Multi-vehicle cooperative transfer method based on hierarchical architecture

Abstract

The application discloses a multi-vehicle cooperative transfer method based on a layered architecture, and belongs to the technical field of multi-AGV vehicle cooperative control. The method realizes high-precision, high-synchronism and strong-robustness cooperative work of the multi-vehicles through the architecture of upper-layer global motion planning (overall motion planning and speed decomposition) and lower-layer distributed execution (single-vehicle execution and error feedback closed loop), and based on a maximum-error-first adaptive cooperative control mechanism, the method realizes the fast convergence of the cooperative overall motion with the smallest control cost by selectively adjusting the vehicle with the largest deviation degree through real-time monitoring of cooperative errors of the vehicles.

Description

Technical Field

[0001] This invention relates to the field of multi-AGV vehicle cooperative control technology, and specifically to a multi-vehicle cooperative transfer method based on a hierarchical architecture. Background Technology

[0002] In large-scale logistics warehousing, port terminals, aerospace manufacturing, and other fields, it is often necessary to move multiple Automated Guided Vehicles (AGVs) to collaboratively transport oversized or overweight goods. The carrying capacity of a single AGV is limited and cannot complete such tasks.

[0003] Current multi-AGV collaborative control methods mainly suffer from the following problems:

[0004] 1. Centralized control bottleneck: Using a single central controller to directly calculate the motion commands of each drive wheel of all vehicles results in a large amount of computation, poor real-time performance, weak system scalability, and a failure of the central controller will paralyze the entire system.

[0005] 2. Insufficient coordination: The lack of a unified global coordinate system and motion planning means that each vehicle tracks only based on its relative position, which can easily lead to problems such as asynchronous movement and internal force conflicts (mutual pushing and pulling between vehicles), resulting in unstable cargo or even overturning.

[0006] 3. Poor anti-interference capability: Traditional methods do not fully consider the execution error of a single vehicle and ground disturbances, lack a closed-loop adjustment mechanism based on real-time information feedback, and cannot guarantee the accuracy and robustness of coordinated motion.

[0007] 4. Limitations of the master-slave mode: Traditional methods use a slave vehicle adjustment strategy, which cannot accurately control the differences and deviations between vehicles, resulting in low adjustment efficiency, high energy consumption, and insufficient coordination accuracy.

[0008] 5. Lack of effective fault handling mechanism: When a part of the system fails, it can easily lead to the paralysis of the entire system or a cargo accident.

[0009] Therefore, there is an urgent need for a collaborative control method that can achieve both unified global planning and precise distributed execution, while also possessing good fault tolerance and scalability. Summary of the Invention

[0010] In view of this, the purpose of this invention is to provide a multi-vehicle cooperative transfer method based on a hierarchical architecture to solve the related problems caused by existing centralized control.

[0011] To achieve the above objectives, the method of this invention utilizes an architecture of upper-level global motion planning (overall motion planning and velocity decomposition) and lower-level distributed execution (single-vehicle execution and error feedback closed loop) to realize high-precision, high-synchronization, and robust collaborative operation among multiple vehicles. Furthermore, based on an adaptive collaborative control mechanism prioritizing maximum error, it selectively adjusts the vehicle with the greatest deviation by real-time monitoring of the collaborative errors of each vehicle, achieving rapid convergence of the overall collaborative motion with minimal control cost. Specifically, the method employs a two-layer control architecture and includes the following steps:

[0012] The upper-level control system receives global instructions for the overall motion control of multiple vehicles in coordinated motion, and performs motion trajectory planning for multiple vehicles based on the global instructions to obtain the motion trajectory planning results of multiple vehicles in the coordinated transfer process. The motion trajectory planning results of multiple vehicles are decomposed into speed control instructions for each individual vehicle. The upper-level controller sends the speed control instructions of each individual vehicle to the individual vehicle controllers corresponding to each vehicle in the lower-level control system.

[0013] The vehicle controller in the lower-level control system uses a kinematic solver to obtain control commands for each drive unit of the vehicle based on the received speed control command. The actuators of each drive unit operate according to the commands to control each drive unit and feed back the actual status information to the vehicle controller in real time. The vehicle controller adjusts the control commands for each drive unit of the vehicle based on the actual status information so that the vehicle tracks the speed control command.

[0014] During the multi-vehicle collaborative transfer process, the upper-level control system acquires the actual position and posture information of the multi-vehicles in real time. It compares the actual position and posture information of the multi-vehicles with the ideal position and posture information obtained from the motion trajectory planning results, and determines the single vehicle with the largest deviation between the actual position and posture information and the ideal position and posture information as the target vehicle. Only the speed control command of the target vehicle is adjusted to correct the result of decomposing the motion trajectory planning results of the multi-vehicles.

[0015] The beneficial effects are as follows: This invention adopts a layered architecture, with the upper layer serving as the overall planning layer and the lower layer responsible for calculating execution instructions based on the upper layer's planning. Compared to centralized control that directly calculates execution instructions for each component, this invention, based on a layered architecture, has a smaller computational load at each layer, thus ensuring computational efficiency and real-time performance. Furthermore, the clear division of labor based on the layered architecture enhances the overall system's scalability. In this invention, the upper and lower control systems not only implement their respective functions but also engage in information feedback. During multi-vehicle collaborative transfer, the upper control system obtains real-time information on the actual position and posture of the collaboratively moving vehicles through sensors installed in the lower control system. Based on this information, it adjusts control instructions to ensure global coordination. Moreover, when adjusting control instructions, it only adjusts the control instructions of the single vehicle with the largest deviation. This approach is more efficient than adjusting overall control instructions and, moreover, it focuses on adjusting the vehicle with the largest deviation, thus achieving higher efficiency. Therefore, this invention improves collaborative transfer efficiency while ensuring the coordination of the collaborative transfer process.

