Unmanned Warehousing Methods Based on Multi-Robot Collaboration for Cross-Floor Material Transfer

By constructing a multi-level dynamic spatiotemporal topology map and an improved contract network protocol, the problem of seamless connection of cross-floor material flow in the warehousing and logistics system was solved, realizing efficient collaboration between vertical conveying equipment and planar handling robots, and improving the overall flow efficiency and stability of the system.

CN123113907APending Publication Date: 2026-11-17RIZHAO PORT GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611007498.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-08
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing warehousing and logistics systems lack a unified scheduling platform for cross-floor material flow, resulting in fragmented management of vertical conveying equipment and horizontal handling robots. This makes it difficult to achieve seamless connection and efficient collaboration across floors and equipment. Furthermore, static path planning methods are unable to cope with order fluctuations and sudden tasks, leading to local congestion and system deadlock, resulting in low efficiency in overall inventory management and dynamic resource allocation.

Method used

A multi-level dynamic spatiotemporal topology map is constructed, and the entrance and exit of the vertical conveying equipment are defined as vertical interaction nodes with time window attributes. An improved contract network protocol is used for dynamic task allocation, and path planning and deadlock prevention are performed through spatiotemporal right-of-way auction. When the robot arrives at the vertical interaction node, it performs handshake control to achieve seamless docking.

Benefits of technology

It achieves direct integration of cross-floor path planning and vertical transportation resources, improves global load balancing and flow efficiency, avoids deadlock conflicts between equipment, shortens cross-floor logistics flow time, and significantly improves the collaborative operation efficiency and operational stability of unmanned warehousing systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113907A_ABST
    Figure CN123113907A_ABST
Patent Text Reader

Abstract

This invention discloses a method for unmanned warehousing based on multi-robot collaboration for cross-floor material handling. The method includes: real-time prediction of robot trajectories; triggering a right-of-way auction mechanism when overlapping areas are detected in the predicted trajectories of multiple robots in space and time; each robot calculating a right-of-way bidding function value, which is a weighted sum of the time delay cost caused by avoidance and the task priority; allocating exclusive right-of-way in the overlapping area to the robot with the highest right-of-way bidding function value; and sending a reservation request when a robot reaches the trigger node corresponding to the vertical conveyor equipment while performing a cross-floor task, receiving feedback from the vertical conveyor equipment control system on the optimal connection time window, and dynamically adjusting its speed based on the difference between this time window and the current time to ensure timely arrival at the vertical conveyor equipment. This invention effectively resolves multi-robot conflicts and achieves efficient collaboration in cross-floor material handling, significantly improving the overall operational efficiency of the warehousing and logistics system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of smart logistics technology, specifically relating to an unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration. Background Technology

[0002] In existing automated warehousing and logistics systems, multi-robot collaborative operations are typically limited to two-dimensional planar scheduling within the same floor. When faced with the need for material transfer across floors, the system often manages vertical conveying equipment and planar handling robots separately. This separate control results in a lack of a unified scheduling platform between the various devices, making it difficult to achieve seamless connection and efficient collaboration across floors and devices.

[0003] Due to a lack of overall planning, robots often face long waiting times or path conflicts when arriving at vertical transport equipment. Existing static path planning methods struggle to cope with order fluctuations and sudden tasks, easily leading to localized congestion or even system deadlock in vertical interaction areas. Furthermore, traditional scheduling algorithms do not take into account the time window attributes of vertical transport resources, resulting in inefficient global inventory management and dynamic resource allocation.

[0004] Existing unmanned warehousing solutions lack intelligent and dynamic task allocation and micro-level deadlock prevention mechanisms when dealing with complex and ever-changing cross-floor flow scenarios. This limits the overall utilization rate of warehouse space and cannot meet the needs of modern warehousing for fully digital closed-loop and deadlock-free scheduling in highly dynamic environments. Summary of the Invention

[0005] To address the problems existing in the background technology, this invention proposes an unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration.

