Modularized unmanned practical training vehicle control method and system under ROS architecture

Abstract

The invention belongs to the technical field of unmanned vehicles, and discloses a modular unmanned practical training vehicle control method and system under an ROS architecture, and the method comprises the steps: obtaining multi-source object data, and constructing a semantic structured cognitive configuration of a scene relation; constructing a path task decision diagram in combination with a path node task target; analyzing and optimizing the path task decision diagram by utilizing a multi-dimensional integrated constraint condition to obtain a priority constraint control strategy; obtaining an actual path execution result, and inputting a priority constraint control strategy for analysis and correction to obtain feedback information of a dynamic adjustment path; carrying out data integration processing by utilizing a modular processing framework, and generating a running state parameter set for controlling execution; and generating a control instruction, and driving the unmanned practical training vehicle to navigate and run according to the optimized path in real time. According to the invention, the problem of path execution under different tasks and scenes is solved through a modularized processing framework, and different traffic environments, road conditions and vehicle performance changes can be coped with.

Description

Technical Field

[0001] This invention relates to the field of autonomous vehicle technology, and in particular to a control method and system for a modular autonomous driving training vehicle under the ROS architecture. Background Technology

[0002] With the continuous advancement of intelligent driving technology, unmanned training vehicles play a crucial role in the research, teaching, and algorithm verification of intelligent connected vehicles. These vehicles integrate multiple sensors such as LiDAR, millimeter-wave radar, cameras, and GPS, and utilize an advanced Robot Operating System (ROS) architecture for modular design, enabling real-time perception of the surrounding environment, path planning, and control execution. The ROS architecture provides the system with flexible modular division and data communication mechanisms, effectively reducing the complexity of system development and integration, and improving the overall system scalability.

[0003] However, existing ROS-based autonomous driving control systems typically rely on traditional geometric calculations for accuracy and adaptability in path planning and execution. Existing autonomous driving training vehicle control methods focus on distance or time optimization, but fail to deeply consider multi-level information such as traffic rules, dynamic obstacles, and environmental perception uncertainties. This can lead to path selection failing to adapt to real-time environmental changes during execution, or even execution errors or non-compliance.

[0004] Furthermore, existing path planning systems lack comprehensive integration and optimization of multi-dimensional constraints, such as traffic speed limits, turning radii, and vehicle dynamics. Although path planning systems can calculate the shortest or fastest path, they often fail to effectively incorporate these constraints when dealing with real-world traffic conditions and complex road situations, thus affecting path safety and execution reliability.

[0005] Furthermore, existing systems generally lack effective closed-loop feedback control mechanisms. During path execution, real-time monitoring and dynamic adjustment of vehicle status are often insufficient, leading to a failure to promptly correct paths or update control strategies when encountering road obstacles, traffic flow changes, or unexpected situations. This limitation results in insufficient robustness and flexibility of autonomous driving systems in dynamic environments, impacting stability and safety during long-term operation. Under the ROS architecture, existing systems face challenges in data flow integration and processing capabilities. The collaboration between different modules is relatively weak, making it difficult to fully utilize the interactive feedback of sensor data, path planning decisions, and vehicle execution information. This leads to low overall system efficiency, affecting the real-time performance and execution accuracy of autonomous driving tasks.

[0006] Therefore, how to provide a modular unmanned training vehicle control method and system under the ROS architecture is an urgent problem to be solved. Summary of the Invention

[0007] This invention provides a modular unmanned training vehicle control method and system under the ROS architecture to solve the problems mentioned above in the prior art.

[0008] According to a first aspect of the present invention, a control method for a modular unmanned training vehicle under the ROS architecture is provided.