[0016] Furthermore, once the target vehicle is determined, in the subsequent multi-vehicle collaborative transfer process, only when the deviation of other vehicles is the greatest, and the deviation of the other single vehicle from the already determined target vehicle at the current moment differs from a fixed threshold, will the other single vehicle be determined as the new target vehicle, so that the speed control command is adjusted only for the new target vehicle.

[0017] This invention addresses the issue of frequent target vehicle switching caused by minor fluctuations in deviation when selecting a target vehicle. To avoid this, a hysteresis interval mechanism is introduced. After a target vehicle is identified, if another vehicle becomes the one with the largest deviation, it is not immediately designated as the new target vehicle. Instead, the target vehicle is switched only when its deviation reaches a certain range above that of the currently adjusted vehicle. This effectively avoids frequent switching and ensures system stability. This dynamic target switching mechanism ensures optimized allocation of adjustment resources, enabling the entire multi-vehicle system to converge quickly and smoothly to a high-precision cooperative motion state.

[0018] Furthermore, the method for determining the single vehicle with the greatest deviation between actual pose information and ideal pose information among multiple vehicles is as follows: by setting a multi-dimensional cost function including pose error penalty term and pose error change rate penalty term, the deviation value of each single vehicle among multiple vehicles is calculated according to the multi-dimensional cost function, and the single vehicle with the largest deviation value is determined as the single vehicle with the greatest deviation between actual pose information and ideal pose information among multiple vehicles.

[0019] This invention employs a multi-indicator fusion method to calculate the deviation degree of each individual vehicle, precisely quantifying the coordination error. This avoids focusing solely on the current difference while ignoring the changing trend that leads to continuous changes in the target being adjusted. Furthermore, compared to considering only the difference or the changing trend, this invention offers higher accuracy through comprehensive consideration. Therefore, by jointly considering both pose error (i.e., the difference between actual and ideal pose information) and the changing trend of pose error (i.e., the rate of change of pose error), this invention accurately identifies the vehicles that need adjustment. Specifically, the more severe the deviation of a vehicle from its ideal motion state at the current moment, the more effective the adjustment of the motion coordination of multiple vehicles can be achieved by focusing on the vehicle with the greatest negative impact on overall coordination performance.

[0020] Furthermore, the pose error penalty term includes a pose error vector obtained based on the actual pose information and the ideal pose information, and an error weight set for the pose error vector. The pose error vector includes an X-axis error component, a Y-axis error component, and a heading angle error component. The X-axis and Y-axis represent the longitudinal and lateral directions of a single vehicle in a planar coordinate system. The positive X-axis is the positive direction of vehicle movement. The positive Y-axis direction in the horizontal plane is determined according to the right-hand rule. Among the error weights, the lateral error weight used to quantify the Y-axis error component is greater than the longitudinal error weight used to quantify the X-axis error component.

[0021] When calculating the deviation value, the method of this invention not only considers the process of integrating multiple indicators, but also designs the error weight of each individual indicator. Specifically, different components of a single indicator correspond to different cost weights. Furthermore, based on the principle of suppressing internal force conflicts that may lead to cargo overturning, this invention sets the lateral error weight to be greater than the longitudinal error weight. The target vehicle determined in this way is the one that most needs to be adjusted, so as to rapidly improve the coordination of multiple vehicles by adjusting this vehicle.

[0022] Furthermore, the adjustment process for adjusting the speed control command of the target vehicle includes: determining the compensation of the speed control command of the target vehicle using an adaptive law in the form of proportional-differential based on the degree of deviation of the target vehicle; adding the control speed in the original speed control command of the target vehicle to the compensation value to obtain the control speed in the new speed control command; and issuing the new speed control command to the corresponding single-vehicle controller of the target vehicle.

[0023] The present invention utilizes the degree of deviation determined when selecting target vehicles in the adjustment process, and then directly uses this degree of deviation to obtain the corresponding compensation value, without the need to obtain other data information, thereby improving calculation efficiency, ensuring adjustment efficiency, and enabling multiple vehicles to coordinate more effectively.

[0024] Furthermore, when the actuators of each drive unit are running according to the instructions controlling each drive unit, the lower-level control system also collects force information of the superstructure components of the vehicle through a six-dimensional force sensor and a pressure sensor installed on the vehicle, and transmits the force information to the vehicle controller. The vehicle controller adjusts the control instructions for each drive unit of the vehicle according to the force information to coordinate the force control between the drive units of the vehicle.

[0025] This invention also takes into account the force coordination between various vehicle drive units. By acquiring force information in real time and making corresponding control adjustments based on the force information, force coordination control between multiple drive units can be realized in the lower-level control system, avoiding single-carrier overload or system dynamic instability.

[0026] Furthermore, the lower-level control system feeds back the force information of each individual vehicle to the upper-level control system. The upper-level control system adjusts the speed control commands of each individual vehicle based on the force information of each individual vehicle to coordinate the force control among the individual vehicles.

[0027] In this invention, not only are force information coordinated and controlled on the single vehicle side to ensure force balance among the drive units of the single vehicle, but the force information of each single vehicle is also transmitted to the upper control system so that the upper control system can adjust the force situation of each single vehicle according to the force information, thereby ensuring force balance among the single vehicles. This achieves balanced force on multiple carriers and avoids overload / underload.

[0028] Furthermore, during the multi-vehicle collaborative transfer process, the upper-level control system obtains the actual position and posture information of the multi-vehicles in real time, and calculates the formation index, including the distance variance and angle deviation between individual vehicles. If the formation index is greater than the preset safety tolerance value, a formation fault alarm is triggered.

[0029] Furthermore, during the multi-vehicle collaborative transfer process, the upper-level control system obtains the operating status of each data processing module used in the upper-level control system in real time. If the planning algorithm for motion trajectory planning times out or the communication link involving information transmission is interrupted, the upper-level control system switches to degraded mode.

