obtaining Euler angles through the IMU, and a "GPS / IMU pose transformation matrix" is constructed by combining GPS position, mapping the local geometric center coordinates of the point clusters to the world coordinate system, with errors controlled at the centimeter level, providing accurate node data for the topology map. Second, the topology map is updated in real time and adapts to dynamic obstacles. It clearly defines a classification mechanism that combines "static nodes (intersections, corners, etc.)" and "dynamic nodes (irregularly moving objects)". Environmental data is uploaded in real time through the flatbed truck's sensors, and node attributes are dynamically updated. For example, if a dynamic object (such as a suddenly appearing test obstacle) is detected on the path "corner 1-corner 3" at time t, the topology map will immediately mark the "accessibility" of the path and update the additional attributes of the node path (such as whether the path width meets the requirements for the flatbed truck to pass). If the width of a certain road surface is lower than the width of the flatbed truck (such as the path "task location-intersection 2" in the embodiment being narrow due to static object obstruction), the path is directly excluded to avoid the flatbed truck planning an "infeasible path", reduce the risk of collision, and significantly reduce the response delay of dynamic obstacles in actual tests, meeting the needs of real-time operations.

[0068] Secondly, a two-layer dynamic evaluation: This achieves scientific quantification of "task difficulty and priority," solving the problems of "single sorting and resource misallocation" in traditional scheduling. First, dynamic scene complexity: This balances "task difficulty + scheduling difficulty," avoiding blind allocation. Unlike existing technologies that only evaluate "static characteristics of the task location," Example 1 simultaneously calculates the scene complexity of the task location and the scene complexity surrounding the nearest flatbed truck, and then weights them together using a β-weighted fusion. This calculation method reflects both the task execution difficulty (e.g., if there are many obstacles around the task to be executed, the execution test requires more precise control) and the resource scheduling difficulty (e.g., if there are many obstacles around the flatbed truck to be executed, scheduling to the target will take longer), avoiding assigning complex tasks to flatbed trucks that are already congested, reducing resource waste, and improving the rationality of task allocation in actual tests. Second, dynamic priority: A three-factor balance including timeliness, fairness, and complexity avoids task delays and starvation by introducing urgency factors + waiting penalty items + scene factors, and quantifies the weights using fixed coefficients (ω1=0.4, ω2=0.3, ω3=0.2). The urgency factor is applied to time-sensitive tasks (such as test tasks simulating "sudden traffic accidents"). When the remaining time for a task is low, its priority is significantly increased to ensure that urgent tasks are assigned first. Actual testing shows that this shortens the response time for urgent tasks. The waiting penalty is applied to edge tasks that have been idle for a long time (such as test tasks simulating low-speed driving in severe weather). If a task has been waiting for a considerable period, its priority is moderately increased to avoid task starvation and improve task completion rate. The scenario factor is applied to highly complex tasks. Maximizing the scenario factor ensures that complex tasks are not ignored indefinitely due to their high execution difficulty. For example, in test tasks simulating "multi-vehicle convergence," the probability of resource acquisition is increased.

[0069] Meanwhile, the adaptive auction algorithm achieves optimal matching between tasks and flatbed carts, resolving resource waste and allocation conflicts. First, the reward function accurately quantifies the fit, avoiding resource mismatch. Each parameter corresponds to actual needs: scheduling cost – if flatbed cart A is closer to the task location, the scheduling cost is low and the reward is high; if flatbed cart B is farther from the task, the reward is low. The algorithm prioritizes flatbed cart A for bidding, reducing unnecessary travel. Actual testing shows a shorter scheduling distance and lower energy consumption for the flatbed carts. Interaction penalty – if a task is "low priority + high complexity," the penalty is low, preventing excessive exclusion due to high difficulty and low priority; if a task is low priority + low complexity, the penalty is high, preventing flatbed carts from preempting low-value tasks and improving resource utilization. Second, conflict resolution rules ensure orderly allocation and avoid deadlocks caused by multiple carts bidding for the same task or a cart being selected by multiple tasks. Flatbed carts prioritize bidding for the task with the highest reward; if rewards are the same, the task closest to the cart is selected. Tasks prioritize the flatbed cart with the highest reward; if rewards are the same, the flatbed cart with the highest battery capacity is selected. Released tasks / flatbeds re-enter the next round of bidding, ensuring no allocation deadlock occurs, shortening conflict resolution time, and improving task allocation completion rate.