[0009] In one embodiment, the control method for a modular unmanned training vehicle under a ROS architecture includes: Multi-source object data is acquired from multi-source sensors in an unmanned training vehicle, and a semantic structured cognitive configuration of scene relationships is constructed based on the multi-source object data. Based on semantic structured cognitive configuration and path node task objectives, a path task decision graph is constructed; and multi-dimensional integrated constraints are used to analyze and optimize the path task decision graph to obtain a priority constraint control strategy. Based on multi-source object data, the actual path execution results are obtained, and the actual path execution results are input into the priority constraint control strategy for analysis and correction, so as to obtain feedback information for dynamically adjusting the path. A modular processing framework is used to integrate and process semantic structured cognitive configurations, priority constraint control strategies, and feedback information to generate a set of operational state parameters for control execution. Control commands are then generated based on the set of operational state parameters to drive the unmanned training vehicle to perform real-time navigation and path driving according to the optimized path.

[0010] In one embodiment, acquiring multi-source object data based on multi-source sensors in an autonomous driving training vehicle and constructing a semantically structured cognitive configuration of scene relationships based on the multi-source object data includes: Raw point clouds, images, and radar signals are collected using lidar, cameras, and millimeter-wave radar, and time synchronization, coordinate unification, and noise reduction are performed to obtain standardized multi-source sensor data. Based on standardized multi-source sensor data, deep learning models and data association algorithms are used to perform image target recognition and fuse the target recognition results to generate a fused target list. The spatial topological and traffic logic relationships between targets in the fusion target list are analyzed, and a dynamic scene semantic network is constructed. The dynamic scene semantic network is transformed into a graph structure, with entities as attribute nodes and spatiotemporal and logical relationships as connecting edges, to obtain a semantically structured cognitive configuration of scene relationships.

[0011] In one embodiment, the multidimensional integration constraint includes: Constraints include traffic rules, vehicle dynamics, driving safety, environmental perception uncertainty, and task execution timeliness.

[0012] In one embodiment, a path task decision graph is constructed based on a semantically structured cognitive configuration and the task objectives of path nodes; and the path task decision graph is parsingly optimized using multi-dimensional integrated constraints to obtain priority constraint control strategies, including: Based on high-precision maps and global tasks, path planning is performed to generate path node task objectives. Then, using graph theory, the path node task objectives and semantic structured cognitive configurations are deduced and analyzed to construct a path task decision graph. By using multidimensional integrated constraints, the path task decision graph is parsed and processed to remove nodes and branches that do not meet the hard constraints, resulting in a set of feasible paths that meet the constraints. By defining a comprehensive cost function, the set of feasible paths that meet the constraints is analyzed and evaluated to obtain the optimal execution path; and a priority constraint control strategy is constructed based on the optimal execution path.

[0013] In one embodiment, by defining a comprehensive cost function, the set of feasible paths that meet the constraints is analyzed and evaluated to obtain the optimal execution path; and the priority constraint control strategy constructed based on the optimal execution path further includes: Based on path efficiency, driving safety, and energy economy, a comprehensive evaluation system is generated. Using the weight adaptive mechanism and the comprehensive evaluation system, the weight coefficients of each dimension of the scene features in the semantic structured cognitive configuration are dynamically adjusted. Based on the comprehensive evaluation results generated by dynamic weight coefficients, and combined with the priority constraint control strategy, real-time path optimization and control command generation are executed to drive the vehicle along the optimal path, while dynamic path correction is implemented to cope with real-time environmental changes. A closed-loop feedback control mechanism is used to monitor the vehicle status in real time and adjust the control strategy to maintain the optimality and safety of the entire path tracking.

[0014] In one embodiment, based on multi-source object data, the actual path execution result is obtained, and the actual path execution result is input into the priority constraint control strategy for analysis and correction, resulting in feedback information for dynamically adjusting the path, including: Based on multi-source object data, the current position, speed and path deviation data of the vehicle are obtained in real time, and the path tracking algorithm is used to calculate the multi-source object data to obtain the actual path execution result. The actual path execution results are input into the priority constraint control strategy to analyze and evaluate path errors and real-time environmental data, and the feasibility of the path is analyzed using path planning algorithms to obtain feedback information for dynamic path adjustment.