[0006] The first aspect of the present invention provides a method for unmanned warehousing based on multi-robot collaboration for cross-floor material transfer, comprising: A multi-level dynamic spatiotemporal topology map is constructed, in which the entrances and exits of the vertical transport equipment on each floor are defined as vertical interaction nodes with time window attributes, and the operating status of the vertical transport equipment is transformed into a dynamic spatiotemporal resource window to generate a topology structure that includes static road network, static obstacles and dynamic vertical transport resources. Based on the multi-level dynamic spatiotemporal topology map, dynamic task allocation is performed using an improved contract network protocol. In this process, the robot calculates the bid price based on a comprehensive bid cost function that includes estimated distance, energy consumption, urgency coefficient, and congestion penalty value, and the robot with the lowest bid price is determined as the successful bidder to perform the handling task. During the process of the winning robot performing the handling task, path planning and deadlock prevention are carried out based on spatiotemporal right-of-way auction. When the predicted trajectories of multiple robots overlap in spatiotemporal regions, the exclusive right-of-way for the overlapping regions is allocated according to the right-of-way bidding function that includes time delay costs and task priorities. When the robot arrives at the vertical interaction node while performing a cross-floor task, it performs a handshake control for vertical flow. The robot sends a reservation request containing the target floor and the estimated arrival time to the vertical conveyor and adjusts its speed according to the feedback of the optimal connection time window to pass through the vertical interaction node.

[0007] Optionally, the construction of a multi-level dynamic spatiotemporal topology map includes: Establish a warehouse environment model containing multiple floors. The warehouse environment model consists of a set of nodes including shelf points, workstations and vertical interaction points, and a set of edges including intra-floor aisle edges and cross-floor virtual edges. The operating parameters of the vertical conveying equipment are acquired in real time, and the car position and door opening / closing time of the vertical conveying equipment are mapped to the time window attribute of the vertical interaction node. The connectivity condition for the cross-floor virtual edge is determined to be that it is currently within the time window interval corresponding to each vertical interaction node.

[0008] Optionally, the dynamic task allocation based on the improved contract network protocol includes: When the warehouse management system issues a new order, it sends out a tender notice to the idle robots; After receiving the task, the robot calculates the comprehensive bidding cost function value, which is composed of the estimated distance from the robot's current position to the task starting point, the movement energy consumption and load energy consumption required to execute the task, the reciprocal of the task's urgency coefficient, and the weighted sum of the estimated path congestion penalty value. The robot sends the calculated comprehensive bidding cost function value to the system as the bid price.

[0009] Optionally, the urgency coefficient of the task is positively correlated with the urgency level of the task. The urgency coefficient of the task corresponding to the emergency order is greater than that of the task corresponding to the regular order, so that the comprehensive bidding cost function value corresponding to the emergency order is less than that of the comprehensive bidding cost function value corresponding to the regular order.

[0010] Optionally, the path planning and deadlock prevention based on spatiotemporal right-of-way auction includes: The robot can predict its trajectory within a preset time period in real time. When the predicted trajectories of two or more robots are detected to overlap in space and time, the right-of-way auction mechanism is triggered. Each robot participating in the auction calculates its own right-of-way bidding function value, which is a weighted sum of the time delay cost caused by avoidance and the task priority. The system compares the right-of-way bidding function values ​​of each party in the conflict area, and assigns the exclusive right-of-way in the overlapping area to the robot with the highest right-of-way bidding function value. The remaining robots execute avoidance strategies such as slowing down and waiting or planning detours.

[0011] Optionally, the time delay cost caused by the avoidance is positively correlated with the robot's current load status and the priority of the current task being performed. The time delay cost caused by the avoidance of a fully loaded robot or a high-priority task robot is greater than the time delay cost caused by the avoidance of an empty robot or a non-urgent task robot.

[0012] Optionally, the handshake control for performing vertical flow includes: When the robot reaches the trigger node corresponding to the vertical conveying equipment, it sends the reservation request to the vertical conveying equipment control system. The vertical conveying equipment control system calculates and feeds back the optimal connection time window based on the current car position, running direction, and load. The robot receives the optimal docking time window and dynamically adjusts its speed based on the difference between the optimal docking time window and the current time to ensure that it arrives at the vertical conveyor within the optimal docking time window.

