Special vehicle automatic driving system based on multi-scene intelligent scheduling and control method thereof

Abstract

The invention discloses a special vehicle automatic driving system based on multi-scene intelligent scheduling and a control method thereof, and relates to the technical field of intelligent driving. The system comprises a scene management module, an execution management module, a state management module, a control module and an operating system layer. The scene management module identifies four driving scenes of manual driving, calling, parking and charging and following and loads a matching component; the state management module collects and feeds back vehicle real-time state data; the execution management module coordinates component interaction, manages task priority and triggers safety interruption; and the control module executes acceleration, deceleration and steering actions. The control method comprises the five steps of scene recognition loading, state synchronous detection, task scheduling decision making, action execution and feedback closed loop. The problems that an existing system is single in scene, complex in control and insufficient in safety are solved, multi-scene intelligent switching and unified management are achieved, the system is suitable for low-speed special vehicles such as scooters for old people, and safety and intelligence are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of automatic driving, and particularly relates to a special vehicle automatic driving system based on multi-scene intelligent scheduling and a control method thereof. BACKGROUND

[0002] With the acceleration of population aging, the special vehicle for the elderly, referred to as the elderly vehicle, gradually becomes the core tool for community travel, and the safety and scene adaptability of its intelligent driving system are increasingly urgent. However, the current special vehicle intelligent driving system is still limited by the technical architecture and has three key defects, which makes it difficult to meet the use requirements of the elderly vehicle in complex scenes: First, the driving function scene is single and lacks intelligent scheduling capability. The existing system only supports simple scenes such as basic straight-line cruising and fixed-distance obstacle avoidance, and cannot adapt to complex scenes such as the winding road network of the community frequently contacted by the elderly, the dense area of the market, and the slope start-stop. At the same time, the scheduling logic is fixed, and it cannot dynamically adjust the driving strategy in combination with the physiological characteristics of the elderly, such as the need for temporary parking for rest after sitting for a long time, and the real-time road conditions, such as sudden children crossing or obstacles occupying the road, which makes the system not flexible enough and requires frequent manual intervention when used by the elderly, thereby increasing the operation burden. Second, the control execution part structure is complex, and the design of driving and steering separation leads to low communication efficiency of the module. The system adopts an independent architecture in which the driving unit is responsible for power output and the steering unit is responsible for direction control, and the two rely on different control chips and sensors, which not only has high hardware redundancy, but also increases the failure rate. More importantly, the data interaction between modules is delayed, and in the emergency avoidance scene, the driving deceleration and steering avoidance actions are often out of sync, thereby increasing the risk of rollover. Third, the execution layer lacks unified scheduling and safety feedback mechanism, which easily causes control conflict or response delay. On the one hand, there is no scheduling hub to coordinate multiple module instructions, such as when the obstacle avoidance module triggers emergency braking, the speed control module may still be outputting constant speed instructions, resulting in chaotic execution. On the other hand, the safety feedback link is broken, and when the sensor detects tire deflation, battery overload and other hidden dangers, it can only be prompted by a single warning light, and cannot be fed back to the driving and steering modules simultaneously, so the modules cannot adjust the action in advance, which poses a safety threat to the elderly users with weak reaction ability. Therefore, there is an urgent need for a special vehicle intelligent driving system with unified execution management and multi-scene scheduling capability to fill this technical gap and ensure the safety of the elderly in travel. SUMMARY