[0015] In one embodiment, a modular processing framework is used to integrate and process semantically structured cognitive configurations, priority constraint control strategies, and feedback information to generate a set of operational state parameters for control execution. Control commands are then generated based on this set of operational state parameters to drive the autonomous driving training vehicle through real-time navigation and path travel along an optimized path. A modular processing framework is generated by combining path planning, environmental perception, and priority control. This modular processing framework is then exchanged through a data interface to obtain a unified modular processing framework. A unified modular processing framework is used to integrate semantic structured cognitive configuration, priority constraint control strategy and feedback information, and the Kalman filter algorithm is used to integrate the integrated results to obtain a set of running state parameters for control execution. The predictive control algorithm is used to calculate the set of operating state parameters for control execution, generate control commands, and drive the unmanned training vehicle to run along the optimized path to ensure that the vehicle performs the task according to the optimal path.

[0016] In one embodiment, a unified modular processing framework is used to integrate semantically structured cognitive configurations, priority constraint control strategies, and feedback information. The Kalman filter algorithm is then used to integrate the integrated results, resulting in a set of operational state parameters for control execution, including: Establish a unified spatiotemporal coordinate system, and perform timestamp synchronization and coordinate transformation on environmental data in semantic structured cognitive configuration, target data in priority constraint control strategy and vehicle status data in feedback information, and output a standardized spatiotemporal aligned data set; The standardized spatiotemporally aligned dataset is input into a Kalman filter, and the noise is processed by real-time estimation and observation noise. The filter gain is dynamically adjusted to output high-precision vehicle dynamic state parameter estimation results. Multi-source data cross-validation and residual analysis are performed on the high-precision vehicle dynamic state parameter estimation results to identify and remove abnormal data, and generate a set of control execution operation state parameters with credibility indicators.

[0017] In one embodiment, the Kalman filter state prediction formula is: ; ; In the formula, This represents the predicted state value at time t. Represents the state transition matrix. This represents the matrix showing the effect of control inputs on the state. Indicates control input, This represents the prediction of state covariance. Represents the process noise covariance; The observation update formula for the Kalman filter is: ; ; ; In the formula, Kalman gain represents the trade-off between prediction and observation. Represents the observation matrix. Represents the observation noise covariance. This represents the current sensor fusion observations. This represents the updated state estimate. This represents the updated state covariance.

[0018] According to a second aspect of the present invention, a modular unmanned training vehicle control system based on ROS architecture is provided, comprising: The multi-source data integration module is used to acquire multi-source object data based on multi-source sensors in the unmanned driving training vehicle, and to construct a semantic structured cognitive configuration of scene relationships based on the multi-source object data. The path task decision module is used to construct a path task decision graph based on semantic structured cognitive configuration and path node task objectives; and to analyze and optimize the path task decision graph using multi-dimensional integrated constraints to obtain priority constraint control strategies. The path execution feedback module is used to obtain the actual path execution results based on multi-source object data, and input the actual path execution results into the priority constraint control strategy for analysis and correction, so as to obtain feedback information for dynamically adjusting the path; The control command generation module is used to integrate and process semantic structured cognitive configuration, priority constraint control strategy and feedback information using a modular processing framework to generate a set of operating state parameters for control execution; and to generate control commands based on the set of operating state parameters to drive the unmanned training vehicle to perform real-time navigation and path driving according to the optimized path.

[0019] According to a third aspect of the present invention, a computer device is provided.

[0020] In some embodiments, the computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0021] According to a fourth aspect of the present invention, a computer-readable storage medium is provided.

[0022] In one embodiment, a computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the above method.

[0023] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: 1) This invention solves the problem of traditional path planning methods ignoring actual constraints by integrating multi-dimensional constraints such as traffic rules, vehicle dynamics, and driving safety, thereby avoiding the risk of path planning failure, vehicle collisions, or violations of traffic rules in complex environments.

[0024] 2) This invention achieves accurate monitoring and dynamic adjustment of the vehicle's real-time status by introducing a closed-loop feedback control mechanism and a Kalman filter algorithm. This enables automatic correction during path execution, maintaining optimal path tracking and ensuring long-term operational stability and safety. Furthermore, the modular processing framework solves path execution problems under different tasks and scenarios, giving the system high flexibility and scalability, easily handling different traffic environments, road conditions, and vehicle performance changes.

