Intelligent control system of shuttle vehicle for tray conveying

Abstract

The invention discloses an intelligent shuttle vehicle control system for tray conveying, which belongs to the technical field of intelligent shuttle vehicle control and comprises a multi-mode sensing module S1, a dynamic path planning module S2, a multi-vehicle collaborative scheduling module S3, an energy efficiency optimization module S4 and an edge calculation module S5. The multi-mode sensing module S1 comprises an environment sensing sub-module S11 and a cargo state monitoring sub-module S12, the environment sensing sub-module S11 comprises laser radar data preprocessing and visual sensor calibration and fusion, and the laser radar data preprocessing adopts a filtering algorithm based on statistics to remove outliers generated by environment interference. Through the distributed consensus algorithm and the task allocation strategy, the system throughput is improved, the path conflict rate is reduced, the work waiting time is remarkably shortened, the improved A * algorithm is combined with the reinforcement learning model, the time consumption of single path planning is shortened, the path generation efficiency is improved, and the method adapts to the high-dynamic storage environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of shuttle vehicle intelligent control, and in particular relates to an intelligent control system for a shuttle vehicle used for pallet transportation. Background Art

[0002] With the rapid development of the logistics industry and intelligent manufacturing, warehouse logistics automation has become key to improving supply chain efficiency. As the core equipment of automated warehousing systems, the intelligence level of pallet shuttles directly affects the overall performance of the warehousing system. However, traditional shuttle control systems have the following problems: Traditional systems often use fixed paths or simple obstacle avoidance strategies, which are difficult to dynamically adapt to complex and changing warehousing environments, resulting in frequent path conflicts and low system efficiency; when multiple vehicles work together, they lack an efficient scheduling mechanism, which is prone to path intersections and resource competition, seriously restricting system throughput; and traditional systems do not fully consider the dynamic characteristics and operating status of the shuttles, resulting in significant energy waste and high operating costs. Summary of the Invention

[0003] The purpose of the present invention is to provide an intelligent control system for a shuttle vehicle for pallet conveying, so as to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: including S1: multimodal perception module, S2: dynamic path planning module, S3: multi-vehicle collaborative scheduling module, S4: energy efficiency optimization module, and S5: edge computing module, wherein the S1: multimodal perception module includes S11: environmental perception submodule and S12: cargo status monitoring submodule, and the S11: environmental perception submodule includes lidar data preprocessing and visual sensor calibration and fusion, wherein the lidar data preprocessing adopts a statistical-based filtering algorithm to remove outliers generated by environmental interference, and uses the RANSAC algorithm to fit the ground plane and divide the point cloud into ground points and non-ground points, wherein the visual sensor calibration and fusion adopts the Zhang Zhengyou calibration method, uses a checkerboard calibration plate to calibrate the binocular camera, obtains the internal and external parameters and distortion coefficient of the camera, and fuses the depth map generated by the binocular camera with the lidar point cloud, utilizing the advantages of each to improve the accuracy and robustness of environmental perception;

[0005] As a further preferred embodiment of the present technical solution: the S12: cargo status monitoring submodule includes cargo positioning and tracking and cargo status anomaly detection, wherein cargo positioning and tracking is based on the YOLOv8 model to achieve real-time detection and identification of pallet cargo, and adopts the DeepSORT algorithm, combined with the appearance characteristics and motion information of the cargo, to achieve stable tracking of multiple cargoes, wherein the cargo status anomaly detection determines whether the cargo has shifted by comparing the difference between the current position of the cargo and the preset position, and determines whether the cargo exceeds the carrying range of the pallet or shuttle vehicle based on the size information of the cargo;