[0003] In order to overcome the shortcomings of the prior art, the purpose of the present application is to provide a special vehicle automatic driving system based on multi-scene intelligent scheduling and a control method thereof, and the above technical purpose of the present application is realized by the following technical scheme: The first objective of this invention is to provide an autonomous driving system for special-purpose vehicles based on multi-scenario intelligent scheduling. The system includes a scenario management module, a state management module, an execution management module, and a control module. The scenario management module (ScM) identifies the current operating scenario, the state management module (SM) monitors the vehicle's operating status and provides real-time feedback, the execution management module (EM) is responsible for instruction scheduling and safety policy control, and each scenario algorithm component communicates with a unified control module (Control) to execute actions such as acceleration, deceleration, and steering. The scenario management module is used to identify manual driving scenarios, summoning scenarios, parking and charging scenarios, and following scenarios according to user input instructions or preset task requirements, and load algorithm component combinations that match the identified scenarios, while sending an execution preparation notification to the execution management module. The status management module is used to continuously collect vehicle operation status data and sensor output data of the special vehicle, perform real-time analysis and processing on the collected data to extract vehicle health status information and potential risk information, and send the health status information and potential risk information to the execution management module in real time. The execution management module is used to receive the scenario task type information issued by the scenario management module, set the task priority and execution strategy in combination with the current operating status of the special vehicle, monitor the potential risk information sent by the status management module in real time, terminate the current non-emergency task and issue a safety interruption command if a safety alarm is detected, and coordinate the interaction between each scenario algorithm component and the control module. The control module is used to receive control quantities sent by the algorithm components of each scenario through a standard communication interface, drive the power system, steering system and braking system of the special vehicle to perform acceleration, deceleration or steering operations, and generate feedback signals in real time to send to the execution management module. Operating system layer: calls the SDK interfaces of different cameras to realize image acquisition, and realizes data interaction with the vehicle controller through the CAN bus. Based on the OpenCV algorithm library, it realizes image preprocessing and feature extraction. The network communication submodule communicates with the cloud or other devices through TCP / IP.

[0004] The second objective of this invention is to provide a control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling, the steps of which are as follows: Step S1 Scene Recognition and Loading The scene management module identifies the current driving scene to be run according to user input instructions or preset task requirements, including but not limited to manual driving scenes, calling scenes, parking and charging scenes, and following scenes; based on the identification result, the algorithm component combination matched with the driving scene is loaded to ensure that the component function is adapted to the scene task; at the same time, an execution preparation notification is sent to the execution management module to trigger the execution management module to enter an execution preparation state, and the initialization is completed for subsequent task scheduling.

[0005] Step S2 state synchronization and detection The state management module continuously collects vehicle running state data (such as vehicle speed, gear position, and battery state) and sensor output data (such as millimeter wave radar, camera, and current sensor data) of the special vehicle; the collected data is analyzed and processed in real time to extract vehicle health state information (such as motor operating temperature and communication link stability) and potential risk information (such as obstacle distance and sensor abnormalities); the above health state information and risk information are sent to the execution management module in real time to realize the bidirectional state synchronization between the state management module and the execution management module, and provide data support for task execution.

[0006] Step S3 task scheduling and safety decision The execution management module receives the scene task type information (such as calling task and parking and charging task) issued by the scene management module, sets the priority (such as safety class task priority higher than regular driving task) and specific execution strategy (such as path planning priority of calling scene and docking accuracy requirement of parking and charging scene) of the corresponding task in combination with the current running state of the special vehicle; during task execution, the execution management module monitors the risk information sent by the state management module in real time, and if a safety alarm (such as too close obstacle and insufficient power) is detected, the current non-urgent task is terminated immediately, and a safety interruption command is generated and issued to the related execution unit to ensure vehicle safety.

[0007] Step S4 action execution control The algorithm component bound to the current driving scene calculates the vehicle control quantity (including driving speed, steering angle, and brake force) according to the scene task logic (such as path planning logic of calling scene and target tracking logic of following scene) and the real-time data fed back by the state management module; the calculated control quantity is sent to the control module through a preset standard communication interface (CAN bus, ROS2); after receiving the control quantity, the control module drives the power system, steering system, and braking system of the special vehicle to perform acceleration, deceleration, or steering operation, and completes the action landing of the scene task.