[0025] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0026] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0027] Figure 1 This is a flowchart illustrating a modular unmanned training vehicle control method under a ROS architecture, according to an exemplary embodiment. Figure 2 This is a schematic diagram illustrating the principle of a modular unmanned training vehicle control system under the ROS architecture, according to an exemplary embodiment. Figure 3 This is a schematic diagram of the structure of a computer device according to an exemplary embodiment. Detailed Implementation

[0028] The following description and accompanying drawings fully illustrate specific embodiments described herein to enable those skilled in the art to practice them. Some portions and features of certain embodiments may be included in or replace portions and features of other embodiments. The scope of the embodiments herein includes the entire scope of the claims and all available equivalents thereof. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments; similar or identical parts between embodiments can be referred to interchangeably.

[0029] The modules in the apparatus or system of this application can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0030] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0031] Figure 1 An embodiment of the control method and system for a modular unmanned training vehicle under the ROS architecture of the present invention is shown.

[0032] In this optional embodiment, the modular unmanned training vehicle control method under ROS architecture includes: Step S101: Acquire multi-source object data based on multi-source sensors in the unmanned driving training vehicle, and construct a semantic structured cognitive configuration of scene relationships based on the multi-source object data; In this optional embodiment, acquiring multi-source object data based on multi-source sensors in the unmanned training vehicle and constructing a semantically structured cognitive configuration of scene relationships based on the multi-source object data includes: Raw point clouds, images, and radar signals are collected using lidar, cameras, and millimeter-wave radar, and time synchronization, coordinate unification, and noise reduction are performed to obtain standardized multi-source sensor data. Based on standardized multi-source sensor data, deep learning models and data association algorithms are used to perform image target recognition and fuse the target recognition results to generate a fused target list. The spatial topological and traffic logic relationships between targets in the fusion target list are analyzed, and a dynamic scene semantic network is constructed. The dynamic scene semantic network is transformed into a graph structure, with entities as attribute nodes and spatiotemporal and logical relationships as connecting edges, to obtain a semantically structured cognitive configuration of scene relationships.

[0033] Step S102: Based on the semantic structured cognitive configuration and the task objectives of the path nodes, construct a path task decision graph; and use multi-dimensional integrated constraints to analyze and optimize the path task decision graph to obtain a priority constraint control strategy. In this optional embodiment, the multidimensional integration constraints include: Constraints include traffic rules, vehicle dynamics, driving safety, environmental perception uncertainty, and task execution timeliness.

[0034] In this optional embodiment, a path task decision graph is constructed based on the semantically structured cognitive configuration and the task objectives of the path nodes; and the path task decision graph is parsingly optimized using multi-dimensional integrated constraints to obtain priority constraint control strategies, including: Based on high-precision maps and global tasks, path planning is performed to generate path node task objectives. Then, using graph theory, the path node task objectives and semantic structured cognitive configurations are deduced and analyzed to construct a path task decision graph. By using multidimensional integrated constraints, the path task decision graph is parsed and processed to remove nodes and branches that do not meet the hard constraints, resulting in a set of feasible paths that meet the constraints. By defining a comprehensive cost function, the set of feasible paths that meet the constraints is analyzed and evaluated to obtain the optimal execution path; and a priority constraint control strategy is constructed based on the optimal execution path.

[0035] In this optional embodiment, by defining a comprehensive cost function, the set of feasible paths that meet the constraints is analyzed and evaluated to obtain the optimal execution path; and the priority constraint control strategy constructed based on the optimal execution path further includes: Based on path efficiency, driving safety, and energy economy, a comprehensive evaluation system is generated. Using the weight adaptive mechanism and the comprehensive evaluation system, the weight coefficients of each dimension of the scene features in the semantic structured cognitive configuration are dynamically adjusted. Based on the comprehensive evaluation results generated by dynamic weight coefficients, and combined with the priority constraint control strategy, real-time path optimization and control command generation are executed to drive the vehicle along the optimal path, while dynamic path correction is implemented to cope with real-time environmental changes. A closed-loop feedback control mechanism is used to monitor the vehicle status in real time and adjust the control strategy to maintain the optimality and safety of the entire path tracking.