[0030] Furthermore, during the multi-vehicle collaborative transfer process, the lower-level control system detects in real time whether each individual vehicle has hardware faults such as motor overload or abnormal sensor data, as well as communication faults such as communication interruption with the upper-level control system. If a hardware fault is detected, the hardware fault result is transmitted to the upper-level control system so that the upper-level control system can readjust the speed control commands for each individual vehicle. If a communication fault is detected, the lower-level control system implements degraded control based on the last speed control command received from the upper-level control system and local sensor data.

[0031] Fault detection and handling mechanisms are embedded in the layered architecture, including collaborative formation fault judgment, hardware fault tolerance, and dynamic control logic in the event of layered faults. Through real-time fault reporting, instruction backup, and task reassignment, the system can still operate safely during partial faults, enhancing the stability of the collaborative formation. By automatically detecting and correcting formation anomalies, internal force conflicts or the risk of overturning can be avoided.

[0032] The method of the present invention has the following advantages:

[0033] 1. Decoupling and scalability: The upper and lower layers have clear division of labor, with the upper layer responsible for planning and the lower layer responsible for execution, which greatly reduces the computing burden on the central controller and makes the system easy to expand to more vehicles.

[0034] 2. High synchronization and stability: Motion decomposition is performed through a unified global coordinate system, and unified compensation is carried out from the upper layer, which effectively avoids asynchronous motion between vehicles and the generation of internal forces, ensuring the smoothness of the loading process.

[0035] 3. Strong robustness: The lower layer is based on distributed closed-loop control with real-time feedback, which enables each vehicle to autonomously cope with local disturbances, improving the overall reliability and anti-interference capability of the system.

[0036] 4. Fault Tolerance: The layered architecture ensures that a single point of failure does not affect the overall system. Faults in lower layers can be reported, and tasks can be reassigned in upper layers. In the event of a failure in an upper layer, the lower layer can maintain operation for a short period based on the last instruction, resulting in high system reliability.

[0037] 5. Improved adjustment efficiency: By adopting the maximum error priority strategy, control oscillations caused by uniform adjustment are avoided, and the adjustment time is reduced by more than 30%.

[0038] 6. Energy consumption optimization: Adjustments are made only for vehicles with the largest deviations, reducing the total control energy consumption of the system.

[0039] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the system architecture used in the method of this invention;

[0041] Figure 2 This is a schematic diagram of the system structure of the upper-level control system used in the method of this invention;

[0042] Figure 3 This is a schematic diagram of the system structure of the lower-level control system used in the method of the present invention;

[0043] Figure 4 This is a schematic diagram of the error compensation process using adaptive control in the method of this invention;

[0044] Figure 5 This is a schematic diagram of the system functions used in the method of the present invention;

[0045] Figure 6 This is a schematic diagram of the emergency response mechanism of the method of the present invention. Detailed Implementation

[0046] The technical solution of the present invention will be clearly and completely described below with reference to specific embodiments. However, those skilled in the art should understand that the embodiments described below are only for illustrating the present invention and should not be regarded as limiting the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Implementation Examples of Multi-Vehicle Cooperative Transfer Method Based on Hierarchical Architecture

[0048] This invention achieves high-precision, high-synchronization, and robust collaborative operation of multiple vehicles through an architecture of upper-level global motion planning (overall motion planning and velocity decomposition) and lower-level distributed execution (single-vehicle execution and error feedback closed loop). Furthermore, it proposes an adaptive collaborative control mechanism based on prioritizing the maximum error. By monitoring the collaborative errors of each vehicle in real time, it selectively adjusts the vehicle with the largest deviation, achieving rapid convergence of the overall collaborative motion with minimal control cost.

[0049] like Figure 5As shown, the method in this embodiment is applied to a system consisting of at least two vehicles with independent driving capabilities, employing a two-layer control architecture. The upper layer receives global motion commands, performs macro-level coordination and online real-time planning to ensure the integrity and adaptability of the actions. The planning results are sent down via high real-time Ethernet to ensure low latency and high reliability of command transmission. After receiving the commands, the lower-layer controller coordinates multiple drive units based on control strategies to enhance synchronization accuracy and dynamic response capabilities. Finally, bidirectional communication between the drive units and the execution module is achieved through the CAN bus, exchanging position, speed, and torque information in real time to form closed-loop control and improve the system's anti-interference capability and stability.

[0050] like Figure 1 As shown, the method in this embodiment is based on Figure 1 The system implementation of the architecture. The upper-level control system and the lower-level control system are described below.

[0051] like Figure 2 As shown, the upper-level control system—the global coordination layer—focuses on task allocation, global planning, and collaborative decision-making, providing precise individual motion commands and collaborative strategies to the lower-level execution layer. Its core modules and functions are as follows:

[0052] 1. Human-computer interaction and task input module:

[0053] As a human-machine interface, it receives external task instructions (such as remote control operation, autonomous task planning, environmental perception data, etc.), transforms task requirements into control signals that the system can recognize, and provides input basis for the "remote control" and "navigation control" sub-modules.

[0054] 2. Remote control and navigation control submodule:

[0055] Remote control: Receives target speed commands (such as forward / backward / turning speed) from manual remote control to achieve direct speed control under manual intervention;

[0056] Navigation and control: Based on technologies such as SLAM and visual navigation, it generates target speed commands for autonomous movement, supporting path planning and autonomous execution of unmanned tasks;

[0057] Output: The two modules output the target velocity signal in parallel, which serves as the input reference for motion planning.

[0058] 3. Motion Planning Module:

[0059] Combining target speed, carrier dynamics constraints, mission time window, obstacle avoidance requirements, etc., the motion trajectory is planned (including spatial path and velocity curve design), and the phased process speed is output (time-sequential speed command for multi-carrier cooperative motion).

[0060] 4. Global Virtual Coordinate System Module:

[0061] Establish a unified virtual coordinate reference system (such as the world coordinate system or the mission coordinate system) to provide a consistent pose reference for all execution vehicles, solve the problem of position and attitude synchronization among multiple vehicles, and ensure the accurate mapping of "process velocity" in the local coordinate systems of each vehicle.

