An AMR robot multi-machine cooperative scheduling method, device and storage medium

Abstract

The application discloses an AMR robot multi-machine cooperative scheduling method and device and a storage medium. The method comprises the following steps: receiving a production order and disassembling the production order into an action sequence; calling a first group of physical parameters containing a basic envelope radius R1 and no-load kinematics parameters to generate a space-time trajectory in the stage of acquiring a trolley under no load; generating a second group of physical parameters containing an extended envelope radius R2 and load kinematics parameters in the stage of carrying a load; establishing a four-dimensional space-time graph matrix and setting a delay penalty weight; using a mixed integer linear programming model to allocate a time window for a high-weight production order and mark an extended occupied area; and if a time overlap is encountered when planning a path for a low-weight production order, increasing a cost penalty and controlling to wait and avoid in a safe area. The application can accurately describe the space and speed changes before and after the docking trolley, eliminate the multi-machine bottom interlocking hidden danger, and improve the logistics throughput.

Description

Technical Field

[0001] This application relates to the fields of computer application and enterprise resource management scheduling technology, and in particular to a method, device and storage medium for multi-machine collaborative scheduling of AMR robots. Background Technology

[0002] Flexible manufacturing workshops are increasingly demanding more automated logistics solutions. In complex logistics management systems that include multiple holstered autonomous mobile robots and multiple unpowered assembly trolleys, multi-machine collaborative scheduling technology is the core to ensure the continuity of workshop production. Traditional workshop logistics scheduling systems typically use a unified mass point model to calculate road network costs and reserve time and space resources when handling collaborative handling tasks between holstered autonomous mobile robots and trolleys.

[0003] However, scheduling schemes based on a single-mass model or fixed-size boundaries have significant physical applicability flaws. This calculation method ignores the substantial physical differences in the spatial occupancy and underlying kinematic characteristics of autonomous mobile robots in their unloaded state and under load state after being lifted by the trolley. The system assigns tasks to the equipment uniformly, without reserving differentiated resources for the dynamic trajectory of the autonomous mobile robot combined with the trolley in the global spatiotemporal dimension. Because the overall physical shape expands when the trolley is fully loaded, increasing the turning radius, and the increased mass prolongs the acceleration and deceleration time of the chassis, multiple autonomous mobile robots with loaded trolleys are prone to encountering each other at intersections or narrow passageways when the workshop logistics scheduling task density is high. Since the passageways reserved by the upper-level scheduling system are too narrow, the underlying obstacle avoidance module will inevitably trigger frequent emergency stops. Frequent stops of the underlying hardware will cause the actual travel time to deviate from the time window pre-planned by the scheduling system, leading to the failure of subsequent work order allocation plans, causing cascading delays in workshop logistics, and even physical multi-machine deadlocks. The existing dispatch system lacks a joint modeling mechanism for the urgency of workshop work orders and the real-time physical dimensions of equipment, making it impossible to perform work order-level task reorganization and resource exchange operations before the risk of congestion occurs. Summary of the Invention

[0004] This application provides a multi-robot collaborative scheduling method, device and storage medium for AMR robots, which solves the technical problem that existing scheduling schemes fail to distinguish the physical changes before and after the trolley is combined, resulting in inaccurate time window prediction and logistics congestion in high-density scenarios.

[0005] In a first aspect, embodiments of this application provide a method for multi-robot collaborative scheduling of autonomous mobile robots, applied to a logistics system comprising multiple autonomous mobile robots and assembly trolleys. The method includes:

[0006] Receive production work orders and break them down into action sequences that include the unloaded trolley acquisition stage and the load transportation stage.

[0007] During the no-load acquisition of the trolley, the first set of physical parameters, including the basic envelope radius R1 and the no-load kinematic parameters, is called to generate the spatiotemporal trajectory.

[0008] During the load transport phase, the dimensions of the assembly trolley are obtained and projected onto the dimensions of the autonomous mobile robot using a three-dimensional Boolean union, generating a second set of physical parameters including the extended envelope radius R2 and load kinematic parameters. .

[0009] Establish a four-dimensional spatiotemporal graph matrix, and calculate the delay penalty coefficient based on the delivery deadline of the production work order to set the weight.