[0006] As a further preferred embodiment of the present technical solution: the S2: dynamic path planning module includes S21: improved A* algorithm optimization and S22: reinforcement learning model application, the S21: improved A* algorithm optimization includes heuristic function design and time window constraint implementation, wherein the heuristic function design dynamically adjusts the weight parameters in the heuristic function according to the real-time status of the storage environment, learns the optimal weight combination through the reinforcement learning algorithm, improves the efficiency of path planning, incorporates multiple optimization objectives into the heuristic function, and realizes multi-objective optimization through weight distribution, wherein the time window constraint is implemented according to the kinematic model and path length of the shuttle vehicle, calculates the time window of each path node, dynamically updates the time window information according to the actual operation of the shuttle vehicle, and ensures the time window coordination between multiple vehicles through real-time communication;

[0007] As a further preferred embodiment of the present technical solution: S22: the application of the reinforcement learning model includes state space construction, action space design and reward function design, wherein the state space construction includes the current node coordinates, target point coordinates, adjacent vehicle positions, order priority, equipment load rate, etc., and through sensor fusion and communication technology, comprehensive environmental state information is obtained, including the position, speed, acceleration, battery level, etc. of the shuttle vehicle, and the vehicle status is monitored in real time through the vehicle's own sensors and battery management system, wherein the action space design defines the shuttle vehicle's movement direction, such as forward, backward, left turn, right turn and speed level, such as low speed, medium speed, and high speed, and simplifies the decision-making process by discretizing the action space, wherein the reward function design designs efficiency reward items based on the time, path length and energy consumption of the shuttle vehicle to complete the task, and encourages the shuttle vehicle to choose an efficient and energy-saving path;

[0008] As a further preferred embodiment of the present technical solution: the S3: multi-vehicle collaborative scheduling module includes S31: distributed consensus algorithm deepening and S32: conflict prediction and collaborative avoidance, the S31: distributed consensus algorithm deepening includes leader election and fault recovery, task allocation and load balancing, wherein the leader election fault recovery adopts the Raft protocol to elect the leader, and maintains the leader's status through the heartbeat mechanism. If the leader fails, a new election process is triggered. When the shuttle or scheduling node fails, the redundancy design and failover mechanism are used to ensure the continuous operation of the system. The task allocation and load balancing are based on the current position of the shuttle, task priority and equipment load, and the Hungarian algorithm or auction algorithm is used to allocate tasks. The shuttle load is monitored in real time, and load balancing is achieved by adjusting the task allocation strategy.

[0009] As a further preferred embodiment of the present technical solution: said S32: conflict prediction and collaborative avoidance includes a conflict detection algorithm and a collaborative avoidance strategy, wherein the conflict detection algorithm predicts the future trajectory of the shuttle vehicle through Kalman filtering or particle filtering, obtains the status information of the adjacent shuttle vehicles through real-time communication, performs conflict detection and triggers collaborative avoidance based on the real-time status information, wherein the collaborative avoidance strategy adjusts the speed of the shuttle vehicle, such as slowing down or stopping, to avoid collision when a potential conflict is detected, and triggers a path replanning process in the event of a serious conflict, and generates a new conflict-free path for the affected shuttle vehicle through a dynamic path planning algorithm;

[0010] As a further preferred embodiment of the present technical solution: the S4: energy efficiency optimization module includes S41: dynamic model deepening and S42: energy management strategy, wherein S41: dynamic model deepening includes motor control strategy and speed optimization algorithm, wherein the motor control strategy adopts a vector control algorithm to achieve independent control of motor torque and magnetic flux, and during braking, kinetic energy is converted into electrical energy and stored through an energy recovery unit, and a maximum power point tracking algorithm is adopted to improve energy recovery efficiency, wherein the speed optimization algorithm generates a smooth acceleration / deceleration curve through model predictive control or optimal control algorithm, reduces the impact of sudden stop and start on the motor and battery, extends the life of the equipment, and adaptively adjusts the speed of the shuttle according to the slope, curvature and obstacle conditions of the path;

[0011] As a further preferred embodiment of the present technical solution: S42: the energy management strategy includes a battery management system and an energy scheduling strategy, wherein the battery management system monitors the battery's voltage, current, temperature, remaining power and other status information in real time, estimates the battery's health status and life through the BMS algorithm, adopts active balancing or passive balancing technology, and balances the power of each battery cell in the battery pack, wherein the energy scheduling measurement reasonably allocates energy resources according to the priority and urgency of the task, ensures that high-priority tasks obtain energy supply first, improves the overall efficiency of the system, and recovers and stores the braking energy through the energy recovery unit for use in subsequent tasks;