[0013] Optionally, the dynamic adjustment of its own driving speed includes: When the robot predicts that the arrival time is earlier than the start time of the optimal pick-up time window, it reduces its driving speed and follows at a constant speed. When the robot predicts that its arrival time will be later than the end of the optimal pick-up time window, it will increase its speed to catch up. When the robot predicts that the arrival time is within the optimal pick-up time window, it maintains its current travel speed.

[0014] Optionally, the path-predicted congestion penalty value is calculated based on historical traffic flow heatmaps, including: The system calculates the traffic frequency of each road segment within a historical time period and generates the historical traffic flow heat map. During the task allocation phase, the frequency of passage through high-heat areas along the task path is used as an input variable for the estimated congestion penalty value of the path, enabling the robot to actively avoid historically high-heat and congested areas.

[0015] Optionally, the weight of the cross-floor virtual edge in the multi-level dynamic spatiotemporal topology map is composed of the running time of the vertical conveying equipment and the door opening and closing time, so that the path planning algorithm regards the cross-floor action as a continuous edge with time constraints when searching the global path.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a method for unmanned warehousing based on multi-robot collaboration for cross-floor material handling. By constructing a multi-level dynamic spatiotemporal topology map, the entrances and exits of vertical conveying equipment are defined as vertical interaction nodes with time window attributes. This breaks through the limitations of traditional static maps, enabling cross-floor path planning to directly integrate dynamic vertical transportation resources and effectively solving the information silo problem in multi-floor environments. Dynamic task allocation based on this map introduces a comprehensive bidding cost function, which can comprehensively evaluate estimated distance, energy consumption, and congestion penalties, allowing the system to automatically assign tasks to the robot with the lowest overall cost, thereby achieving global load balancing and improving overall flow efficiency. During the execution of cross-floor tasks, the system uses a spatiotemporal right-of-way auction mechanism for path planning and deadlock prevention. When multiple robot trajectories overlap in spatiotemporal areas, exclusive right-of-way is allocated based on time delay cost and task priority, avoiding deadlock conflicts between devices and ensuring the smooth passage of high-priority tasks. In addition, the vertical flow handshake control enables the robot to dynamically adjust its speed according to the optimal connection time window when it arrives at the vertical interaction node, realizing seamless docking between the robot and the vertical conveying equipment, greatly shortening the flow time of cross-floor logistics, and significantly improving the collaborative operation efficiency and operational stability of the unmanned warehouse system. Attached Figure Description

[0017] Figure 1 This is a flowchart of a warehouse unmanned method for cross-floor material transfer based on multi-robot collaboration in one embodiment of the present invention; Figure 2 This is another flowchart of a warehouse unmanned method for cross-floor material transfer based on multi-robot collaboration in one embodiment of the present invention; Figure 3 This is another flowchart of a warehouse unmanned method for cross-floor material transfer based on multi-robot collaboration, according to one embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0019] In one embodiment, such as Figure 1 As shown, a method for unmanned warehousing based on multi-robot collaboration for cross-floor material transfer is provided, and this method is applied to... Figure 1 Taking China as an example, the following specific steps will be used: S10: Construct a multi-level dynamic spatiotemporal topology map, wherein the entrances and exits of the vertical transport equipment on each floor are defined as vertical interaction nodes with time window attributes, and the operating status of the vertical transport equipment is converted into a dynamic spatiotemporal resource window to generate a topology structure that includes static road networks, static obstacles and dynamic vertical transport resources.

[0020] Specifically, the steps for constructing a multi-level dynamic spatiotemporal topology map first include establishing a warehouse environment model containing multiple floors. This model consists of a set of nodes including shelf points, workstations, and vertical interaction points, as well as a set of edges including intra-floor aisle edges and cross-floor virtual edges. The system defines the entrances and exits of vertical conveyor equipment on each floor as vertical interaction nodes with time window attributes, and acquires the operating parameters of the vertical conveyor equipment in real time, mapping the car position and door opening / closing time of the vertical conveyor equipment to the time window attributes of the vertical interaction nodes. Simultaneously, the system transforms the operating status of the vertical conveyor equipment into a dynamic spatiotemporal resource window, determining that the connectivity condition of cross-floor virtual edges is that the current moment falls within the time window interval corresponding to the vertical interaction node, thereby generating a topology that integrates static road networks, static obstacles, and dynamic vertical transportation resources.