[0070] Finally, collaborative path planning enables conflict-free multi-vehicle driving, solving the problems of local optima and global congestion inherent in traditional path planning. First, multiple candidate paths cover the optimal probability, avoiding path homogenization. An A* heuristic search algorithm generates M candidate paths for each flatbed truck, with the starting point being the flatbed truck's current location and the destination being the task location. Multiple candidate paths provide selection space for subsequent conflict assessment, preventing planning failure due to the infeasibility of a single path. Second, a global conflict score filters for the "optimal combination," avoiding multi-vehicle collisions. A conflict factor is defined, quantifying the spatial conflict (node ​​overlap) and temporal conflict (length difference; similar lengths may lead to simultaneous arrival at a node) between two flatbed truck paths: ensuring no spatial / temporal conflicts during multi-vehicle driving. Real-world testing shows that the collision risk is reduced to zero when multiple vehicles are driving in parallel, and traffic efficiency is improved, meeting the requirements of "multi-traffic participant collaborative simulation" in autonomous driving testing.

[0071] Example 2 Unlike the previous embodiments, this embodiment provides a multi-panel collaborative planning system for dynamic and complex scenarios based on dynamic priorities, such as... Figure 2 As shown, a multi-platform collaborative planning method for complex scenarios based on dynamic priority includes a data analysis module, which includes an electronic device. The electronic device includes a memory, a processor 101, and a computer program 1041 stored in the memory and executable on the processor. When the processor 101 executes the computer program, it enables the electronic device to implement a multi-platform collaborative planning method for complex scenarios based on dynamic priority, as described in any of the above embodiments.

[0072] Specifically, such as Figure 3 As shown, the electronic device may include: one or more processors 101, one or more input devices 102, one or more output devices 103, one or more memories 104, and a computer program stored in the memory 104 and executable on the processor. The processor 101, input devices 102, output devices 103, and memory 104 are interconnected via a bus 105. The memory 104 stores the computer program 1041, which includes program instructions. The processor 101 is configured to invoke the program instructions to execute the method steps described in the above method embodiments.

[0073] It should be understood that, in this embodiment, the processor 101 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0074] Input device 102 may include a keyboard, etc., and output device 103 may include a display (LCD, etc.), a speaker, etc.

[0075] The memory 104 may include read-only memory and random access memory, and provides instructions and data to the processor 101. A portion of the memory 104 may also include non-volatile random access memory. For example, the memory 104 may also store device type information.

[0076] In specific implementations, the processor 101, input device 102, and output device 103 described in the embodiments of the present invention can execute the implementation methods described in the relevant embodiments of the multi-platform collaborative planning method and system based on dynamic priority in complex scenarios provided by the embodiments of the present invention, which will not be repeated here.

[0077] It should be noted that for a more detailed description of the electronic device's workflow and the implementation of a multi-tablet collaborative planning method in complex scenarios based on dynamic priorities, please refer to the aforementioned method implementation section, which will not be repeated here.

[0078] Example 3 Unlike the previous embodiments, the memory described in this embodiment should be interpreted broadly. It can be not only a hardware component in a computer system used for temporary data storage, but also a physical medium capable of storing digital information and being read by a computer. These media can be permanent or temporary, including but not limited to hard disks and solid-state drives.

[0079] Specifically, the memory can be an internal storage unit of the electronic device described in any of the embodiments, such as a system hard drive or memory. The memory can also be an external storage device of the system, such as a plug-in hard drive, SmartMediaCard (SMC), Secure Digital (SD) card, FlashCard, etc., equipped on the system. Furthermore, the memory can include both internal storage units and external storage devices. The memory is used to store the computer program and other programs and data required by the system. The memory can also be used to temporarily store data that has been output or will be output.

[0080] Storage devices include: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media that can store program code.