[0010] A mixed-integer linear programming model is used to prioritize the allocation of time windows to production work orders with high weights. During this period, the extended occupied area is marked in the four-dimensional spatiotemporal graph matrix based on the second set of physical parameters. The occupation duration of the extended occupied area is calculated based on the preset node distance and load kinematic parameters.

[0011] When planning a path for a production order with low weight, if the corresponding envelope overlaps with the marked extended occupied area in time, the cost penalty value of the corresponding path is increased, and the corresponding autonomous mobile robot is controlled to wait and avoid in a safe area.

[0012] In one embodiment, the method includes: in response to receiving a highest priority order, iterating through the estimated arrival times of each owner mobile robot in the idle trolley acquisition phase; issuing a task change frame carrying new coordinates and trolley verification codes based on a verified cutoff time window, and returning the original task to the scheduling pool.

[0013] In one embodiment, the unloaded kinematic parameters include the unloaded driving speed and the unloaded angular velocity; the loaded kinematic parameters include the loaded driving speed and the loaded angular velocity after deceleration.

[0014] In one embodiment, the occupation duration of the extended occupation area is configured to be calculated by: obtaining the quotient of the physical distance between preset nodes and the load driving speed; and adding the quotient to the rotation buffer time set based on the extended envelope radius R2 to obtain the occupation duration.

[0015] In one embodiment, the rotation buffer time is configured as the quotient of the rotation angle and the load angular velocity; the spatial coordinate system in the four-dimensional spatiotemporal graph matrix is ​​configured to extend the extended envelope radius R2 outward by a preset safety margin distance to define the physically impassable boundary.

[0016] In one embodiment, a mixed-integer linear programming model is used to prioritize the allocation of time windows to production work orders with high weights, including: configuring mutual exclusion constraints for time windows and non-overlapping constraints for space occupancy; and calculating the available time window sequence based on the mutual exclusion constraints for time windows and non-overlapping constraints for space occupancy.

[0017] In one embodiment, controlling the corresponding autonomous mobile robot to wait and avoid obstacles in a safe area includes: backtracking to the nearest non-overlapping node in the four-dimensional spatiotemporal graph matrix as the safe area; generating a waiting time window in the safe area and maintaining the waiting state until the extended occupied area corresponding to the production work order with high weight is released.

[0018] In one embodiment, the cost penalty value for the corresponding path is configured to be positively adjusted according to the length of the waiting time window; in response to the cost penalty value exceeding a preset penalty threshold, an alternative path to avoid the expanded occupied area is replanned.

[0019] Secondly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory. When the processor executes the program, it implements the autonomous mobile robot multi-machine cooperative scheduling method described in the first aspect or any embodiment of the first aspect.

[0020] Thirdly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the autonomous mobile robot multi-machine cooperative scheduling method described in the first aspect or any embodiment of the first aspect.

[0021] The beneficial effects of this application are as follows: By introducing a dual-state polygonal spatiotemporal envelope model for equipment form state switching, the system accurately describes the spatial occupancy and speed time changes before and after the autonomous mobile robot docks with the trolley at the underlying physical level; combined with the above-mentioned spatiotemporal resource reservation mechanism and work order weight ranking, it realizes high-precision feedback support from the mixed integer linear programming model to the high-level task prediction, eliminating the hidden danger of multiple load trolleys interlocking at the underlying obstacle avoidance level in narrow areas; in addition, relying on the dynamic scheduling algorithm based on truncation logic, it can ensure the timely fulfillment of emergency production processes and improve the overall logistics throughput of the workshop. Attached Figure Description

[0022] Figure 1 This is a flowchart of a multi-machine collaborative scheduling method for autonomous mobile robots provided in an embodiment of the present invention.

[0023] Figure 2 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0024] The specific embodiments of this application will now be described in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustrating and explaining the technical solutions of this application and do not limit the scope of protection of this application.