[0008] Step S5 feedback loop and task ending After the control module executes the action, a feedback signal is generated in real time, the feedback signal including an action execution state (such as "execution completed", "execution in progress", "execution exception") and a control error (such as a deviation of an actual vehicle speed from a target vehicle speed, a deviation of an actual steering angle from a target angle); the feedback signal is sent to the execution management module, the feedback signal is summarized and verified by the execution management module; after verification, the execution management module forwards the feedback signal to the state management module and the scene management module respectively: the state management module updates the vehicle state database based on the feedback signal, and the scene management module judges the execution progress of the scene task based on the feedback signal; a closed-loop control mechanism is constructed based on the above feedback link, and the control parameters of the subsequent task are optimized based on the historical feedback data, forming a system self-learning mechanism; the execution management module judges whether the task is completed according to the task progress fed back by the scene management module, and if the task is completed, triggers a task ending process, notifies the scene management module to unload the algorithm components of the current scene, and the system returns to the initial standby state.

[0009] Compared with the prior art, the present application has the following beneficial effects: (1) The system is designed for low-speed special vehicles such as old people's scooters, the scene management module can identify four scenes of manual driving, calling, parking and charging, and following, and load corresponding algorithm components; the state management module collects real-time vehicle state and sensor data and feeds back; the execution management module coordinates component interaction, manages task priority and triggers safety interruption; the control module uniformly executes acceleration, deceleration and steering actions; the operating system layer provides a tool library and basic function support, laying a bottom layer foundation for the operation of each module.

[0010] (2) The system performs outstandingly in scene adaptation, breaking through the limitation of single scene of existing systems, covering high-frequency daily travel scenes of old people, and forming a complete use closed loop. The scene triggering mode is flexible, which not only supports users to send requests through an APP, but also can be triggered automatically by the system, such as triggering the parking and charging scene when the power is too low, and triggering the manual driving scene when the driving personnel is detected, which is in line with the operation habits of old users, reduces the frequency of manual intervention, and reduces the difficulty of use.

[0011] (3) On the control architecture, the system uses a unified control module and a standardized communication interface, changing the traditional design of separating driving and steering. Each scene algorithm component interacts with the control module through interfaces such as CAN bus and ROS2, the control delay is not more than 100ms, avoiding the risk of action asynchronization in emergency situations. At the same time, unified control reduces hardware redundancy, reduces failure rate, simplifies data interaction link, significantly improves overall control efficiency, and adapts to the demand of low-speed special vehicles for timely response.

[0012] (4) The system builds a three-level safety mechanism. The state management module monitors the vehicle state and potential risks in real time, the execution management module sets the task priority, the safety task priority is higher than the regular task, the non-urgent task is terminated and the interrupt command is issued when the safety alarm is detected, and the control module executes the deceleration or braking operation preferentially. In addition, the control module will feedback the execution state and control error, the execution management module optimizes the parameters combined with historical data to form a self-learning mechanism, and the safety of the elderly users is guaranteed.

[0013] (5) The control method is divided into five steps of scene recognition loading, state synchronization detection, task scheduling decision, action execution and feedback loop. From the scene recognition loading corresponding components to the state synchronization providing data support, then to the task scheduling guaranteeing safe execution, followed by the action execution landing the scene task, and finally through the feedback loop to update the state, judge the progress and optimize the parameters, the process is complete and the logic is close, realizing the intelligent switching and unified management of multiple scenes, effectively solving the problems of complex control and insufficient safety of existing systems, and improving the intelligence and safety of the automatic driving of the special vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is the overall structure schematic diagram of the special vehicle automatic driving system based on multi-scene intelligent scheduling of the present application; Figure 2 the work flow chart of the human driving scene; Figure 3 the work flow chart of the calling scene; Figure 4 the flow chart of the parking and charging scene; Figure 5 the flow chart of the following scene; Figure 6 the communication relationship schematic diagram among the scene management module, the state management module, the execution management module and the control module; Figure 7 the system scene switching flow chart. DETAILED DESCRIPTION

[0015] The present application will be further described in detail below in combination with the drawings.