[0036] Specifically, a semantically structured cognitive configuration encompassing traffic participant dynamics, lane topology, and traffic signal status is constructed through multi-source sensor fusion, providing a complete scenario understanding foundation for decision-making and planning. Path planning based on high-precision maps generates sequential node task objectives. A hierarchical decision graph containing task, behavior, and motion layers is established using graph theory methods. Node pruning is performed using multi-dimensional integrated conditions such as traffic rule constraints, vehicle dynamics constraints, and safety constraints to form a set of feasible paths. In the evaluation phase, an adaptive weighting mechanism dynamically adjusts the cost function weights for path efficiency, driving safety, and energy economy. A multi-objective optimization algorithm selects the optimal constant-speed passage scheme with a score of 85 from three candidate paths, generating a control strategy with lane keeping as the primary priority. In the execution phase, a model predictive controller calculates the path geometric features in real time into control commands such as steering wheel angle and throttle opening. Kalman filtering is used to fuse positioning and inertial measurement data to achieve centimeter-level path tracking.

[0037] Step S103: Based on multi-source object data, obtain the actual path execution results, and input the actual path execution results into the priority constraint control strategy for analysis and correction, and obtain feedback information for dynamically adjusting the path; In this optional embodiment, based on multi-source object data, the actual path execution result is obtained, and the actual path execution result is input into the priority constraint control strategy for analysis and correction, resulting in feedback information for dynamically adjusting the path, including: Based on multi-source object data, the current position, speed and path deviation data of the vehicle are obtained in real time, and the path tracking algorithm is used to calculate the multi-source object data to obtain the actual path execution result. The actual path execution results are input into the priority constraint control strategy to analyze and evaluate path errors and real-time environmental data, and the feasibility of the path is analyzed using path planning algorithms to obtain feedback information for dynamic path adjustment.

[0038] Specifically, vehicle status data is collected in real time through multi-source sensor fusion, and GPS and IMU provide centimeter-level positioning. Visual and lidar are combined to monitor the path tracking status. When a lateral deviation of 0.3 meters, a heading angle deviation of 5 degrees, and a speed error of 0.2 meters per second are detected, the data is fused through the Kalman filter algorithm to form a structured actual path execution result. This result is input to the priority constraint control strategy module in real time and analyzed synchronously with environmental perception data such as the position of obstacles 80 meters ahead and changes in road curvature. Local path replanning and feasibility assessment are completed within 200 milliseconds, and finally, feedback information for dynamically adjusted path is generated, including specific parameters such as speed adjustment to 8 meters per second and path curvature optimization.

[0039] Step S104: Using a modular processing framework, the semantic structured cognitive configuration, priority constraint control strategy, and feedback information are integrated and processed to generate a set of operating state parameters for control execution; and control commands are generated based on the set of operating state parameters to drive the unmanned training vehicle to perform real-time navigation and path driving according to the optimized path.

[0040] In this optional embodiment, a modular processing framework is used to integrate and process semantically structured cognitive configurations, priority constraint control strategies, and feedback information to generate a set of operational state parameters for control execution. Control commands are then generated based on this set of operational state parameters to drive the unmanned driving training vehicle to perform real-time navigation and path driving along an optimized path, including: A modular processing framework is generated by combining path planning, environmental perception, and priority control. This modular processing framework is then exchanged through a data interface to obtain a unified modular processing framework. A unified modular processing framework is used to integrate semantic structured cognitive configuration, priority constraint control strategy and feedback information, and the Kalman filter algorithm is used to integrate the integrated results to obtain a set of running state parameters for control execution. The predictive control algorithm is used to calculate the set of operating state parameters for control execution, generate control commands, and drive the unmanned training vehicle to run along the optimized path to ensure that the vehicle performs the task according to the optimal path.