[0012] As a further preferred embodiment of the present technical solution: the S5: edge computing module includes S51: hardware configuration deepening and S52: cloud-edge collaboration deepening, the S51: hardware configuration deepening includes AI chip selection and optimization, sensor fusion and data synchronization, wherein the AI ​​chip selection and optimization selects an AI chip suitable for the shuttle vehicle intelligent control system, such as NVIDIA Jetson AGX Orin, taking into account factors such as the chip's computing power, power consumption, interface and cost to ensure that the system requirements are met, and by optimizing the chip's operating frequency, voltage and temperature and other parameters, the chip's performance and stability are improved, wherein the sensor fusion and data synchronization adopt algorithms such as Kalman filtering, extended Kalman filtering or particle filtering to fuse data from multiple sensors, and adopt technologies such as timestamp synchronization or hardware synchronization to ensure the time consistency of multi-sensor data, and improve the accuracy of subsequent processing and analysis through the data synchronization mechanism;

[0013] As a further preferred embodiment of the present technical solution: S52: Deepening of cloud-edge collaboration includes data synchronization and storage, model iteration and upgrade, wherein data synchronization and storage adopt protocols such as MQTT, CoAP or HTTP to realize data synchronization between the edge computing unit and the cloud, and deploy a big data platform such as Hadoop, Spark or Flink on the cloud to store and manage the synchronized data, wherein model iteration and upgrade utilizes the big data platform on the cloud to train and optimize the operation data of the shuttle vehicle, and deploys the trained model to the edge computing unit to realize real-time reasoning and decision-making of the model.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] 1. This invention uses a distributed consensus algorithm and task allocation strategy to improve system throughput, reduce path conflict rate, and significantly reduce job waiting time. The improved A* algorithm is combined with a reinforcement learning model to shorten the time consumption of single path planning and improve path generation efficiency. It adapts to highly dynamic warehousing environments and supports seamless collaboration between shuttle vehicles and equipment such as AGVs and stackers, achieving seamless connection of cross-regional transportation tasks and improving overall operation efficiency.

[0016] 2. This invention uses a deep learning model to reduce the accuracy of cargo offset detection and shorten fault response time, significantly improving operational safety. The multimodal perception module integrates lidar and visual data to achieve three-dimensional reconstruction and obstacle detection of the storage environment, adapting to different lighting conditions.

[0017] 3. The present invention supports remote training and optimization of the model through the cloud big data platform, realizes model iteration of the edge computing unit through OTA technology, and continuously improves system performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1This is a structural diagram of an intelligent control system for a pallet transport shuttle vehicle according to the present invention;

[0019] Figure 2 This is the process of the intelligent control system of a pallet transport shuttle in the present invention Figure 1 ;

[0020] Figure 3 This is the process of the intelligent control system of a pallet transport shuttle in the present invention Figure 2 ;

[0021] Figure 4 This is the process of the intelligent control system of a pallet transport shuttle in the present invention Figure 3 ;

[0022] Figure 5 This is the process of the intelligent control system of a pallet transport shuttle in the present invention Figure 4 ;

[0023] Figure 6 This is the process of the intelligent control system of a pallet transport shuttle in the present invention Figure 5 . DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] Example:

[0026] See also Figures 1-6As shown, the present invention provides a technical solution: including S1: multimodal perception module, S2: dynamic path planning module, S3: multi-vehicle collaborative scheduling module, S4: energy efficiency optimization module, S5: edge computing module, S1: multimodal perception module includes S11: environmental perception submodule and S12: cargo status monitoring submodule, S11: environmental perception submodule includes lidar data preprocessing and visual sensor calibration and fusion, wherein the lidar data preprocessing adopts a statistical-based filtering algorithm to remove outliers caused by environmental interference. In the specific implementation, the neighborhood statistics of each point are calculated. If the difference between a point and its neighboring points exceeds a threshold, such as three times the standard deviation, it is determined to be a noise point. And eliminate it, and use the RANSAC algorithm to fit the ground plane, divide the point cloud into ground points and non-ground points, and improve the accuracy of ground segmentation by iteratively optimizing the plane parameters, providing a basis for subsequent obstacle detection. The visual sensor calibration and fusion adopts Zhang Zhengyou calibration method, and uses the checkerboard calibration plate to calibrate the binocular camera to obtain the internal and external parameters and distortion coefficient of the camera. According to the calibration results, the binocular image is corrected to eliminate distortion and ensure that the corresponding points are on the same horizontal line. The depth map generated by the binocular camera is fused with the lidar point cloud, taking advantage of their respective advantages, such as the texture information of the camera and the precise distance measurement of the lidar, to improve the accuracy and robustness of environmental perception;

[0027] In this embodiment, specifically: S12: The cargo status monitoring submodule includes cargo positioning and tracking and cargo status anomaly detection, wherein cargo positioning and tracking is based on the YOLOv8 model to realize real-time detection and identification of pallet cargo, and through the training data set, contains images of different angles, lighting conditions and cargo types to improve the generalization ability of the model, and adopts the DeepSORT algorithm, combined with the appearance characteristics and motion information of the cargo, to achieve stable tracking of multiple cargoes, and predicts the cargo position through Kalman filtering to reduce missed detection and false detection, wherein the cargo status anomaly detection determines whether the cargo has shifted by comparing the difference between the current position of the cargo and the preset position, and adopts the threshold comparison method. If the offset exceeds the set threshold (such as 5mm), an alarm is triggered, and according to the size information of the cargo, it is determined whether the cargo exceeds the carrying range of the pallet or shuttle, and the cargo size is measured in real time and compared with the preset threshold to achieve over-limit detection;

[0028] In this embodiment, specifically: S2: dynamic path planning module includes S21: improved A* algorithm optimization and S22: reinforcement learning model application, S21: improved A* algorithm optimization includes heuristic function design and time window constraint implementation, wherein the heuristic function design dynamically adjusts the weight parameters in the heuristic function according to the real-time status of the storage environment, learns the optimal weight combination through the reinforcement learning algorithm, improves the efficiency of path planning, integrates multiple optimization objectives into the heuristic function, realizes multi-objective optimization through weight distribution, adopts the Pareto optimal solution set, provides multiple candidate paths for the scheduling module to select, wherein the time window constraint implementation calculates the time window of each path node according to the kinematic model and path length of the shuttle, considers the start, acceleration, constant speed, deceleration and stop stages of the shuttle, ensures the accuracy of the time window, dynamically updates the time window information according to the actual operation of the shuttle, and ensures the time window coordination between multiple vehicles through real-time communication;

[0029] In this embodiment, specifically: S22: The application of the reinforcement learning model includes state space construction, action space design and reward function design, wherein the state space construction includes the current node coordinates, target point coordinates, adjacent vehicle positions, order priorities, equipment load rate, etc., and through sensor fusion and communication technology, comprehensive environmental state information is obtained, including the position, speed, acceleration, battery level, etc. of the shuttle vehicle. The vehicle status is monitored in real time through the vehicle's own sensors and battery management system, wherein the action space design defines the shuttle vehicle's movement direction, such as forward, backward, left turn, right turn, and speed level, such as low speed, medium speed, and high speed. By discretizing the action space, the decision-making process is simplified. On the basis of discrete actions, continuous speed control is introduced, and precise speed control of the shuttle vehicle is achieved through PID controller or model predictive control. wherein the reward function design designs efficiency reward items based on the time, path length and energy consumption of the shuttle vehicle to complete the task, encouraging the shuttle vehicle to choose an efficient and energy-saving path, and designs safety reward items based on the distance between the shuttle vehicle and the obstacle, collision risk, etc., punishing the unsafe behavior of the shuttle vehicle, thereby improving the safety of the system;