[0062] 5. Cooperative Motion Decomposition Algorithm Module:

[0063] Based on the global virtual coordinate system and process velocity, algorithms such as inverse kinematics and distributed trajectory tracking are used to decompose the overall cooperative motion into the underlying execution commands such as velocity, angle, and displacement of each individual carrier, thereby realizing the mapping of "global motion → local action".

[0064] 6. Relative pose error detection system:

[0065] Using sensors such as LiDAR, vision cameras, and IMUs, the position and attitude information of each carrier (i.e., each individual vehicle) is collected in real time, the relative pose error (position deviation, angle deviation) is calculated, and the error signal is fed back to the "synchronization and force compensation algorithm" module to correct the cooperative motion decomposition results.

[0066] 7. Synchronization and Force Compensation Algorithm Module:

[0067] By integrating the error signal from the "relative posture error detection system" with the "feedback information" from the lower-level control system, an adaptive control algorithm is adopted to calculate the synchronization correction (to eliminate synchronization errors between carriers) and the force compensation (to balance the forces on multiple carriers and avoid overload / underload). The output command of the "cooperative motion decomposition algorithm" is dynamically optimized, and finally, the speed command of a single vehicle is generated and sent to the lower execution layer.

[0068] This embodiment proposes an adaptive control method, such as... Figure 4 As shown, the algorithm is applied to the synchronization and force compensation algorithm module in this embodiment to perform synchronization compensation. The adaptive control algorithm in this embodiment includes the following steps:

[0069] 1. Quantitative calculation of cooperative error.

[0070] Input: Using a global pose detection system (such as laser SLAM, UWB positioning system, etc.), acquire in real-time the actual pose state A of each vehicle in the cluster within the global virtual coordinate system. i (A) i =(X i ,Y i ,φ i )) and its corresponding ideal pose state A i_ref (A) i_ref =(X i_ref ,Y i_ref ,φ i_ref()). Where (X,Y) represent the planar coordinates in the global virtual coordinate system, and φ represents the heading angle. In this embodiment, the positive X-axis of the global virtual coordinate system is the positive direction of vehicle movement, and the positive Y-axis direction in the horizontal plane is determined according to the right-hand rule.

[0071] Error calculation: A multi-index fusion method for calculating vehicle tracking accuracy is adopted to accurately quantify the collaborative error.

[0072] Pose error vector e i (Represents the pose error vector of the i-th vehicle): Calculate the deviation between the actual pose and the ideal pose, defined as:

[0073] e i =A i -A i_ref =[△X i △Y i △φ i ] T ;

[0074] That is, △X i =X i -X i_ref , △Y i =Y i -Y i_ref and △φ i =φ i -φ i_ref , and respectively represent the position deviation in the X direction, the position deviation in the Y direction, and the heading angle deviation.

[0075] Error change rate u i (Representing the error change rate of the i-th vehicle): The error change trend is estimated in real time through difference operations, defined as:

[0076] u i ≈(e i (t)-e i (t-△t))÷△t;

[0077] Where e i (t) represents e at time t. i Value, e i (t-△t) represents e at time t-△t. i The value is Δt, where Δt is the control period, and the error change rate is used to reflect the dynamic characteristics of the error.

[0078] Multidimensional Cost Function: To comprehensively evaluate tracking performance and reflect the importance of different error components, a quadratic cost function based on optimal control theory is defined. The core of this function lies in two designed weight matrices:

[0079] Error weight matrix Q: is a positive definite diagonal matrix, i.e., Q = diag(qx q y q φ Its diagonal elements are used to quantify the cost weights of each component of the pose error vector. For example, to suppress internal force conflicts that could lead to cargo overturning, a lateral error weight q is typically set to quantify the Y-axis error component. y Greater than the longitudinal error weight q used to quantize the X-axis error component x .

[0080] Control weight matrix R: Also a positive definite matrix, it is used to quantify the cost weight of the error rate of change, to penalize the amplitude or rate of change of the control command, so as to avoid overly drastic adjustment and ensure system stability.

[0081] Tracking accuracy of each vehicle (B) i The tracking level of the i-th vehicle (represented by ) is ultimately quantified by the following cost function:

[0082] B i =(e i T ×Q×e i +u i T ×△t×R×u i ) 0.5 ;

[0083] Among them, e i T and u i T They represent e respectively i transpose and u i Transpose, tracking degree (B) i ) is a nonnegative scalar. B i The higher the value, the more serious the deviation of the vehicle from its ideal motion state at the current moment, and the greater the negative impact on the overall cooperative performance. Therefore, in this embodiment, the tracking degree is used to represent the degree of deviation between the actual pose information and the ideal pose information of the vehicle, and corresponding adjustments can be made based on the tracking degree.

[0084] 2. Adjustment target decision based on comparison of tracking degree.

[0085] The calculated tracking degree B of each vehicle i The input vehicle tracking accuracy is compared and ranked. The core of this decision-making mechanism is to identify the unit that currently poses the greatest threat to the system's collaborative performance, namely:

[0086] k=argmax(B i );

[0087] Where k is the vehicle index identified as a key unit, i.e., the vehicle that maximizes the tracking accuracy. For example, when the tracking accuracy of the second vehicle is maximized, k=2, and the selected vehicle has a tracking accuracy of B. k =max(B1,B2,...,B n ), where n represents the number of vehicles coordinating the transfer of the same goods. The algorithm will lock this vehicle as the adjustment target for the current control cycle and perform error compensation and command adjustment processes.

[0088] 3. Selective error compensation and instruction generation.

[0089] Selective compensation: The adjustment strategy is targeted, only applying the speed command V of the key vehicle k determined above. cmdk Compensation will be provided.