[0041] In this optional embodiment, a unified modular processing framework is used to integrate semantically structured cognitive configurations, priority constraint control strategies, and feedback information. The Kalman filter algorithm is then used to integrate the integrated results, resulting in a set of operational state parameters for control execution, including: Establish a unified spatiotemporal coordinate system, and perform timestamp synchronization and coordinate transformation on environmental data in semantic structured cognitive configuration, target data in priority constraint control strategy and vehicle status data in feedback information, and output a standardized spatiotemporal aligned data set; The standardized spatiotemporally aligned dataset is input into a Kalman filter, and the noise is processed by real-time estimation and observation noise. The filter gain is dynamically adjusted to output high-precision vehicle dynamic state parameter estimation results. Multi-source data cross-validation and residual analysis are performed on the high-precision vehicle dynamic state parameter estimation results to identify and remove abnormal data, and generate a set of control execution operation state parameters with credibility indicators.

[0042] Specifically, a unified spatiotemporal coordinate system with the vehicle's center of mass as the origin is established. Using a 10ms cycle, the environmental data (including relative positions of obstacles and lane line equations) in the semantic structured cognitive configuration, the target data (including path point coordinates and expected speed) in the priority constraint control strategy, and the vehicle status data (including actual position and heading angle) in the feedback information are time-stamped and transformed. The obstacle coordinates in the lidar coordinate system, the lane line parameters in the camera coordinate system, and the GPS global coordinate system are transformed into the vehicle coordinate system, and a standardized spatiotemporal aligned data set is output.

[0043] Specifically, the acquired data set is input into an extended Kalman filter. By estimating the system process noise covariance matrix Q (value 0.01) and observation noise covariance matrix R (value 0.1) in real time, the filter gain coefficient is dynamically adjusted. At a 100Hz update frequency, high-precision vehicle dynamic state parameter estimation results with position estimation error less than 2cm and heading angle estimation error less than 0.5 degrees are output. At the same time, the estimation results are cross-validated with multi-source data. By comparing the residual sequences of GPS positioning data and visual odometry data (threshold set to 3σ), abnormal positioning data caused by signal obstruction are identified and eliminated. At the same time, motion consistency is checked by combining the angular velocity data of the inertial measurement unit. Finally, a set of control execution operation state parameters with a 95% confidence interval is generated, providing reliable state input for subsequent control command generation.

[0044] In this optional embodiment, the Kalman filter state prediction formula is: ; ; In the formula, This represents the predicted state value at time t. Represents the state transition matrix. This represents the matrix showing the effect of control inputs on the state. Indicates control input, This represents the prediction of state covariance. Represents the process noise covariance; The observation update formula for the Kalman filter is: ; ; ; In the formula, Kalman gain represents the trade-off between prediction and observation. Represents the observation matrix. Represents the observation noise covariance. This represents the current sensor fusion observations. This represents the updated state estimate. This represents the updated state covariance.

[0045] Figure 2 An embodiment of a modular unmanned training vehicle control system under the ROS architecture of the present invention is shown.

[0046] In this optional embodiment, a modular unmanned training vehicle control system based on ROS architecture includes: The multi-source data integration module 201 is used to acquire multi-source object data based on multi-source sensors in the unmanned driving training vehicle, and to construct a semantic structured cognitive configuration of scene relationships based on the multi-source object data. The path task decision module 202 is used to construct a path task decision graph based on semantic structured cognitive configuration and path node task objectives; and to analyze and optimize the path task decision graph using multi-dimensional integrated constraints to obtain priority constraint control strategies. The path execution feedback module 203 is used to obtain the actual path execution result based on multi-source object data, and input the actual path execution result into the priority constraint control strategy for analysis and correction, so as to obtain feedback information for dynamically adjusting the path. The control command generation module 204 is used to integrate and process semantic structured cognitive configuration, priority constraint control strategy and feedback information using a modular processing framework to generate a set of operating state parameters for control execution; and to generate control commands based on the set of operating state parameters to drive the unmanned training vehicle to perform real-time navigation and path driving according to the optimized path.

[0047] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores static and dynamic information data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0048] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0049] In addition, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0050] In addition, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0051] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0052] This invention is not limited to the structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this invention is limited only by the appended claims.