[0025] It should be noted that the application scenario involved in this application includes a manufacturing execution system server, a central scheduling server, a wireless communication base station, and multiple autonomous mobile robots deployed in a flexible manufacturing workshop. The workshop contains multiple storage areas, processing stations, and multiple assembly trolleys. The central scheduling server, as the core processing node, maintains a data synchronization connection with the manufacturing execution system server via an industrial Ethernet network to receive production work orders. Each autonomous mobile robot's controller is equipped with a wireless network communication module, enabling it to periodically report its current two-dimensional physical coordinates, battery level, and whether it is currently in a lifting state to the central scheduling server via the wireless communication base station. The assembly trolley, as a passive carrier, has a globally unique RFID tag fixed to its bottom for equipment identification and docking. The central scheduling server runs a business logic module and a path calculation module.

[0026] Figure 1 This is a flowchart of a multi-robot collaborative scheduling method for autonomous mobile robots provided in an embodiment of the present invention. The main execution entity of this method is the processor in the aforementioned central scheduling server. Figure 1 As shown, this method specifically includes the following logical steps:

[0027] Receive production work orders and break them down into action sequences S101. Receive production work orders and break them down into action sequences that include the empty trolley acquisition stage and the load transportation stage.

[0028] The production work order issued by the Manufacturing Execution System (MES) includes the target material number, starting workstation coordinates, target workstation coordinates, and a specified delivery deadline. The business logic module of the central scheduling server is responsible for breaking down the macro-level production work order into detailed equipment action steps. The specific action sequence after decomposition strictly follows a sequential relationship, including: acquiring idle equipment, assigning it to the trolley coordinates, controlling the lifting of the trolley, transporting the trolley to the target workstation, and lowering the trolley. This sequence is logically divided into two core physical stages: the empty trolley acquisition stage before lifting and the loaded transport stage after lifting. This decomposition provides a timeline boundary for subsequent polymorphic modeling.

[0029] To ensure a closed-loop hardware sequence during action execution, the central scheduling server's underlying communication interface is equipped with a real-time clock synchronization module and packet loss retransmission compensation logic. The real-time clock synchronization module employs a precise time protocol to ensure that the clock deviation between the central scheduling server and multiple chassis controllers is within milliseconds, thus avoiding spatiotemporal map prediction errors caused by network latency. Based on the retransmission compensation logic, when consecutive heartbeat packets or status feedback frames are detected as being lost, the system triggers predictive calculations and initiates a handshake reconnection. By employing high-precision clock synchronization and packet loss retransmission compensation, the consistency of the timestamp reference for multi-machine collaboration can be ensured, reducing the risk of physical collisions.

[0030] In the no-load acquisition trolley stage, the spatiotemporal trajectory S102 is generated by calling the first set of physical parameters, which includes the basic envelope radius R1 and the no-load kinematic parameters.

[0031] When the business logic module issues an unloaded trolley instruction, the path calculation module extracts the physical dimension parameters of the equipment's factory design. The system constructs the equipment's top-view outer contour into a basic polygon envelope and extracts the circumcircle of this envelope to determine the collision detection boundary. During this process, the radius of the circumcircle of the basic polygon is set as the basic envelope radius R1. At the dynamics setting level, the system assigns the first type of kinematic parameter matrix to the current stage, namely the unloaded kinematic parameters. These unloaded kinematic parameters specifically include the unloaded travel speed and the unloaded angular velocity. According to the global parameter dictionary mapping in the central scheduling server, in this embodiment, the first type of kinematic parameter—maximum travel speed—is strictly set to 1.5 m / s, and the first type of kinematic parameter—maximum angular velocity—is set to 0.8 rad / s. In one embodiment, the system calculates the time taken for the spatiotemporal trajectory during the unloaded stage based on the first set of physical parameters using the following formula:

[0032]

[0033] The system performs an accumulation division operation, where, S represents the idle prediction time variable between node i and node j. ij Represents the straight-line physical distance between nodes. This indicates that the maximum driving speed is set to 1.5 m / s. Indicates the relative deflection angle between nodes. This indicates a maximum angular velocity of 0.8 rad / s. It should be noted that the above parameters are only typical examples of equipment calibration values; those skilled in the art can adjust them according to the specific motor model, as long as they reflect the high mobility characteristics under no-load conditions. By using R1 calculated for dimensions and the first type of kinematic parameters, the equipment's mobility when unloaded can be utilized to shorten the time spent on no-load runs.