[0021] In this invention, the operating status of vertical conveying equipment can be sensed in real time, transforming the originally isolated cross-floor physical actions into continuous digital edges with time constraints. This enables the path planning algorithm to directly utilize dynamic spatiotemporal resource windows for global optimization, thereby effectively solving the problem of information silos in complex multi-floor environments and providing accurate data support for subsequent dynamic task allocation and path planning.

[0022] S20: Based on the multi-level dynamic spatiotemporal topology map, dynamic task allocation is performed using the improved contract network protocol. The robot calculates the bid price based on a comprehensive bidding cost function that includes estimated distance, energy consumption, urgency coefficient, and congestion penalty value, and determines the robot with the lowest bid price as the winner to perform the handling task.

[0023] Specifically, the improved contract network protocol in this invention is a dynamic task allocation method for warehousing and logistics systems. When the warehousing management system issues a new order, the system publishes a tender notice to robots in an idle state. After receiving the task, the robot reads the path information in a multi-level dynamic spatiotemporal topology map and calculates a comprehensive bidding cost function value. This comprehensive bidding cost function value is a weighted sum of the estimated distance from the robot's current position to the task's starting point, the movement energy consumption and load energy consumption required to execute the task, the reciprocal of the task's urgency coefficient, and the estimated path congestion penalty value. The robot sends the calculated comprehensive bidding cost function value as its bid price to the system, and the system determines the robot with the lowest bid price as the successful bidder to execute the handling task. For example, the urgency coefficient of the task is positively correlated with the urgency level of the task. The urgency coefficient of a sudden urgent order is greater than that of a regular order, making the comprehensive bidding cost function value of the sudden urgent order less than that of a regular order. In addition, the path prediction congestion penalty value is calculated based on historical traffic flow heat maps. The system counts the traffic frequency of each road segment within a historical period and uses the traffic frequency of the task path passing through high-heat areas as the input variable for the path prediction congestion penalty value during the task allocation stage.

[0024] In this invention, the system can take into account physical distance, energy consumption and road network congestion during the task allocation stage, and automatically assign the handling task to the robot with the lowest overall cost, thereby achieving global load balancing, effectively avoiding traffic congestion in local hot spots, and significantly improving the overall circulation efficiency of the warehousing and logistics system.

[0025] S30: During the process of the winning robot performing the handling task, path planning and deadlock prevention are performed based on spatiotemporal right-of-way auction. When the predicted trajectories of multiple robots overlap in spatiotemporal regions, the exclusive right-of-way for the overlapping regions is allocated according to the right-of-way bidding function that includes time delay costs and task priorities.

[0026] Specifically, the robot predicts its trajectory in real time over a preset time period. When the predicted trajectories of two or more robots overlap in space and time, a right-of-way auction mechanism is triggered. Each participating robot calculates its own right-of-way bidding function value, which is a weighted sum of the time delay cost caused by avoidance and the task priority. The system compares the right-of-way bidding function values ​​of each party in the conflict area and allocates the exclusive right-of-way in the overlapping area to the robot with the highest right-of-way bidding function value. The remaining robots execute avoidance strategies such as slowing down and waiting or planning detours. In one embodiment, the time delay cost caused by avoidance is positively correlated with the robot's current load status and the priority of the currently executed task. The time delay cost caused by avoidance for a fully loaded robot or a robot with a high-priority task is greater than the time delay cost caused by avoidance for an empty robot or a robot with a non-urgent task.

[0027] In this invention, the system can dynamically resolve path conflicts during robot task execution, intelligently allocate right-of-way based on task urgency and operational status, avoid the efficiency loss caused by traditional deadlock prevention mechanisms, ensure priority passage for high-value tasks, and significantly improve the smoothness and response speed of multi-robot collaborative operation systems.

[0028] S40: When the robot arrives at the vertical interaction node while performing a cross-floor task, it performs a handshake control for vertical flow, wherein the robot sends a reservation request containing the target floor and the estimated arrival time to the vertical conveyor, and adjusts its travel speed according to the feedback of the optimal connection time window to pass through the vertical interaction node.