[0016] As shown in Figure 1 , the special vehicle automatic driving system based on multi-scene intelligent scheduling comprises a scene management module, a state management module, an execution management module, a control module and an operating system layer; the current running scene is recognized through the scene management module, the vehicle running state is monitored and real-time feedback is provided through the state management module, the execution management module is responsible for instruction scheduling and safety strategy control, and each scene algorithm component communicates with the unified control module to realize the execution of actions such as acceleration, deceleration and steering.

[0017] The operating system layer provides tool libraries such as camera SDK, CAN, and OpenCV, and implements basic functions such as network protocol stack, IO and security based on POSIX OS, providing underlying technical support for the operation of various modules of the intelligent driving system.

[0018] like Figure 2 As shown, the execution process of the vehicle manual driving scenario is as follows: after the control module detects the driver's intervention, it enters the manual driving scenario. Then, according to the task of the manual driving scenario, the manual driving component is activated to collect throttle and steering wheel signals to control the vehicle in real time. Finally, the control module controls the vehicle to move.

[0019] like Figure 3 As shown, the execution flow of the vehicle summoning function is as follows: After the user sends a summoning request on the app to enter the summoning scene, the system starts a series of components according to the summoning task; among them, the ultrasonic component detects surrounding obstacles, the inertial measurement component detects the vehicle's running status and positioning in real time; the summoning component monitors the relationship between the vehicle and the summoning location in real time and controls the vehicle's operation, and finally the control module controls the vehicle to move, completing the summoning process.

[0020] like Figure 4 As shown, the automatic parking and charging process is as follows: When the user sends a parking and charging request on the app or the control module detects that the vehicle's battery is too low, the parking and charging scenario is entered. First, the point-to-point component in the parking and charging scenario is activated, and the path planning component plans the vehicle's path to the vicinity of the charging station. At the same time, the camera component is activated to collect the surrounding environment of the charging station, and the data algorithm identifies the location of the charging station. Then, the path planning component communicates with the charging station component in real time to adjust the positional relationship between the vehicle and the charging station and issue control commands. Finally, the control module controls the vehicle to move to the charging station location to complete the charging.

[0021] like Figure 5 As shown, the process of automatic vehicle following is as follows: After the user sends a follow request on the app, the camera component and target detection component in the following scene are activated; the camera component then collects the feature information of the person in front of the vehicle, and the target detection component identifies and tracks the target based on this feature information; then the path following component communicates with the path planning component and radar component in real time, sends out the path trajectory and converts it into control commands; finally, the control module controls the vehicle movement according to the commands to achieve automatic following.

[0022] like Figure 6As shown, the interaction process of an intelligent vehicle from startup to scene activation is as follows: After the user initiates the vehicle startup command, the execution management, status management, and scene management sequentially transmit the startup signal. The scene management triggers the startup self-check and intelligent driving system readiness process, and then starts the human-driving system. The human-driving component provides feedback on the human-driving scene status. Afterward, the user actively activates the scene, triggering the startup summon / parking / follow process. Through the coordinated execution of various modules, the intelligent driving system finally provides feedback on the status, completing the entire process of interaction and status management from vehicle startup to intelligent scene activation.

[0023] like Figure 7 As shown, after the vehicle is powered on by the circuit breaker and the intelligent driving system is initialized, it enters the manual driving mode, which can be switched to three intelligent modes: Follow, Parking & Charging, and Summon. The Follow mode requires target activation, following, and reactivation or system failure handling after target loss. The Parking & Charging mode follows the process of arriving at the parking point, recognizing the feature plate, initial posture adjustment, recognizing the QR code, accurately aligning with the charging station, and starting charging. The Summon mode performs operations such as navigation to the pick-up point after receiving the command. If there are no obstacles, it successfully reaches the pick-up point; if there are obstacles, it stops and reports. If decoding fails, it performs a self-assessment and determines whether manual intervention is required. The modes are switched through a switch or conditional trigger. In case of abnormality, it reverts to manual driving or enters a sleep state.