[0090] V cmdk ’ =V cmdk +f(B k );

[0091] Where V cmdk ’ This indicates the new speed command after the key vehicle k is adjusted, V cmdk This indicates the original speed command of the key vehicle k before adjustment, f(B) k The compensation function is f(B). k (B) can be based on the degree of tracking of the vehicle. k And / or its error components are designed, and as one implementation method, an adaptive law in the form of proportional-derivative (PD) can be used:

[0092] f(B k )=K p ×e k +K d ×e k ′ ;

[0093] Among them, K p ,K d For the control gain matrix, e k Let e ​​represent the pose error vector of the critical vehicle k. k ′ e k The differential form of .

[0094] Command issuance: For other vehicles in the cluster, their original speed commands remain unchanged, thereby achieving efficient suppression of overall system coordination errors with minimal control cost.

[0095] 4. Dynamic monitoring and target switching.

[0096] The system continuously monitors the tracking accuracy of key vehicles. k To avoid due to B k The value fluctuates slightly around the threshold, causing the adjustment target to switch frequently, so a hysteresis interval mechanism is introduced.

[0097] The system continuously monitors the tracking accuracy of key vehicles. k When the compensation control takes effect, it makes B k The value decreases significantly, introducing a lag interval. That is, once a vehicle is selected as the adjustment target, the lag interval only applies if the tracking degree B of another vehicle is greater than that of the current target vehicle. k A fixed threshold is determined (B≥B) k The target will only switch when the threshold value is +C (where C is a set threshold). This effectively avoids high-frequency switching and ensures system stability. This dynamic target switching mechanism ensures the optimal allocation of adjustment resources, enabling the entire multi-vehicle system to converge quickly and smoothly to a high-precision cooperative motion state.

[0098] In this embodiment, the force compensation is calculated based on the force information of each individual vehicle fed back by the lower-level control system. Specifically, it determines whether overload / underload conditions exist, and in this case, calculates the force compensation and adjusts the speed control command. Specifically, in overload / underload conditions, the speed control command of each individual vehicle can be adjusted, or the vehicle with the greatest influence on stability factors can be selected for adjustment, as described above. Therefore, this embodiment integrates the force information of "relative posture error" and "upper structure feedback (force)," employs an adaptive control algorithm, calculates the synchronization correction (eliminating synchronization error between carriers) and the force compensation (balancing the force on multiple carriers and avoiding overload / underload), dynamically optimizes the output command of the "cooperative motion decomposition algorithm," and finally generates the speed command of each vehicle and sends it to the lower-level execution layer, thus achieving both posture synchronization and force balance.

[0099] like Figure 3 As shown, the lower-level control system—the distributed execution layer—focuses on command execution, local feedback, and force closed-loop regulation to achieve precise motion control and force response for a single carrier. Its core modules and functions are as follows:

[0100] 1. Single vehicle speed command reception:

[0101] The vehicle receives speed commands (including linear speed and angular speed) from the upper-level "Synchronization and Force Compensation Algorithm" module, which are used as the target input for local control.

[0102] 2. Bicycle controller:

[0103] By combining the vehicle speed command, the "local drive feedback" and the "drive unit cooperative controller" signal, and using algorithms such as PID and adaptive control, the control quantities (torque, speed, and position commands) of the drive unit are generated to achieve closed-loop control of the vehicle's motion.

[0104] 3. Single-vehicle kinematics solver:

[0105] The control input of the bicycle controller is converted into the kinematic parameters of the actuator (such as joint angle, wheel speed, and rudder deflection angle), thus completing the dynamic mapping of "control command → physical motion" and ensuring that the actuator's actions are consistent with the command.

[0106] 4. Motor driver:

[0107] It receives instructions from the vehicle's kinematics solver to drive the execution motor; it collects local drive feedback (current, speed, angle) through current sensors, encoders, and gyroscopes to form a local closed loop of "instruction → execution → feedback" and corrects drive errors in real time.

[0108] 5. Drive unit cooperative controller:

[0109] It receives force information from the "upper body feedback (force)", interacts with the vehicle controller, and realizes force coordination control among multiple drive units to avoid overload of a single carrier or instability of system dynamics.

[0110] 6. Upper structure feedback (force):

[0111] Force information of the upper components is collected by six-dimensional force sensors and pressure sensors. On the one hand, it is fed back to the upper-level "synchronization and force compensation algorithm" module to optimize global synchronization and force compensation. On the other hand, it is fed back to the "drive unit collaborative controller" at this level to optimize force coordination among multiple drive units.

[0112] 7. Local drive feedback (current, speed, angle):

[0113] The system uses acquisition units such as motor drivers and encoders to provide real-time feedback on the status parameters of the actuator, such as current, speed, and angle, to provide local status awareness for the vehicle controller and ensure the accuracy and stability of the underlying control.

[0114] Based on the functions of the upper-level control system and the lower-level control system described above, the upper-level control system and the lower-level control system in this embodiment can implement the following control methods, which are described below:

[0115] The upper-level control method includes the following steps:

[0116] S1: Construct a global virtual coordinate system: Based on task requirements, establish a two-dimensional or three-dimensional global virtual coordinate system to describe the motion of the entire collaborative system.

[0117] S2: Global Motion Command Reception and Resolution: Receiving motion commands from the target (such as velocity V along the X-axis of the global coordinate system). x (Move), which will be parsed into the desired motion velocity vector V in the global virtual coordinate system. global =[V x V y ,ω z ], where V x and V y Let ω be the linear velocity along the X-axis and along the Y-axis. z ω is the angular velocity.

[0118] S3: Calculation of desired speed for a single vehicle: Based on the real-time position of each vehicle in the global virtual coordinate system (i.e., actual pose state (X)). i ,Y i ,φ i Using a kinematic model, the desired velocity motion vector V is... global Decomposed into the speed command V of the i-th vehicle cmdi V cmdi =[V xi V yi ,ω zi ], where V xi and V yi Let ω be the linear velocity of the i-th car along the X-axis and along the Y-axis. zi Let be the angular velocity of the i-th car.