[0029] Specifically, when the robot performs a cross-floor task and reaches the vertical interaction node, the specific implementation of the vertical flow handshake control is as follows: When the robot reaches the trigger node corresponding to the vertical conveyor equipment, it sends a reservation request containing the target floor and the estimated arrival time to the vertical conveyor equipment control system. After receiving the request, the vertical conveyor equipment control system calculates and feeds back the optimal connection time window based on the current car position, running direction, and load. The robot receives the optimal connection time window and dynamically adjusts its own speed according to the difference between the optimal connection time window and the current time to ensure that it arrives at the vertical conveyor equipment within the optimal connection time window. In one embodiment, the dynamic adjustment of its own speed includes reducing the speed to maintain a constant speed when the robot predicts an arrival time earlier than the start time of the optimal connection time window, increasing the speed to catch up when the robot predicts an arrival time later than the end time of the optimal connection time window, and maintaining the current speed when the robot predicts an arrival time within the optimal connection time window.

[0030] In this invention, the system achieves precise coordination between the robot and the vertical conveying equipment, eliminating the idle time caused by waiting for elevators in traditional cross-floor transportation, realizing seamless connection of vertical flow links, greatly shortening the flow cycle of cross-floor logistics, and significantly improving the overall operating efficiency of the unmanned warehousing system.

[0031] In one embodiment, step S10, namely constructing a multi-level dynamic spatiotemporal topology map, further includes the following steps: S11: Establish a warehouse environment model containing multiple floors. The warehouse environment model consists of a set of nodes including shelf points, workstations and vertical interaction points, and a set of edges including intra-floor aisle edges and cross-floor virtual edges.

[0032] S12: Obtain the operating parameters of the vertical conveying equipment in real time, and map the car position and door opening / closing time of the vertical conveying equipment to the time window attribute of the vertical interaction node.

[0033] S13: Determine that the connectivity condition of the cross-floor virtual edge is that it is currently within the time window interval corresponding to each vertical interaction node.

[0034] Specifically, the purpose of constructing a multi-level dynamic spatiotemporal topology map is to establish a computational model that reflects the spatial structure of the warehousing environment and the dynamic operating status of vertical conveying equipment. During the initial construction of this computational model, a three-dimensional spatial model encompassing all floors needs to be established based on the warehouse building's design drawings or actual measurement data. During the operation of the three-dimensional spatial model, the system collects the operating parameters of the vertical conveying equipment in real time through a sensor network and dynamically updates the relevant attributes in the model based on these parameters. This dynamically updated parameter model represents the warehousing environment in the form of a topology graph, where nodes represent key locations in the space, and edges represent traversable paths between locations. The connectivity of edges is not fixed but changes dynamically over time. This dynamic characteristic allows the dynamically updated parameter model to accurately reflect the accessibility of the vertical conveying equipment at different times, providing accurate spatial basis for subsequent path searching and task scheduling.

[0035] In this invention, the warehouse environment model is a digital spatial description encompassing multiple floors. This model consists of two main parts: a set of nodes and a set of edges. The node set contains three types of elements: shelf points, workstations, and vertical interaction points. A shelf point refers to the coordinates of a shelf in the warehouse environment used for storing goods. Each shelf point corresponds to a specific storage location or a group of storage locations, typically represented in the warehouse environment model by the coordinates of the shelf's center point or entrance point. A workstation refers to the coordinates of a workbench or work area in the warehouse environment used for performing operations such as loading, unloading, sorting, and packaging. Workstations may include workbenches in the receiving area, packing stations in the shipping area, and interface locations for automated sorting equipment. A vertical interaction point refers to the coordinates of the location where horizontal handling equipment and vertical conveying equipment dock during vertical transfer of goods between different floors. Vertical interaction points are typically located at the entrances and exits of vertical conveying equipment on each floor. For example, a three-story warehouse building may have one vertical interaction point on the first floor, located at the elevator exit on the first floor. Another vertical interaction point may be located on the second floor, located at the same elevator exit on the second floor. There is also a vertical interaction point on the third floor, located at the elevator exit on the third floor. These vertical interaction points are spatially located at the same vertical projection position, but belong to different floors.