Claims

1. A control method of a modular unmanned training vehicle under a ROS architecture, characterized in that, include: Multi-source object data is acquired from multi-source sensors in an unmanned training vehicle, and a semantic structured cognitive configuration of scene relationships is constructed based on the multi-source object data. Based on semantic structured cognitive configuration and path node task objectives, a path task decision graph is constructed; and multi-dimensional integrated constraints are used to analyze and optimize the path task decision graph to obtain a priority constraint control strategy. Based on multi-source object data, the actual path execution results are obtained, and the actual path execution results are input into the priority constraint control strategy for analysis and correction, so as to obtain feedback information for dynamically adjusting the path. A modular processing framework is used to integrate and process semantic structured cognitive configurations, priority constraint control strategies, and feedback information to generate a set of operational state parameters for control execution. Control commands are then generated based on the set of operational state parameters to drive the unmanned training vehicle to perform real-time navigation and path driving according to the optimized path. 2.The ROS architecture-based modular unmanned training vehicle control method of claim 1, wherein, The semantic structured cognitive configuration of scene relationships, which is based on multi-source object data acquired by multi-source sensors in an unmanned training vehicle and constructed based on the multi-source object data, includes: Raw point clouds, images, and radar signals are collected using lidar, cameras, and millimeter-wave radar, and time synchronization, coordinate unification, and noise reduction are performed to obtain standardized multi-source sensor data. Based on standardized multi-source sensor data, deep learning models and data association algorithms are used to perform image target recognition and fuse the target recognition results to generate a fused target list. The spatial topological and traffic logic relationships between targets in the fusion target list are analyzed, and a dynamic scene semantic network is constructed. The dynamic scene semantic network is transformed into a graph structure, with entities as attribute nodes and spatiotemporal and logical relationships as connecting edges, to obtain a semantically structured cognitive configuration of scene relationships.

3. The ROS architecture-based modular unmanned training vehicle control method according to claim 1, wherein, The multidimensional integrated constraints include: Constraints include traffic rules, vehicle dynamics, driving safety, environmental perception uncertainty, and task execution timeliness.

4. The ROS architecture-based modular unmanned training vehicle control method according to claim 3, characterized in that, The path task decision graph is constructed based on semantic structured cognitive configuration and path node task objectives; Furthermore, the path task decision graph is analytically optimized using multidimensional integrated constraints, resulting in priority constraint control strategies including: Based on high-precision maps and global tasks, path planning is performed to generate path node task objectives. Then, using graph theory, the path node task objectives and semantic structured cognitive configurations are deduced and analyzed to construct a path task decision graph. By using multidimensional integrated constraints, the path task decision graph is parsed and processed to remove nodes and branches that do not meet the hard constraints, resulting in a set of feasible paths that meet the constraints. By defining a comprehensive cost function, the set of feasible paths that meet the constraints is analyzed and evaluated to obtain the optimal execution path; and a priority constraint control strategy is constructed based on the optimal execution path.

5. The ROS architecture-based modular unmanned training vehicle control method according to claim 4, characterized in that, The method utilizes a defined comprehensive cost function to analyze and evaluate the set of feasible paths that meet the constraints, thereby obtaining the optimal execution path. The priority constraint control strategy based on the optimal execution path also includes: Based on path efficiency, driving safety and energy economy, a comprehensive evaluation system is generated, and the weight self-adaptive mechanism and the comprehensive evaluation system are used to dynamically adjust the weight coefficient of each dimension in the scene feature of the semantic structured cognitive configuration; Based on the comprehensive evaluation results generated by the dynamic weight coefficient, and combined with the priority constraint control strategy, real-time path optimization and control instruction generation are executed to drive the vehicle to travel along the optimal path, while dynamic path correction is implemented to respond to real-time environmental changes; Real-time monitoring of vehicle state and adjustment of control strategy are realized by using closed-loop feedback control mechanism to maintain the optimality and safety of path tracking throughout the journey.