[0119] S4: Information Fusion and Compensation: Receives feedback information (such as actual position, speed, motor load, etc.) from lower-level vehicles, performs data fusion, and determines whether the overall system motion state matches the expectation. If a deviation exists, the calculated speed command V for the key vehicle k is adjusted. cmdk Dynamic compensation and adjustment are performed to achieve synchronization of vehicle movement and suppression of internal forces.

[0120] The lower-level control method operates on each vehicle controller and includes the following steps:

[0121] S5: Vehicle kinematics calculation: The vehicle controller receives the vehicle's speed command V from the upper level. cmdi Based on the vehicle's coordinate system (usually with the vehicle's center of mass as the origin) and the installation positions and angles of each drive unit (such as Mecanum wheels, steering wheels, etc.) in the vehicle coordinate system, inverse kinematics calculations are performed to obtain the desired speed commands [α1, α2, ..., α] for each drive unit of the vehicle. N ], where N represents the number of drive units.

[0122] S6: Distributed closed-loop control: The actuators of each drive unit operate according to the desired speed command and provide real-time feedback of actual speed, current and other information to the vehicle controller.

[0123] S7: Torque Adjustment and Fault Tolerance: The single-vehicle controller compares the actual and desired states of each drive unit, and uses control algorithms to independently adjust the output torque of each motor to overcome disturbances such as ground friction and gradient, ensuring that the vehicle accurately tracks the speed commands issued from the upper level. Simultaneously, the lower-level controller has preliminary fault tolerance capabilities; if a drive unit malfunctions, it can perform compensation within its capabilities and report the abnormal state to the upper level.

[0124] Based on the detailed descriptions of the upper-level control system and the lower-level control system described above in this embodiment, the following steps of the multi-vehicle collaborative transfer method are implemented in actual multi-vehicle collaborative transfer:

[0125] The upper-level control system receives global instructions for the overall motion control of multiple vehicles in coordinated movement, and performs motion trajectory planning for multiple vehicles based on the global instructions. The motion trajectory planning results of multiple vehicles in the coordinated transfer process are obtained, and the motion trajectory planning results of multiple vehicles are decomposed into speed control instructions for each individual vehicle (i.e., the speed instructions mentioned above). The upper-level controller sends the speed control instructions of each individual vehicle to the individual vehicle controllers corresponding to each vehicle in the lower-level control system.

[0126] The vehicle controller in the lower-level control system uses a kinematic solver to obtain control commands for each drive unit of the vehicle based on the received speed control commands. The actuators of each drive unit operate according to the commands to control each drive unit and feed back the actual status information to the vehicle controller in real time. The vehicle controller adjusts the control commands for each drive unit of the vehicle based on the actual status information so that the vehicle tracks the speed control commands.

[0127] During the multi-vehicle collaborative transfer process, the upper control system acquires the actual pose information of the multi-vehicles in real time, compares the actual pose information of the multi-vehicles with the ideal pose information obtained from the motion trajectory planning results, and determines the single vehicle with the largest deviation between the actual pose information and the ideal pose information as the target vehicle. Only the speed control command of the target vehicle is adjusted to correct the result of decomposing the motion trajectory planning results of the multi-vehicles.

[0128] Once the target vehicle is identified, during the subsequent multi-vehicle collaborative transfer process, only when the deviation of other vehicles is the greatest and the deviation of the other single vehicle from the already identified target vehicle at the current moment differs from that of the target vehicle by a fixed threshold will the other single vehicle be identified as the new target vehicle, so that the speed control command is adjusted only for the new target vehicle.

[0129] This embodiment of the method also incorporates an emergency response mechanism to ensure that a single point of failure does not affect overall operation. For example... Figure 6 As shown, a fault detection and handling mechanism is embedded in the layered architecture, including collaborative formation fault judgment, hardware fault tolerance, and dynamic control logic in case of layered faults; through real-time fault reporting, instruction backup, and task reassignment, the system can still operate safely in the event of partial faults; the stability of the collaborative formation is enhanced by automatically detecting and correcting formation anomalies through algorithms to avoid internal force conflicts or the risk of overturning.

[0130] Specifically, to enhance fault handling capabilities, this embodiment adds the following modules to the upper-level control system - global coordination layer:

[0131] Fault Detection and Diagnosis Module: This module monitors the operational status of each upper-layer module in real time, including whether the planning algorithm times out, whether the communication link is interrupted, and whether data consistency is abnormal. It detects faults through a heartbeat mechanism and a checksum algorithm.

[0132] Collaborative formation monitoring module: Based on vehicle position and speed data fed back from the lower layer, it calculates formation indicators (such as the variance of distance between vehicles, angle deviation, etc.). If the indicators exceed the threshold (such as the distance deviation being greater than the safety tolerance), a formation fault alarm is triggered.

[0133] Fault response strategy library: Stores predefined fault handling strategies, such as switching to degrade mode when the upper layer fails, and reassigning tasks when the lower layer fails.

[0134] The upper-level control method adds fault handling steps to the original steps (S1-S4):

[0135] S1a: Fault detection initialization: When the system starts, the fault detection module is initialized and monitoring parameters (such as heartbeat interval and fault threshold) are set.

[0136] S4a: Real-time fault detection: During information fusion and compensation, the status of upper-layer modules is detected synchronously. If a fault is detected, the fault handling process is triggered.

[0137] Add the following to the lower-level control system - distributed execution layer:

[0138] Local fault detection unit: This unit is built into each vehicle controller and is used to detect hardware faults (such as motor overload, abnormal sensor data) and communication faults (such as interruption of communication with the upper layer). Detection methods include threshold comparison (such as current exceeding limits).

[0139] Fault-tolerant execution module: When a fault is detected, this module implements degraded control (such as switching to speed hold mode) based on the last valid command and local sensor data.