[0030] In this embodiment, specifically: S3: multi-vehicle collaborative scheduling module includes S31: distributed consensus algorithm deepening and S32: conflict prediction and collaborative avoidance, S31: distributed consensus algorithm deepening includes leader election and fault recovery, task allocation and load balancing, wherein leader election fault recovery adopts Raft protocol for leader election, and maintains the leader status through heartbeat mechanism. If the leader fails, a new election process is triggered, and the stability of the system is ensured by optimizing the election algorithm, such as reducing election delay and improving election success rate. When the shuttle or scheduling node fails, the continuous operation of the system is ensured through redundancy design and failover mechanism, and backup nodes and hot backup technology are used to reduce the impact of the failure on the system. Task allocation and load balancing are based on the current position of the shuttle, task priority and equipment load, and the Hungarian algorithm or auction algorithm is used for task allocation. By optimizing the task allocation strategy, the throughput and resource utilization of the system are improved, the load of the shuttle is monitored in real time, and the task allocation strategy is adjusted to achieve load balancing to avoid the situation where some shuttles are overloaded while other shuttles are idle.

[0031] In this embodiment, specifically: S32: conflict prediction and collaborative avoidance include a conflict detection algorithm and a collaborative avoidance strategy, wherein the conflict detection algorithm predicts the future trajectory of the shuttle vehicle through Kalman filtering or particle filtering, and detects potential conflict points based on the predicted trajectory, such as path intersection, too close distance, etc., obtains status information of adjacent shuttle vehicles through real-time communication, performs conflict detection and triggers collaborative avoidance based on the real-time status information, wherein the collaborative avoidance strategy adjusts the speed of the shuttle vehicle, such as slowing down or stopping, to avoid collision when a potential conflict is detected, and adopts a collaborative speed control algorithm to ensure speed coordination between multiple vehicles. In the case of severe conflict, the path replanning process is triggered, and a new conflict-free path is generated for the affected shuttle vehicle through a dynamic path planning algorithm;

[0032] In this embodiment, specifically: S4: Energy efficiency optimization module includes S41: Dynamic model deepening and S42: Energy management strategy, wherein S41: Dynamic model deepening includes motor control strategy and speed optimization algorithm, wherein the motor control strategy adopts vector control algorithm to realize independent control of motor torque and magnetic flux, and improves the response speed and stability of the motor by optimizing control parameters (such as PI controller gain). During braking, kinetic energy is converted into electrical energy and stored by energy recovery unit, and maximum power point tracking algorithm is adopted to improve energy recovery efficiency, wherein the speed optimization algorithm generates a smooth acceleration / deceleration curve through model predictive control or optimal control algorithm, reduces the impact of sudden stop and start on the motor and battery, and prolongs the life of the equipment, and adaptively adjusts the speed of the shuttle according to the slope, curvature and obstacle conditions of the path, increases the speed on flat sections to improve efficiency, and reduces the speed on ramps or near obstacles to ensure safety;

[0033] In this embodiment, specifically: S42: The energy management strategy includes a battery management system and an energy scheduling strategy, wherein the battery management system monitors the battery's voltage, current, temperature, remaining power and other status information in real time, and estimates the battery's health status and life through the BMS algorithm. Active balancing or passive balancing technology is used to balance the power of each battery cell in the battery pack to avoid overcharging or over-discharging of certain battery cells and extend the service life of the battery pack. The energy scheduling measurement reasonably allocates energy resources according to the priority and urgency of the task, ensuring that high-priority tasks obtain energy supply first, thereby improving the overall efficiency of the system. The braking energy is recovered and stored through the energy recovery unit for use by subsequent tasks. The energy management strategy is used to optimize energy use and distribution and reduce energy waste.