6. The ROS architecture-based modular unmanned training vehicle control method according to claim 1, wherein, The actual path execution result is obtained based on multi-source object data, and the actual path execution result is input into the priority constraint control strategy for analysis and correction to obtain feedback information of dynamically adjusted path, including: Real-time acquisition of current position, speed and path deviation data of the vehicle based on multi-source object data, and calculation of multi-source object data by using path tracking algorithm to obtain actual path execution result; The actual path execution result is input into the priority constraint control strategy, and the path error and real-time environmental data are analyzed and evaluated, and the feasibility of the path is analyzed by using path planning algorithm to obtain feedback information of dynamically adjusted path.

7. The ROS architecture-based modular unmanned training vehicle control method according to claim 1, characterized in that, The data integration processing of semantic structured cognitive configuration, priority constraint control strategy and feedback information is realized by using modular processing framework to generate control execution running state parameter set; And according to the running state parameter set, control instructions are generated to drive the unmanned real training vehicle to navigate and travel along the optimized path in real time, including: A modular processing framework is generated by combining path planning, environment perception and priority control, and the modular processing framework is exchanged through data interface to obtain a unified modular processing framework; The semantic structured cognitive configuration, priority constraint control strategy and feedback information are integrated by using the unified modular processing framework, and the integrated results are integrated by using Kalman filtering algorithm to obtain the running state parameter set of control execution; The running state parameter set of control execution is calculated by using running prediction control algorithm to generate control instructions, and the unmanned real training vehicle is driven to run according to the optimized path to ensure that the vehicle executes the task according to the optimal path.

8. The ROS architecture-based modular unmanned training vehicle control method according to claim 7, characterized in that, The semantic structured cognitive configuration, priority constraint control strategy and feedback information are integrated by using the unified modular processing framework, and the integrated results are integrated by using Kalman filtering algorithm to obtain the running state parameter set of control execution, including: A unified space-time coordinate system is established, and the time stamp synchronization and coordinate conversion of environmental data in semantic structured cognitive configuration, target data in priority constraint control strategy and vehicle state data in feedback information are realized to output a standardized space-time alignment data set; The standardized space-time alignment data set is input into Kalman filter, and is processed by real-time estimation of noise and observation noise to dynamically adjust the filtering gain, and output high-precision vehicle dynamic state parameter estimation result; The high-precision vehicle dynamic state parameter estimation result is subjected to multi-source data cross verification and residual analysis, abnormal data is identified and removed, and a control execution running state parameter set with a reliability index is generated.

9. The ROS architecture-based modular unmanned training vehicle control method according to claim 8, characterized in that, The state prediction formula of the Kalman filter is: ； ； wherein denotes a state prediction at time t, denotes a state transition matrix, denotes a matrix representing the influence of control inputs on the state, denotes a control input, denotes a state covariance prediction, denotes a process noise covariance; The observation update formula of the Kalman filter is: ; ; ; wherein denotes the Kalman gain, which controls the trade-off between prediction and observation, denotes the observation matrix, denotes the observation noise covariance, denotes the current sensor fused observation, denotes the updated state estimate, denotes the updated state covariance.

10. A ROS architecture-based modular unmanned training vehicle control system, characterized in that, It comprises: A multi-source data integration module is configured to acquire multi-source object data based on multi-source sensors in the unmanned training vehicle, and to construct a semantic structured cognitive configuration of scene relationships according to the multi-source object data; A path task decision module is configured to construct a path task decision graph based on the semantic structured cognitive configuration and path node task targets, to analyze and optimize the path task decision graph using multi-dimensional integrated constraint conditions, and to obtain a priority constraint control strategy; A path execution feedback module is configured to acquire actual path execution results based on the multi-source object data, to input the actual path execution results into the priority constraint control strategy for analysis and correction, and to obtain feedback information of a dynamically adjusted path; A control instruction generation module is configured to perform data integration processing on the semantic structured cognitive configuration, the priority constraint control strategy and the feedback information using a modular processing framework, to generate a running state parameter set for control execution, and to generate a control instruction according to the running state parameter set to drive the unmanned training vehicle to navigate and travel along the optimized path in real time.

Images