[0024] The control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling includes the following steps: Step S1 Scene Recognition and Loading The scenario management module identifies the driving scenario to be run based on user input commands or preset task requirements, including but not limited to manual driving scenarios, summoning scenarios, parking and charging scenarios, and following scenarios; it loads a combination of algorithm components that match the driving scenario based on the identification results to ensure that the component functions are adapted to the scenario task; at the same time, it sends an execution preparation notification to the execution management module, triggering the execution management module to enter the execution preparation state and initialize it for subsequent task scheduling.

[0025] Step S2: State Synchronization and Detection The status management module continuously collects vehicle operating status data (such as vehicle speed, gear position, and battery status) and output data from various sensors (such as millimeter-wave radar, camera, and current sensor data). It performs real-time analysis and processing of the collected data to extract vehicle health status information (such as motor operating temperature and communication link stability) and potential risk information (such as obstacle distance and sensor anomalies). The module then sends the aforementioned health status information and risk information to the execution management module in real time, achieving bidirectional status synchronization between the status management module and the execution management module, and providing data support for task execution.

[0026] Step S3: Task Scheduling and Security Decisions The execution management module receives scenario task type information (such as summoning tasks and parking / charging tasks) from the scenario management module. Based on the current operating status of the special vehicle, it sets the priority of the corresponding tasks (such as safety tasks having higher priority than regular driving tasks) and specific execution strategies (such as the path planning priority for summoning scenarios and the docking accuracy requirements for parking / charging scenarios). During task execution, the execution management module monitors the risk information sent by the status management module in real time. If it detects a safety alarm issued by the status management module (such as an obstacle being too close or insufficient power), it immediately terminates the current non-emergency task, generates and issues a safety interruption command to the relevant execution unit to ensure vehicle safety.

[0027] Step S4 Action Execution Control The algorithm components, which are bound to the current driving scenario, calculate vehicle control quantities (including driving speed, steering angle, braking force, etc.) based on the scenario task logic (such as path planning logic for the summoning scenario and target tracking logic for the following scenario) and real-time data fed back by the state management module. The calculated control quantities are sent to the control module through a preset standard communication interface (CAN bus, ROS2). After receiving the control quantities, the control module drives the power system, steering system and braking system of the special vehicle to perform acceleration, deceleration or steering operations to complete the action implementation of the scenario task.

[0028] Step S5: Feedback Loop Closure and Task Closure After the control module executes an action, it generates feedback signals in real time. These signals include the action execution status (e.g., "execution completed," "in execution," "execution error") and control errors (e.g., the deviation between actual vehicle speed and target vehicle speed, and the deviation between actual steering angle and target angle). The feedback signals are then sent to the execution management module, which summarizes and verifies them. After successful verification, the execution management module forwards the feedback signals to the status management module and the scenario management module. The status management module updates the vehicle status database based on the feedback signals, while the scenario management module determines the execution progress of the scenario task based on the feedback signals. A closed-loop control mechanism is constructed based on the above feedback chain, and the control parameters for subsequent tasks are optimized by combining historical feedback data, forming a system self-learning mechanism. The execution management module determines whether the task is completed based on the task progress feedback from the scenario management module. If the task is completed, the task completion process is triggered, and the scenario management module is notified to unload the algorithm components of the current scenario, and the system returns to its initial state.

[0029] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling, characterized in that, It includes a scene management module, a state management module, an execution management module, a control module, and an operating system layer; The scene management module identifies the current operating scene, the status management module monitors the vehicle's operating status and the output data of each sensor, and provides real-time feedback, the execution management module is responsible for command scheduling and safety policy control, and each scene algorithm component communicates with the unified control module to achieve acceleration, deceleration and steering. The operating system layer provides camera SDK, CAN, and OpenCV tool libraries, and implements network protocol stack, IO, and basic security functions based on POSIX OS, providing underlying support for the operation of the scene management module, state management module, execution management module, and control module.

2. The special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 1, characterized in that, The vehicle operating status data includes vehicle speed, gear position, and battery status; the sensor output data includes millimeter-wave radar data, camera data, and current sensor data.