[0034] During the load transport phase, a second set of physical parameters S103, including the extended envelope radius, is generated. In this phase, the dimensions of the assembly trolley are obtained and projected in a three-dimensional Boolean union with the dimensions of the autonomous mobile robot to generate a second set of physical parameters, including the extended envelope radius R2 and load kinematic parameters. .

[0035] When the business logic module issues an instruction for the trolley to proceed to the target workstation, a physical change occurs. The system acquires target vehicle information through the RFID reader hardware interface and signal decoding module. An RFID antenna is mounted on the top of the chassis. When aligned with the trolley chassis, it scans and reads the electronic tag containing a globally unique identifier fixed to the bottom of the trolley, decoding and extracting the specific trolley model code. The central dispatch server then queries the system database using this model code to retrieve pre-stored spatial dimension data for that model of trolley, including its length, width, height, and the geometric coordinates of its bottom lifting interface.

[0036] After obtaining the dimensions, the system performs a Boolean union operation on the acquired trolley outline and its own outline in a 3D coordinate system, fusing them into an irregular 3D polygonal entity. Next, this 3D polygonal entity is orthogonally projected along the Z-axis onto the XY 2D plane to generate an extended polygonal envelope. The system calculates the circumcircle of this extended polygonal envelope and sets its radius as the extended envelope radius R2. Based on the physical law that geometric superposition inevitably leads to outward expansion, the system has a rigid radius constraint that must be satisfied. .

[0037] As the overall system mass increases and the center of gravity rises under load, the system synchronously assigns a second type of kinematic parameter matrix to the device, namely the load kinematic parameters. These parameters include the load-driven speed and load angular velocity after deceleration. In this embodiment, the second type of kinematic parameter—maximum driving speed—is reduced to 0.8 m / s, and the second type of kinematic parameter—maximum angular velocity—is reduced to 0.3 rad / s. At the chassis drive execution level, these parameters are directly sent to the chassis motor drive and inverse kinematics module, which converts the linear velocity and angular velocity commands into specific duty cycle control signals for the bottom-level left and right drive wheels. By using the logic of generating an extended envelope radius R2 through three-dimensional Boolean union two-dimensional projection and reducing the speed parameters, the system can accurately grasp the actual footprint and long-tail inertial characteristics of the merged body, providing low-level data warnings for corner scrapes that are prone to occur in the subsequent road network.

[0038] Establish a four-dimensional spatiotemporal graph matrix and set the delay penalty weight S104. Calculate the delay penalty coefficient based on the delivery deadline of the production work order to set the weight.

[0039] In the task matching and right-of-way planning stages, the central scheduling server establishes a four-dimensional spatiotemporal graph matrix composed of three-dimensional spatial coordinates and a time axis. The system reads the current system timestamp in real time, calculates the difference between the delivery deadline and the system timestamp as the delay margin, and uses the delay penalty coefficient calculated in reverse based on this margin as a weight parameter. In one embodiment, the quantification of the weight is achieved through the following formula:

[0040]

[0041] The system performs exponentially decaying gain calculation, where W k This represents the overall weight variable for the k-th production order. This indicates the delivery deadline for work order k. Represents the current system absolute time. This represents the penalty amplification gain coefficient. This represents the time sensitivity adjustment factor. This represents the basic priority constant. As the time margin decreases, the exponential term tends to amplify, causing the weight W to... k Increase.

[0042] The mixed-integer linear programming model is used to allocate time windows and mark the extended occupied area S105. The mixed-integer linear programming model prioritizes the allocation of time windows for production work orders with high weights. During this period, the extended occupied area is marked in the four-dimensional spatiotemporal graph matrix based on the second set of physical parameters. The occupation duration of the extended occupied area is calculated based on the preset node distance and load kinematic parameters.

[0043] The system substitutes the calculated weights into the objective function of the mixed-integer linear programming model. For the top N priority work orders by weight, the system prioritizes locking the time segments passing through nodes along the time axis in the topological network of the four-dimensional spatiotemporal graph matrix. When planning the path during the load carrying phase, the system forcibly marks the spatiotemporal graph nodes within the corresponding time slice with an extended occupied area marker with an extended envelope radius R2.