[0034] In this embodiment, specifically: S5: The edge computing module includes S51: Hardware configuration deepening and S52: Cloud-edge collaboration deepening, S51: Hardware configuration deepening includes AI chip selection and optimization, sensor fusion and data synchronization, wherein AI chip selection and optimization selects an AI chip suitable for the shuttle vehicle intelligent control system, such as NVIDIA Jetson AGX Orin, considering factors such as the chip's computing power, power consumption, interface and cost to ensure that the system requirements are met, and by optimizing the chip's operating frequency, voltage and temperature and other parameters, the chip's performance and stability are improved, and dynamic voltage and frequency adjustment technology is used to adjust the chip's working state according to load conditions to reduce power consumption, wherein sensor fusion and data synchronization use algorithms such as Kalman filtering, extended Kalman filtering or particle filtering to fuse data from multiple sensors, and adopt timestamp synchronization or hardware synchronization and other technologies to ensure the time consistency of multi-sensor data, and improve the accuracy of subsequent processing and analysis through the data synchronization mechanism;

[0035] In this embodiment, specifically: S52: Deepening of cloud-edge collaboration includes data synchronization and storage, model iteration and upgrade, wherein data synchronization and storage adopts protocols such as MQTT, CoAP or HTTP to realize data synchronization between the edge computing unit and the cloud, and designs different synchronization strategies according to the importance and real-time requirements of the data, such as real-time synchronization, timed synchronization or triggered synchronization, and deploys a big data platform such as Hadoop, Spark or Flink on the cloud to store and manage the synchronized data, and improves the quality and security of the data through data cleaning, deduplication, compression and encryption technologies, wherein model iteration and upgrade utilizes the big data platform on the cloud to train and optimize the operation data of the shuttle vehicle, and improves the accuracy and generalization ability of the model through deep learning, reinforcement learning or federated learning technologies, and deploys the trained model on the edge computing unit to realize real-time reasoning and decision-making of the model, and realizes remote upgrade and maintenance of the model through OTA upgrade technology, thereby improving the maintainability and scalability of the system.

[0036] Working principle or structural principle: Through the fusion perception of lidar and visual sensors, three-dimensional reconstruction and obstacle detection of the warehouse environment are realized. The lidar data is processed by statistical filtering and RANSAC ground segmentation to effectively eliminate noise points and extract the drivable area; the visual sensor is calibrated by Zhang Zhengyou's calibration method and then fused with the lidar point cloud to improve the accuracy and robustness of environmental perception. The deep learning model based on YOLOv8 detects the status of pallet cargo in real time, and combines with the DeepSORT algorithm to realize multi-cargo tracking. The threshold comparison method is used to judge cargo offset and overrun, trigger abnormal alarms, design a heuristic function for dynamic weight adjustment, integrate multi-objective optimization such as path length, energy consumption, time, etc., generate a Pareto optimal path set, and the time window constraint mechanism ensures that there is no conflict in multi-vehicle paths. Construct a state space including vehicle status, environmental status and task priority, define a discrete-continuous mixed action space, design a dual reward function of efficiency and safety, realize dynamic optimization of the path, and use Raft coordination The system elects a leader and uses the Hungarian algorithm and load balancing strategy to achieve efficient task allocation. Redundant design and failover mechanism ensure high availability of the system. Trajectory prediction and V2V communication based on Kalman filtering achieve real-time detection of conflicts. Speed ​​collaborative control and path replanning ensure safe operation of multiple vehicles. Vector control algorithm and model predictive control are used to achieve precise control of the motor and smooth adjustment of speed. The energy recovery unit recovers kinetic energy during the braking phase to improve energy efficiency. The battery status is monitored in real time through BMS. Active balancing technology is used to extend battery life. The energy scheduling strategy dynamically allocates energy resources according to task priority to reduce waste. The NVIDIA Jetson AGX Orin AI chip is deployed. Sensor fusion and data synchronization mechanism ensure real-time and accuracy of environmental perception. The MQTT protocol is used to synchronize data between the edge and the cloud. The big data platform is used on the cloud for model training and optimization. The remote upgrade of the model is achieved through OTA technology.