[0081] Numerous specific details are set forth in this specification. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, systems, and techniques have not been shown in detail so as not to obscure the understanding of this specification. In the description of this specification, references to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., mean that a specific feature, method, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this specification.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A multi-plank collaborative planning method for complex scenarios based on dynamic priority, characterized in that, include: A: Construct an undirected topology graph of a dynamic and complex scene, and plan a list of tasks to be executed for the flatbed truck cluster. Obtain the environmental and location information of each flatbed truck, and update the dynamic and complex scene topology graph in real time based on the environmental information. B: Perform a two-layer dynamic evaluation on the task to be executed. The two-layer dynamic evaluation includes dynamic scenario complexity evaluation and dynamic priority evaluation, to obtain the dynamic scenario complexity and dynamic priority of the task to be executed. C: Using the dynamic scenario complexity, dynamic priority, and location information of multiple flatbed trucks as input variables, construct an adaptive auction algorithm, input the input variables into the adaptive auction algorithm, and generate the task allocation result of the flatbed truck cluster. D: The flatbed truck cluster generates a collaborative planning path based on the real-time updated dynamic complex scene topology map and task allocation results, and converts the collaborative planning path into a control command sequence, which is used to collaboratively control the flatbed truck cluster.

2. The multi-plank collaborative planning method for complex scenarios based on dynamic priority as described in claim 1, characterized in that, Obtain the first key node, the node paths between passable key nodes, and the additional attributes of the node paths in a dynamic and complex scene, and construct an undirected topology graph of the dynamic and complex scene. The first key node types in the dynamic complex scene include intersections, corners, task locations, static objects, and dynamic objects; The first key node includes static nodes and dynamic nodes; The types of static nodes include intersections, corners, task locations, and static objects; The flatbed vehicle senses environmental information of dynamic and complex scenes and updates dynamic nodes in real time. The type of dynamic node is a dynamic object, which is characterized as an object that moves irregularly in a dynamic and complex scene. The additional attributes of the node path include path length and path width; The types of passable key nodes include intersections, corners, and mission locations.

3. The multi-plank collaborative planning method for complex scenarios based on dynamic priority as described in claim 2, characterized in that, By installing multi-source sensors on a single flatbed truck, the environmental and location information of the flatbed truck is perceived in real time, and the dynamic complex scene topology map is updated in real time based on the environmental information, including: The multi-source sensor includes a lidar, a pose sensor, a positioning device, and a communication module; The laser pulses are emitted around the flatbed vehicle by a lidar, and environmental point cloud data of spatial points around the flatbed vehicle is generated based on the reflected laser signals. The environmental point cloud data includes several spatial points. The RANSAC method was used to fit the ground plane, and spatial points belonging to the ground were removed. The remaining spatial points were then clustered using Euclidean clustering to obtain multiple spatial point clusters, which were then used as environmental objects around the flatbed truck. The positioning device is used to obtain the location information of the flatbed vehicle. The communication module is used to analyze the spatial point cluster and the location information of the flatbed vehicle to calculate the coordinates of the first geometric center of the spatial point cluster. The pose sensor is used to sense the Euler angles of the flatbed truck. The pose matrix of the flatbed truck is calculated by using the Euler angles. Combined with the position coordinates of the flatbed truck, the pose estimation transformation matrix is ​​constructed. The first geometric center coordinates are converted to the second geometric center coordinates using the pose estimation transformation matrix. The second geometric center coordinates are the coordinates in the world coordinate system where the first key node is located. The minimum distance between the second geometric center coordinates and the static node position coordinates is calculated.

4. The multi-plank collaborative planning method for complex scenarios based on dynamic priority as described in claim 1, characterized in that, Extract the tasks to be executed from the list of tasks to be executed, along with the location, generation time, and expected completion time of each task. Calculate the scene complexity of the location where the task to be performed is to be executed, and the scene complexity of the flatbed truck closest to the location: ； ； in, Indicates the location of the task to be performed. The complexity of the scene, Representing the topology of a dynamic and complex scene by task location Centered on, with radius The number of static nodes in the circular region. Representing the topology of a dynamic and complex scene by task location Centered on, with radius The number of dynamic nodes not on the flatbed vehicle within the circular region. Indicates the dynamic node weight factor; This indicates the location of the flatbed truck closest to the task's destination. Representing the topology of a dynamic and complex scene using the flatbed truck's position Centered on, with radius The number of static nodes in the circular region; Representing the topology of a dynamic and complex scene using the flatbed truck's position Centered on, with radius The number of dynamic nodes in the circular region; Indicates distance from mission location The dynamic scene complexity of recent flatbed trucks; The dynamic scene complexity of the task location is obtained by weighting the scene complexity of the location to be executed and the scene complexity of the flatbed truck closest to the task location.