[0036] The edge set contains two types of elements: intra-floor aisle edges and cross-floor virtual edges. Intra-floor aisle edges refer to walkable paths connecting any two nodes within the same floor. Intra-floor aisle edges can connect two shelf points, a shelf point and a workstation, or a workstation and a vertical interaction point. The existence of intra-floor aisle edges depends on the actual physical layout of the warehouse environment, including aisle width, obstacle locations, and the open areas of the access control system. Cross-floor virtual edges are logical connections between vertical interaction points on different floors. Cross-floor virtual edges do not correspond to fixed aisles in physical space but represent the possibility of transferring goods between different floors via vertical conveyor equipment. The target of cross-floor virtual edges is vertical interaction points located on different floors but at the same vertical projection position. For example, a cross-floor virtual edge can be established between a vertical interaction point on the first floor and a vertical interaction point on the second floor. A cross-floor virtual edge can also be established between a vertical interaction point on the second floor and a vertical interaction point on the third floor.

[0037] Vertical conveying equipment refers to equipment used for vertically transporting goods between different floors, including but not limited to freight elevators, hoists, and vertical conveyors. The system acquires the equipment's operating parameters in real time through sensors installed on the vertical conveying equipment or through a communication interface with the equipment's control system. Operating parameters include at least the current position of the car and the time information for door opening and closing. The car position refers to the floor location of the vertical conveying equipment's cargo platform at the current moment. Door opening and closing time refers to the time points involved in the complete process of the vertical conveying equipment's car door opening and closing, including the opening moment, opening duration, closing moment, and closing duration.

[0038] The system maps the real-time car position and door opening / closing time to time window attributes of vertical interaction nodes. Vertical interaction nodes are the vertical interaction points in the aforementioned node set. The time window attribute describes the availability of a vertical interaction node in the time dimension. Specifically, it indicates the time periods during which the vertical conveyor equipment is available for horizontal transport equipment to enter, exit, or stay on a given floor. When the car is on a floor and the car door is open, the corresponding vertical interaction node for that floor is available. When the car leaves that floor or the car door is closed, the corresponding vertical interaction node for that floor is unavailable. The time window attribute is expressed as a time interval, with each interval having a start time and an end time. For example, if the car of the vertical conveyor equipment arrives at the second floor and opens the door at time 10:00, and closes the door and leaves the second floor at time 10:15, then the vertical interaction node on the second floor is available during the time interval from time 10:00 to time 10:15. The system records this time interval as a time window for the vertical interaction node. When the car arrives at the second floor again and opens the door at 10 minutes and 30 seconds, the system records the second time window. Multiple time windows are arranged in chronological order to form the time window sequence of the vertical interaction node.

[0039] By mapping the car position and door opening / closing time to time window attributes of vertical interaction nodes, the system transforms the physical motion state of the vertical conveyor into the time dimension attributes of nodes in the topology model. This mapping allows the warehousing environment model to add a time dimension on top of the spatial dimension, thus enabling it to describe dynamic changes. This mapping process is performed in real time; whenever the operating state of the vertical conveyor changes, the system updates the time window attributes of the corresponding vertical interaction nodes.

[0040] The connectivity condition of a cross-floor virtual edge refers to the decision rule that must be met to determine whether passage between the two vertical interaction nodes connected by the edge is permitted at the current moment. The connectivity condition of a cross-floor virtual edge is determined as follows: the current moment is within the time window interval corresponding to the vertical interaction node connected by the cross-floor virtual edge.

[0041] Specifically, a virtual edge spanning two floors connects two vertical interaction nodes located on different floors. When a horizontal transport device needs to move from one floor to the other, the system needs to determine whether passage through the virtual edge is permitted at that moment. The system obtains the current time value and then retrieves the time window attributes of each of the two vertical interaction nodes connected by the virtual edge. The system then determines whether the current moment falls within the time window intervals of both vertical interaction nodes simultaneously. If the current moment falls within both time window intervals, the virtual edge is connected at that moment, and passage is permitted. If the current moment does not fall within both time window intervals simultaneously, the virtual edge is disconnected at that moment, and passage is not permitted.