[0044] The calculation process for the occupancy duration of the extended occupied area is as follows: obtain the physical distance between preset nodes; calculate the quotient between this physical distance and the load driving speed. The system then adds this quotient to the rotation buffer time set based on the extended envelope radius R2 to obtain the actual occupancy duration of the road segment.

[0045] In one embodiment, the logic for deriving the duration of occupation based on the second set of parameters is implemented through the following formula:

[0046]

[0047] The system performs calculations by dividing the physical distance by the velocity and the rotation angle by the angular velocity, and then accumulates these calculations. D represents the time axis length of the expanded occupied area, and D represents the linear physical distance between nodes. This represents the driving speed corresponding to the aforementioned load of 0.8 m / s. This indicates the path deflection and rotation angle of the node. This represents the load angular velocity corresponding to the aforementioned 0.3 rad / s. The second term in the formula is the rotational buffer time set based on the R2 volume.

[0048] To ensure the safety of spatiotemporal reservation, the core mechanism for allocating time windows using a mixed-integer linear programming model includes two fundamental constraints: mutual exclusion constraints for time windows and non-overlapping constraints for spatial occupancy. Based on these constraints, a sequence of available time windows is calculated. The mutual exclusion constraint for time windows is configured to allow only a single autonomous mobile robot with an occupancy marker to exist within the same physical grid node of the same timestamp slice. The non-overlapping constraint for spatial occupancy is configured to extend the spatial coordinate system in the four-dimensional spatiotemporal graph matrix outwards by a preset safety margin distance from the extended envelope radius R2 to define the physically impassable boundary, ensuring that the sphere of influence of the extended occupancy area constructed based on R2 has no geometric intersection with the real-time envelopes of other autonomous mobile robots in adjacent areas. By employing a mechanism that prioritizes combining the spatiotemporal imprint of the extended envelope size with the mixed-integer linear programming model, the large volume and low-speed operation characteristics of high-priority tasks can be solidified into a dynamic projection on the global map, cutting off the possibility of interference from subsequent scheduling.

[0049] When there is a time overlap in the path, the cost penalty is increased and the robot waits in the safe area to avoid the obstacle. S106 When planning a path for a production order with low weight, if the corresponding envelope overlaps with the marked extended occupied area in time, the cost penalty value of the corresponding path is increased, and the corresponding autonomous mobile robot is controlled to wait in the safe area to avoid the obstacle.

[0050] When the system begins path searching for ordinary work orders with lower weight rankings, the path calculation algorithm performs a collision scan. If it finds that the generated envelope of a trial path intrudes into the extended occupied area left by a previous high-weight task in a time slice, the system immediately increases the cost penalty value of that trial path branch in the graph search algorithm.

[0051] When it is determined that the corresponding autonomous mobile robot needs to wait in a safe area to avoid a collision, the system searches along the negative direction of the time axis and the backtracking direction of the topology link in the four-dimensional spatiotemporal graph matrix to find the blank node closest to the current collision point that has no overlapping spatiotemporal markers as the safe area. A waiting time window is generated at this safe area node, and the device with the low-weight task is driven into the point and kept in a brake-locked waiting state. The system continues to monitor until the large extended area occupied by the high-weight production work order is cleared and the relevant right-of-way is released.

[0052] The cost penalty value for corresponding paths is configured using a progressive mechanism: the system positively adjusts this cost penalty value based on the length of the waiting time window required at the original conflict location. If the calculation finds that the waiting time is too long, causing the accumulated cost penalty value for the path to exceed the preset penalty threshold set internally by the system, the system will trigger a rerouting mechanism, directly abandoning the waiting at this location and replanning an alternative detour path that bypasses the corresponding road segment of the expanded occupied area. By adopting a backtracking mechanism that applies cost penalties on overlapping time windows and forcibly guides low-priority tasks back to safe areas to wait, the system uses static waiting in space to exchange for smooth passage of high-priority, heavily loaded vehicles, thus preventing the underlying deadlock loop that is prone to occur when multiple vehicles meet.