3. The special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 1, characterized in that, The task priority settings in the execution management module satisfy the following condition: safety-related tasks have a higher priority than regular driving tasks.

4. The special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 1, characterized in that, The algorithm component combinations loaded by the scene management module satisfy the following: the summoning scene is loaded with ultrasonic components and inertial measurement components; the parking and charging scene is loaded with point-to-point components, path planning components and camera components; the following scene is loaded with camera components, target detection components and radar components; and the manual driving scene is loaded with human driving components.

5. A control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling, characterized in that, Includes the following steps: S1: Scene Recognition and Loading The scenario management module identifies manual driving scenarios, summoning scenarios, parking and charging scenarios, or following scenarios based on user input instructions or preset task requirements, loads a combination of algorithm components that match the identified scenarios, and sends an execution preparation notification to the execution management module to trigger the execution management module to enter the execution preparation state. S2: State Synchronization and Detection The status management module continuously collects vehicle operation status data and sensor output data of the special vehicle, performs real-time analysis and processing on the collected data to extract vehicle health status information and potential risk information, and sends the health status information and potential risk information to the execution management module in real time to achieve two-way status synchronization. S3: Task Scheduling and Security Decisions The execution management module receives the scenario task type information issued by the scenario management module, sets the task priority and execution strategy in combination with the current operating status of the special vehicle, monitors potential risk information in real time, and terminates the current non-emergency task and issues a safety interruption command if a safety alarm is detected. S4: Action Execution Control The algorithm component bound to the current scene calculates the vehicle control quantity based on the scene task logic and the real-time data fed back by the state management module, and sends the control quantity to the control module through the standard communication interface. The control module drives the special vehicle to perform acceleration, deceleration or steering operations. S5: Feedback Loop and Task Closure The control module generates a feedback signal containing the action execution status and control error, and sends it to the execution management module. After verifying the feedback signal, the execution management module forwards it to the status management module and the scene management module. The status management module updates the vehicle status database, and the scene management module determines the task progress. If the task is completed, the execution management module triggers the task completion process, notifies the scene management module to unload the current scene algorithm component, and the system returns to the initial standby state.

6. The control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 5, characterized in that, In step S1, the scenario management module identifies the triggering methods of the scenario, including: the user sending a call request through the APP to trigger the call scenario, the user sending a parking charging request through the APP or the control module detecting that the battery is too low to trigger the parking charging scenario, the user sending a follow request through the APP to trigger the follow scenario, and the detection of driver intervention to trigger the manual driving scenario.

7. The control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 5, characterized in that, In step S3, the execution strategy includes: setting the path planning priority for the summoning scenario and setting the docking accuracy requirement for the charging scenario; the task priority satisfies the following: safety tasks have a higher priority than regular driving tasks, and the safety tasks include obstacle avoidance tasks and low battery return tasks.

8. The control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 5, characterized in that, In step S4, the vehicle control quantities include driving speed, steering angle, and braking force; The scenario task logic includes: path planning logic for the summoning scenario, docking and charging logic for the parking scenario, target tracking logic for the following scenario, and accelerator and steering wheel signal acquisition logic for the manual driving scenario.

9. The control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 5, characterized in that, In step S4, the control delay of the control module does not exceed 100ms; in step S3, after the execution management module issues a safety interruption command, the control module prioritizes deceleration or braking operations to ensure that the special vehicle stops or avoids risks within a safe distance.

10. The control method for a special-purpose vehicle autonomous driving system based on multi-scenario intelligent scheduling according to claim 5, characterized in that, In step S5, the action execution status includes "execution completed", "execution in progress" and "execution error"; the control error includes the deviation between the actual vehicle speed and the target vehicle speed, and the deviation between the actual steering angle and the target steering angle; while forwarding feedback signals, the execution management module optimizes the control parameters of subsequent tasks by combining historical feedback data, forming a system self-learning mechanism.

Images