[0037] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent control system for a pallet transport shuttle, characterized by: It includes S1: multimodal perception module, S2: dynamic path planning module, S3: multi-vehicle collaborative scheduling module, S4: energy efficiency optimization module, and S5: edge computing module. The S1: multimodal perception module includes S11: environmental perception submodule and S12: cargo status monitoring submodule. The S11: environmental perception submodule includes lidar data preprocessing and visual sensor calibration and fusion. The lidar data preprocessing adopts a statistical-based filtering algorithm to remove outliers caused by environmental interference, and uses the RANSAC algorithm to fit the ground plane and divide the point cloud into ground points and non-ground points. The visual sensor calibration and fusion adopts Zhang Zhengyou calibration method, and uses a checkerboard calibration plate to calibrate the binocular camera to obtain the internal and external parameters and distortion coefficient of the camera, and fuses the depth map generated by the binocular camera with the lidar point cloud, taking advantage of their respective advantages to improve the accuracy and robustness of environmental perception.

2. The intelligent control system for a pallet transport shuttle according to claim 1, characterized in that: The S12: cargo status monitoring submodule includes cargo positioning and tracking and cargo status anomaly detection, wherein cargo positioning and tracking is based on the YOLOv8 model to achieve real-time detection and identification of pallet cargo, and adopts the DeepSORT algorithm, combined with the appearance characteristics and motion information of the cargo, to achieve stable tracking of multiple cargoes, wherein the cargo status anomaly detection determines whether the cargo has shifted by comparing the difference between the current position of the cargo and the preset position, and determines whether the cargo exceeds the carrying range of the pallet or shuttle vehicle based on the size information of the cargo.

3. The intelligent control system for a pallet transport shuttle according to claim 2, characterized in that: The S2: dynamic path planning module includes S21: improved A* algorithm optimization and S22: reinforcement learning model application. The S21: improved A* algorithm optimization includes heuristic function design and time window constraint implementation, wherein the heuristic function design dynamically adjusts the weight parameters in the heuristic function according to the real-time status of the storage environment, learns the optimal weight combination through the reinforcement learning algorithm, improves the efficiency of path planning, incorporates multiple optimization objectives into the heuristic function, and realizes multi-objective optimization through weight distribution, wherein the time window constraint is implemented according to the kinematic model and path length of the shuttle vehicle, calculates the time window of each path node, dynamically updates the time window information according to the actual operation of the shuttle vehicle, and ensures the time window coordination between multiple vehicles through real-time communication.

4. The intelligent control system for a pallet transport shuttle according to claim 3, characterized in that: Said S22: reinforcement learning model application includes state space construction, action space design and reward function design, wherein the state space construction includes the current node coordinates, target point coordinates, adjacent vehicle positions, order priorities, equipment load rate, etc., and obtains comprehensive environmental status information through sensor fusion and communication technology, including the position, speed, acceleration, battery power, etc. of the shuttle vehicle. The vehicle status is monitored in real time through the vehicle's own sensors and battery management system, wherein the action space design defines the shuttle vehicle's movement direction, such as forward, backward, left turn, right turn and speed level, such as low speed, medium speed, and high speed. By discretizing the action space, the decision-making process is simplified, wherein the reward function design designs efficiency reward items based on the time, path length and energy consumption of the shuttle vehicle to complete the task, and encourages the shuttle vehicle to choose an efficient and energy-saving path.

5. The intelligent control system for a pallet transport shuttle according to claim 4, characterized in that: The S3: multi-vehicle collaborative scheduling module includes S31: deepening of distributed consensus algorithm and S32: conflict prediction and collaborative avoidance. The S31: deepening of distributed consensus algorithm includes leader election and fault recovery, task allocation and load balancing. The leader election fault recovery adopts Raft protocol to elect the leader, and maintains the status of the leader through the heartbeat mechanism. If the leader fails, a new election process is triggered. When the shuttle or scheduling node fails, the continuous operation of the system is ensured through redundant design and failover mechanism. The task allocation and load balancing are based on the current position of the shuttle, task priority and equipment load. The Hungarian algorithm or auction algorithm is used to allocate tasks, the load of the shuttle is monitored in real time, and load balancing is achieved by adjusting the task allocation strategy.