[0140] Fault reporting mechanism: The lower layer reports fault information to the upper layer's fault detection and diagnosis module in real time through the communication network (wireless module).

[0141] The lower-level control method adds fault handling steps to the original steps (S5-S7):

[0142] S5a: Local Fault Monitoring: Monitors the vehicle's hardware status in real time before kinematics calculation. If a fault is detected (such as a drive unit malfunction), the fault type and severity are recorded.

[0143] S7a: Fault Response and Reporting: When comparing the actual state with the expected state, if the deviation continues to exceed the limit, fault-tolerant control is triggered (adjusting the output of the drive unit), and fault details are reported through the communication module.

[0144] This embodiment significantly improves the reliability and robustness of the multi-vehicle collaborative transfer system by integrating a multi-layered fault detection and handling mechanism. Specifically, the fault handling logic covers scenarios such as upper-layer system failures, lower-layer single-vehicle failures, abnormal collaborative formations, and hardware failures, forming a complete fault-tolerant system. In upper-layer failure scenarios, when the lower-layer system detects a crash in the upper-layer planning module or a communication interruption through a heartbeat mechanism, it immediately switches to a degraded operation mode: the lower-layer controller bases its operation on the last cached valid instructions (V... cmdi The system uses a combination of data from local sensors and timestamps for autonomous control. The vehicle controller employs a PID algorithm to track the last speed command and fine-tunes the trajectory using a relative pose error detection system to avoid collisions. Simultaneously, backup commands are stored in non-volatile memory to prevent data loss in case of power failure, ensuring continuity during short-term operation. Once an upper-level fault is repaired, the system achieves a smooth transition by comparing the new command with the cached command, avoiding abrupt changes. This design ensures that a single point of failure at the upper level does not affect the overall system, maintaining the basic execution capability of the task.

[0145] In a lower-level fault scenario, when a single vehicle experiences a hardware failure (motor overheating or jamming), the local fault detection unit identifies and categorizes the fault type in real time (minor faults allow for degraded operation, while severe faults require shutdown). It then broadcasts the fault information (including vehicle ID, type, and location data) to the upper-level system and other vehicles via the communication module. Upon receiving the information, the upper-level fault detection and diagnosis module updates the system status diagram. Simultaneously, the system performs software isolation on the faulty vehicle (e.g., disabling control commands) and triggers audible and visual alarms to prompt manual intervention.

[0146] In response to formation failures, the system uses a formation monitoring module to calculate in real time distance deviation (an alarm is triggered when the difference between the actual distance between vehicles and the expected value exceeds ±10%) and angle deviation (when the difference between the heading angle and the reference angle exceeds...). (Timely anomaly detection) promptly identifies formation disorder or internal force conflicts. Once an anomaly is detected, the upper-level synchronization and force compensation algorithm module dynamically adjusts the instructions: for distance deviations, vehicle speed is fine-tuned via PID control; for angle deviations, the steering angle is corrected using an inverse kinematics model; if adjustment fails, an emergency stop protocol is triggered, causing the formation to gradually decelerate to a stop. After the fault is cleared, the system reinitializes the formation parameters and gradually accelerates to restore stability.

[0147] In terms of hardware fault tolerance, the system employs redundancy design and degradation strategies to enhance robustness. Depending on the fault level, the system enters different degradation modes: for minor faults (such as occasional sensor errors), the vehicle speed is limited or the load continues to operate; for severe faults (such as actuator failure), the system immediately stops and reports for maintenance.

[0148] The process of adding an emergency response mechanism includes:

[0149] System initialization: The upper layer constructs a global virtual coordinate system, and each vehicle in the lower layer starts a self-check program. The fault detection module is set with a heartbeat interval of 100ms and a queuing deviation threshold of 5%.

[0150] Normal operation: The upper layer issues commands, and the lower layer tracks the movement. The coordinated formation monitoring module calculates the distance between vehicles in real time to ensure formation stability.

[0151] Fault Occurrence: If a bicycle motor malfunctions, the local fault detection unit identifies the fault and reports it via the network. The upper layer receives the information and makes a judgment.

[0152] Fault recovery: After the faulty vehicle is repaired, it is reinstated into the system, and the formation is smoothly restored.

[0153] Overall, the aforementioned fault handling mechanisms achieve high fault tolerance: the layered architecture ensures that a single point of failure does not affect the overall system; when an upper layer fails, the lower layer degrades to maintain task continuity; and when a lower layer fails, resources are dynamically optimized through real-time reporting and task reallocation. Cooperative formation stability is enhanced through real-time monitoring and automatic adjustment to prevent internal conflicts. Hardware robustness is significantly improved through redundancy and degradation strategies. Simultaneously, the modular design makes the system easily scalable to large-scale vehicle applications. These advantages collectively guarantee the system's reliability and efficiency in complex industrial environments, providing key technological support for intelligent logistics and intelligent manufacturing.

[0154] The method in this embodiment has the following significant advantages:

[0155] 1. Decoupling and scalability: The upper and lower layers have clear division of labor, with the upper layer responsible for planning and the lower layer responsible for execution, which greatly reduces the computing burden on the central controller and makes the system easy to expand to more vehicles.

[0156] 2. High synchronization and stability: Motion decomposition is performed through a unified global coordinate system, and unified compensation is carried out from the upper layer, which effectively avoids asynchronous motion between vehicles and the generation of internal forces, ensuring the smoothness of the loading process.

[0157] 3. Strong robustness: The lower layer is based on distributed closed-loop control with real-time feedback, which enables each vehicle to autonomously cope with local disturbances, improving the overall reliability and anti-interference capability of the system.

[0158] 4. Fault Tolerance: The layered architecture ensures that a single point of failure does not affect the overall system. Faults in lower layers can be reported, and tasks can be reassigned in upper layers. In the event of a failure in an upper layer, the lower layer can maintain operation for a short period based on the last instruction, resulting in high system reliability.