[0053] Based on the aforementioned mainline scheduling scheme, this application relates to an emergency scheduling variant for sudden production situations in the workshop. In another embodiment, the system has the capability to forcibly reorganize interrupted orders. In response to the Manufacturing Execution System (MES) issuing an emergency interrupted order task with the highest priority to the central scheduling server, the scheduling system immediately suspends the regular queuing sequence and initiates dynamic preemption logic. The system iterates through the estimated arrival times of each of the current idle mobile robots that are in the idle trolley acquisition stage and are en route. It extracts the idle device closest to the emergency trolley and calculates a new time window to truncate its current task and reassign it to the trolley required for the emergency interrupted order. If, in the spatiotemporal graph matrix, the calculation result of this new time window indicates that it satisfies the interrupted order response time and does not cause global congestion, the system will generate a task change frame. The central scheduling server issues a verified task change frame to the underlying controller of the target device via a wireless network. The data structure of this frame message carries the new physical two-dimensional coordinates of the target trolley and the corresponding RFID code verification information of the trolley. Once confirmed, the previously bound lower-priority work order will be removed from the central scheduling server and returned to the scheduling pool to await reassignment. By adopting the interception of the highest-priority inserted order and the mechanism of truncating and reorganizing idle equipment, the rigid logic of task binding can be broken. Under the premise of ensuring that no secondary congestion is caused, highly mobile idle resources can be used to respond to emergency production gaps, thereby enhancing the flexible workshop's ability to handle emergencies agilely.

[0054] Figure 2 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Figure 2 As shown, the electronic device 200 in this embodiment includes a processor 201, a memory 202, and a communication interface 203. The components can be electrically and data connected through an internal communication bus 204.

[0055] The processor 201 is the control core of the electronic device and can be configured as a central processing unit, image processor, application-specific integrated circuit, or programmable logic device. The memory 202 stores operating system processes, network communication protocol stacks, and computer programs and configuration files for the aforementioned envelope radius calculations, parameter matrix calibrations, and mixed-integer linear programming models. The memory 202 can include high-speed random access memory or non-volatile memory, such as flash memory. The communication interface 203 includes a wired Ethernet network card chip and a wireless communication transceiver antenna module, responsible for exchanging instruction messages with external devices such as the manufacturing execution system and the underlying controller. In this embodiment, the processor 201 is configured to call and execute the computer program stored in the memory 202 and use the communication interface 203 to send and receive data, thereby achieving the functions described above in this application. Figure 1 The diagram illustrates all the logical control and deduction steps of the multi-robot collaborative scheduling method for autonomous mobile robots. By constructing the scheduling hub using server hardware integrating processors and memory, it can provide parallel computational power support for four-dimensional spatiotemporal graphs, ensuring the real-time output of planning results under high-density workshop road networks.

[0056] In one embodiment, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the multi-machine cooperative scheduling method for autonomous mobile robots described in any of the above embodiments. Those skilled in the art will understand that implementing all or part of the processes in the methods of the above embodiments can be accomplished by a computer program instructing related hardware. The computer program can be stored in a non-volatile or volatile computer-readable storage medium, and when executed, it can include the processes of the above method embodiments.

[0057] Based on the above system architecture and detailed steps, the working principle of this application is as follows: When the system receives a logistics work order, it first divides it into two states on the time axis: idle and loaded. In the idle phase, the equipment exhibits small and fast characteristics; after confirming docking and lifting the trolley via RFID, the system automatically generates a large and slow-moving extended polygon envelope based on material superposition and fusion. The central scheduling server projects these two distinct physical forms onto a four-dimensional spatiotemporal graph matrix composed of three-dimensional space and a time axis, and sorts the tasks by weight calculated using delay penalties. High-weight tasks, based on their expanded R² size and long latency due to slow speed, occupy a large area of ​​the spatiotemporal continuum within the spatiotemporal graph; once the computational probe of a low-weight task touches this occupied area on the time axis, its cost skyrockets, forcing it to find the nearest safe node to brake and wait or detour. In this way, the entire method achieves a seamless mirror mapping of the heterogeneous physical form of the equipment to the top-level mathematical operations planning model, ensuring the accurate and consistent execution of scheduling.