6. The intelligent control system for a pallet transport shuttle according to claim 5, characterized in that: Said S32: conflict prediction and collaborative avoidance includes a conflict detection algorithm and a collaborative avoidance strategy, wherein the conflict detection algorithm predicts the future trajectory of the shuttle vehicle through Kalman filtering or particle filtering, obtains the status information of the adjacent shuttle vehicles through real-time communication, performs conflict detection and triggers collaborative avoidance based on the real-time status information, wherein when a potential conflict is detected, the collaborative avoidance strategy adjusts the speed of the shuttle vehicle, such as slowing down or stopping, to avoid collision. In the case of severe conflict, the path replanning process is triggered, and a new conflict-free path is generated for the affected shuttle vehicle through the dynamic path planning algorithm.

7. The intelligent control system for a pallet transport shuttle according to claim 6, characterized in that: The S4: Energy Efficiency Optimization module includes S41: Dynamic Model Deepening and S42: Energy Management Strategy, wherein S41: Dynamic Model Deepening includes a motor control strategy and a speed optimization algorithm, wherein the motor control strategy adopts a vector control algorithm to achieve independent control of the motor torque and magnetic flux. During braking, kinetic energy is converted into electrical energy and stored through an energy recovery unit, and a maximum power point tracking algorithm is adopted to improve energy recovery efficiency. The speed optimization algorithm generates a smooth acceleration / deceleration curve through model predictive control or optimal control algorithm, reduces the impact of sudden stops and starts on the motor and battery, extends the life of the equipment, and adaptively adjusts the speed of the shuttle according to the slope, curvature and obstacle conditions of the path.

8. The intelligent control system for a pallet transport shuttle according to claim 7, characterized in that: Said S42: energy management strategy includes a battery management system and an energy scheduling strategy, wherein the battery management system monitors the battery status information such as voltage, current, temperature and remaining power in real time, estimates the health status and life of the battery through the BMS algorithm, and adopts active balancing or passive balancing technology to balance the power of each battery cell in the battery pack, wherein the energy scheduling measurement reasonably allocates energy resources according to the priority and urgency of the task, ensures that high-priority tasks obtain energy supply first, improves the overall efficiency of the system, and recovers and stores the braking energy through the energy recovery unit for use in subsequent tasks.

9. The intelligent control system for a pallet transport shuttle according to claim 8, characterized in that: The S5: edge computing module includes S51: hardware configuration deepening and S52: cloud-edge collaboration deepening. The S51: hardware configuration deepening includes AI chip selection and optimization, sensor fusion and data synchronization. The AI ​​chip selection and optimization selects AI chips suitable for the shuttle vehicle intelligent control system, such as NVIDIA Jetson AGX Orin, considering factors such as the chip's computing power, power consumption, interface and cost to ensure that the system requirements are met. By optimizing parameters such as the chip's operating frequency, voltage and temperature, the chip's performance and stability are improved. The sensor fusion and data synchronization use algorithms such as Kalman filtering, extended Kalman filtering or particle filtering to fuse data from multiple sensors, and use technologies such as timestamp synchronization or hardware synchronization to ensure the time consistency of multi-sensor data. Through the data synchronization mechanism, the accuracy of subsequent processing and analysis is improved.

10. The intelligent control system for a pallet transport shuttle according to claim 9, characterized in that: The S52: Deepening of cloud-edge collaboration includes data synchronization and storage, model iteration and upgrade, wherein data synchronization and storage adopt protocols such as MQTT, CoAP or HTTP to realize data synchronization between the edge computing unit and the cloud, and deploy a big data platform such as Hadoop, Spark or Flink on the cloud to store and manage the synchronized data. The model iteration and upgrade uses the big data platform on the cloud to train and optimize the operation data of the shuttle vehicle, and deploy the trained model to the edge computing unit to realize real-time reasoning and decision-making of the model.

Images