5. The multi-plank collaborative planning method for complex scenarios based on dynamic priority as described in claim 1, characterized in that, Based on the current time, the expected completion time of the task to be executed, and the task generation time, the urgency index of the task to be executed is calculated. ; Based on the current time and the time the task to be executed was generated, the waiting time index of the task to be executed is calculated. ; An urgency index for the task to be performed using a non-linear function. and waiting time index The mapping process is performed to obtain the urgency factor and waiting penalty term of the task to be executed, and combined with the dynamic scene complexity of the task location, the dynamic priority of the task to be executed is generated: ； ； ； in, Indicates the dynamic priority of tasks to be executed. Indicates the urgency factor of the task to be performed. This indicates the pending penalty items for tasks to be performed. Indicates the location of the task to be performed. Indicates the location of the mission The complexity of dynamic and complex scenarios, This represents the Sigmoid function, which... As a scenario factor for the task to be performed, Indicates the urgency coefficient. This represents the waiting penalty coefficient. Represents the scene coefficient. This represents the nonlinear control coefficient.

6. The multi-plank collaborative planning method for complex scenarios based on dynamic priority as described in claim 1, characterized in that, Construct a task revenue function for the task to be executed. The task revenue function takes the dynamic scenario complexity and dynamic priority of the task to be executed as fixed parameters and the real-time position of the flatbed truck as a variable. The output of the task revenue function is the revenue generated by different flatbed trucks executing the task. The revenue generated by the flatbed truck in executing any task is calculated using the task revenue function, and the task with the highest current revenue is selected. The flatbed truck then submits a bidding request to the selected task. The task to be executed selects the flatbed truck with the highest profit from all flatbed trucks that have submitted bidding requests as the executor, and removes the task to be executed with an executor from the task to be executed list, and proceeds to the next round of bidding, until the task to be executed list is empty.

7. The multi-plank collaborative planning method for complex scenarios based on dynamic priority as described in claim 6, characterized in that, The expression for the task reward function is: ； ； ； in, This represents the revenue generated by the nth flatbed truck performing the task to be performed. This indicates the normalized dynamic priority of the tasks to be executed. This represents the normalized dynamic scenario complexity of the task to be executed. This represents the interactive penalty items for the task to be performed. This represents the normalized scheduling cost of dispatching the nth flatbed truck to the location of the task to be performed. N represents the total number of flatbed trucks; Indicates the position of the nth flatbed truck With tasks to be performed Mission location The shortest path distance between them This indicates the maximum preset path distance.

8. The multi-plank collaborative planning method for complex scenarios based on dynamic priority as described in claim 1, characterized in that, Multiple candidate planning paths for each flatbed truck are generated in a dynamic and complex scene topology map using a heuristic search algorithm. The candidate planning paths start from the current position of the flatbed truck and end at the task location of the task to be performed by the flatbed truck. Calculate the conflict factor between candidate planned paths for different flatbed trucks; Enumerate candidate planning path combinations consisting of at least one flatbed truck, and calculate the global path conflict score for each candidate planning path combination. The candidate planning path combination consists of a group of candidate planning paths consisting of at least one flatbed truck. The global path conflict score is the sum of the conflict factors of any two candidate planning path combinations in the candidate planning path combination. The candidate path combination with the lowest global path conflict score is selected as the collaborative path combination for the N flatbed trucks: ,in This represents the collaborative planning path for the nth flatbed truck. N is the number of flatbed trucks.

9. A multi-plank collaborative planning method for complex scenarios based on dynamic priority, as described in claim 8, is characterized in that... The conflict factors include: ; in, This represents the m-th candidate planning path for the nth flatbed truck. The g-th candidate path planning group for the q-th flatbed truck Conflict factors between them , , , N represents the total number of flatbed trucks. Indicate candidate planning paths The set of static nodes visited Indicate candidate planning paths The set of static nodes visited This indicates the number of nodes in the set. Indicate candidate planning paths Path length, Indicate candidate planning paths The path length.

10. A multi-plank collaborative planning method for complex scenarios based on dynamic priority, as described in claim 5, is characterized in that... It is 0.

4. It is 0.

3. It is 0.2.

Images