[0058] It should be noted that the various specific equipment models, data sizes, and numerical equations mentioned in this specification are merely illustrative examples to aid in explaining the inventive concept of this application. Those skilled in the art can use similar equivalent algorithm rules and boundary thresholds to make alternative settings based on actual engineering conditions such as the friction coefficient of the workshop floor and the tonnage of the trolley. These alternative solutions can also achieve the scheduling purpose of dynamically reserving right-of-way based on morphological changes in this application, and therefore should all be included within the protection scope of this application.

Claims

1. A multi-robot collaborative scheduling method for AMR robots, applied to a logistics system comprising multiple autonomous mobile robots and assembly trolleys, characterized in that, include: Receive production work orders and break them down into action sequences that include an empty trolley acquisition stage and a load transport stage. During the unloaded acquisition trolley stage, a first set of physical parameters, including the basic envelope radius R1 and the unloaded kinematic parameters, is used to generate a spatiotemporal trajectory. During the load transport phase, the dimensions of the assembly trolley are obtained and projected in a three-dimensional Boolean union with the dimensions of the autonomous mobile robot to generate a second set of physical parameters, including the extended envelope radius R2 and load kinematic parameters. ; Establish a four-dimensional spatiotemporal graph matrix, and calculate the delay penalty coefficient based on the delivery deadline of the production work order to set the weight; A mixed-integer linear programming model is used to prioritize the allocation of time windows to the production work orders with high weights. During this period, the extended occupied area is marked in the four-dimensional spatiotemporal graph matrix based on the second set of physical parameters. The occupation duration of the extended occupied area is calculated based on the preset node distance and the load kinematic parameters. When planning a path for a production order with low weight, if the corresponding envelope overlaps with the marked extended occupied area in time, the cost penalty value of the corresponding path is increased, and the corresponding autonomous mobile robot is controlled to wait and avoid obstacles in a safe area.

2. The method as described in claim 1, characterized in that, include: In response to receiving the highest priority order, the estimated arrival time of each of the autonomous mobile robots in the idle acquisition trolley stage is iterated. Based on the verified truncation time window, a task change frame carrying the new coordinates and trolley verification code is issued, and the original task is returned to the scheduling pool.

3. The method as described in claim 1, characterized in that, The unloaded kinematic parameters include unloaded driving speed and unloaded angular velocity; The load kinematic parameters include the load travel speed after deceleration and the load angular velocity.

4. The method as described in claim 3, characterized in that, The duration of occupation of the extended occupied area is configured to be calculated in the following way: Obtain the quotient of the physical distance between preset nodes and the driving speed of the load; The quotient is superimposed with the rotation buffer time set based on the extended envelope radius R2 to obtain the occupation duration.

5. The method as described in claim 4, characterized in that, The rotation buffer time is configured as the quotient of the rotation angle and the load angular velocity; The spatial coordinate system in the four-dimensional spatiotemporal graph matrix is ​​configured to extend the extended envelope radius R2 outward by a preset safety margin distance to define the physically impassable boundary.

6. The method as described in claim 1, characterized in that, The method of using a mixed-integer linear programming model to prioritize the allocation of time windows to production orders with higher weights includes: Configure time window mutual exclusion constraints and space occupation non-overlapping constraints; The available time window sequence is calculated based on the mutual exclusion constraint of the time window and the non-overlapping constraint of the spatial occupancy.

7. The method as described in claim 1, characterized in that, The control corresponding to the autonomous mobile robot waiting to avoid obstacles in a safe area includes: The safe region is defined by tracing back to the nearest non-overlapping node in the four-dimensional spatiotemporal graph matrix. A waiting time window is generated in the safe area, and the waiting state is maintained until the extended occupied area corresponding to the production work order with high weight is released.

8. The method as described in claim 7, characterized in that, The cost penalty value for the corresponding path is configured to be positively adjusted based on the length of the waiting time window; In response to the cost penalty value exceeding a preset penalty threshold, an alternative path is replanned to avoid the expanded occupied area.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored on the memory, wherein the processor executes the program to implement the method of claim 1.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method of claim 1.

Images