[0159] 5. Improved adjustment efficiency: By adopting the maximum error priority strategy, control oscillations caused by uniform adjustment are avoided, and the adjustment time is reduced by more than 30%.

[0160] 6. Energy consumption optimization: Adjustments are made only for vehicles with the largest deviations, reducing the total control energy consumption of the system.

[0161] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A multi-vehicle collaborative transfer method based on a hierarchical architecture, characterized in that, The method employs a two-layer control architecture and includes the following steps: The upper-level control system receives a global command for the overall motion control of multiple vehicles in coordinated motion. The global command is a target speed command. Based on the global command, the upper-level controller performs motion trajectory planning for multiple vehicles to obtain the motion trajectory planning result of multiple vehicles in the coordinated transfer process. The motion trajectory planning result is a time-sequential speed command for the coordinated motion of multiple carriers. The motion trajectory planning result of multiple vehicles is decomposed into speed control commands for each individual vehicle. The upper-level controller sends the speed control commands for each individual vehicle to the individual vehicle controllers corresponding to each vehicle in the lower-level control system. The vehicle controller in the lower-level control system uses a kinematic solver to obtain control commands for each drive unit of the vehicle based on the received speed control command. The actuators of each drive unit operate according to the commands to control each drive unit and feed back the actual status information to the vehicle controller in real time. The vehicle controller adjusts the control commands for each drive unit of the vehicle based on the actual status information so that the vehicle tracks the speed control command. During multi-vehicle collaborative transfer, the upper-level control system acquires the actual pose information of the collaboratively moving vehicles in real time. It compares the actual pose information with the ideal pose information obtained from the motion trajectory planning results, and determines the single vehicle with the largest deviation between the actual pose information and the ideal pose information as the target vehicle. Only the speed control command of the target vehicle is adjusted to correct the result of decomposing the motion trajectory planning results of the multi-vehicles. Once the target vehicle is determined, in the subsequent multi-vehicle collaborative transfer process, only when the deviation of other single vehicles is the largest, and the deviation of other single vehicles from the already determined target vehicle at the current moment differs from that of the target vehicle by a fixed threshold, will the other single vehicle be determined as the new target vehicle, so that the speed control command of the new target vehicle will be adjusted only.

2. The multi-vehicle cooperative transfer method according to claim 1, characterized in that, The method for determining the single vehicle with the greatest deviation between actual pose information and ideal pose information among multiple vehicles is as follows: by setting a multi-dimensional cost function including pose error penalty term and pose error change rate penalty term, the deviation value of each single vehicle in the multiple vehicles is calculated according to the multi-dimensional cost function, and the single vehicle with the largest deviation value is determined as the single vehicle with the greatest deviation between actual pose information and ideal pose information among multiple vehicles.

3. The multi-vehicle cooperative transfer method according to claim 2, characterized in that, The pose error penalty term includes a pose error vector obtained based on the actual pose information and the ideal pose information, and error weights set for the pose error vector. The pose error vector includes an X-axis error component, a Y-axis error component, and a heading angle error component. The X-axis and Y-axis represent the longitudinal and lateral directions of a single vehicle in a planar coordinate system. The positive X-axis is the positive direction of vehicle movement. The positive Y-axis direction in the horizontal plane is determined according to the right-hand rule. Among the error weights, the lateral error weight used to quantify the Y-axis error component is greater than the longitudinal error weight used to quantify the X-axis error component.

4. The multi-vehicle cooperative transfer method according to claim 1 or 3, characterized in that, The adjustment process for adjusting the speed control command of the target vehicle includes: determining the compensation of the speed control command of the target vehicle using an adaptive law in the form of proportional-differential based on the degree of deviation of the target vehicle; adding the control speed in the original speed control command of the target vehicle to the compensation value to obtain the control speed in the new speed control command; and sending the new speed control command to the corresponding single-vehicle controller of the target vehicle.

5. The multi-vehicle cooperative transfer method according to claim 1, characterized in that, When the actuators of each drive unit run according to the instructions controlling each drive unit, the lower-level control system also collects force information of the superstructure components of the vehicle through a six-dimensional force sensor and a pressure sensor installed on the vehicle, and transmits the force information to the vehicle controller. The vehicle controller adjusts the control instructions for each drive unit of the vehicle according to the force information to coordinate the force control between the drive units of the vehicle.

6. The multi-vehicle cooperative transfer method according to claim 5, characterized in that, The lower-level control system feeds back the force information of each individual vehicle to the upper-level control system. The upper-level control system adjusts the speed control commands of each individual vehicle based on the force information of each individual vehicle to coordinate the force control among the individual vehicles.

7. The multi-vehicle cooperative transfer method according to claim 1, characterized in that, During the multi-vehicle collaborative transfer process, the upper-level control system obtains the actual position and posture information of the multi-vehicles in real time and calculates the formation index, which includes the distance variance and angle deviation between individual vehicles. If the formation index is greater than the preset safety tolerance value, a formation fault alarm is triggered.

8. The multi-vehicle cooperative transfer method according to claim 7, characterized in that, During the multi-vehicle collaborative transfer process, the upper-level control system obtains the operating status of each data processing module used in the upper-level control system in real time. If the planning algorithm for motion trajectory planning times out or the communication link involving information transmission is interrupted, the upper-level control system switches to degraded mode.

9. The multi-vehicle cooperative transfer method according to claim 8, characterized in that, During multi-vehicle collaborative transfer, the lower-level control system detects in real time whether each vehicle has hardware faults such as motor overload or abnormal sensor data, as well as communication faults such as communication interruption with the upper-level control system. If a hardware fault is detected, the hardware fault result is transmitted to the upper-level control system so that the upper-level control system can readjust the speed control commands for each vehicle. If a communication fault is detected, the lower-level control system implements degraded control based on the last speed control command received from the upper-level control system and local sensor data.

Images