[0042] For example, a virtual edge spanning multiple floors connects vertical interaction node A on the first floor and vertical interaction node B on the second floor. The current time window for vertical interaction node A is from 10:00 to 10:20. The current time window for vertical interaction node B is from 10:05 to 10:25. The current time is 10:10. The system determines that the current time of 10:10 falls within both the time window from 10:00 to 10:20 and from 10:05 to 10:25, therefore the virtual edge spanning multiple floors is connected at this time. If the current time is 10:22, which is outside the time window of vertical interaction node A, then the virtual edge spanning multiple floors is disconnected.

[0043] In another example, a warehouse building is equipped with two vertical conveyor systems. The first vertical conveyor system has its own time window attributes for its vertical interaction nodes on the first and second floors. The second vertical conveyor system also has its own time window attributes for its vertical interaction nodes on the first and third floors. The system independently maintains the connectivity status determination for the cross-floor virtual edges corresponding to each vertical conveyor system. The connectivity of the first set of cross-floor virtual edges depends on the time window attributes of the first vertical conveyor system on the first and second floors. The connectivity of the second set of cross-floor virtual edges depends on the time window attributes of the second vertical conveyor system on the first and third floors. The connectivity status of the two sets of cross-floor virtual edges does not affect each other and is determined independently.

[0044] In this invention, the method of determining connectivity conditions allows cross-floor path planning to no longer rely on static floor connectivity assumptions, but rather on dynamic judgment based on the real-time operating status of the vertical conveyor equipment. When a vertical conveyor equipment is unable to provide service at a certain time due to failure, maintenance, or task conflicts, the corresponding time window attribute will be updated accordingly, and the cross-floor virtual edge will automatically become disconnected. The path planning module can thus avoid planning cross-floor paths that cannot be actually executed. This method effectively reduces path planning failures and task interruptions caused by the unavailability of vertical conveyor equipment. This method also makes load balancing among multiple vertical conveyor equipment possible. The system can select the cross-floor virtual edge corresponding to the vertical conveyor equipment with an available time window for path planning based on the time window distribution of each vertical conveyor equipment, thereby avoiding multiple horizontal transport equipment from simultaneously competing for the same vertical conveyor equipment. This method improves the success rate of cross-floor transportation tasks and the overall operating efficiency in multi-floor warehousing systems.

[0045] In one embodiment, in step S10, the weight of the cross-floor virtual edge in the multi-level dynamic spatiotemporal topology map is composed of the running time of the vertical conveying equipment and the door opening and closing time, so that the path planning algorithm regards the cross-floor action as a continuous edge with time constraints when searching the global path.

[0046] Specifically, the weight of the cross-floor virtual edge is not calculated based on physical distance, but rather on time cost. This weight consists of two parts: the running time of the vertical transport equipment and the door opening / closing time. The running time of the vertical transport equipment refers to the physical movement time required for the car to travel vertically from one floor to another. The door opening / closing time refers to the total time required for the car to open, maintain the open state, and close the door after reaching the target floor. When the path planning algorithm searches for a global path, the system accumulates the running time and door opening / closing time, and assigns the accumulated result to the corresponding cross-floor virtual edge as the travel cost of that edge. Through this weighting method, the path planning algorithm treats the cross-floor action as a continuous edge with time constraints, rather than a simple physical spatial connection, when searching for a global path. This means that the algorithm considers the time consumption of vertical transportation when calculating the total path cost, thus enabling a more accurate assessment of the actual travel efficiency of different paths.

[0047] For example, suppose a warehouse environment consists of a first floor and a second floor, connected by a vertical conveyor system. The system pre-determines the travel time required for the vertical conveyor to travel from the first floor to the second floor as 10 seconds. Simultaneously, the system determines the door opening and closing time after reaching the floor as 5 seconds. When the path planning algorithm needs to calculate the weight of the vertical interaction point from the first floor to the second floor via a cross-floor virtual edge, the system adds the 10-second travel time to the 5-second door opening and closing time, resulting in a weight of 15 seconds for the cross-floor virtual edge. If another path exists, such as a staircase or ramp for cross-floor movement, although its physical distance may be shorter, if its corresponding travel time cost is higher than 15 seconds, the path planning algorithm will prioritize the cross-floor virtual edge via the vertical conveyor system. Conversely, if the vertical conveyor system experiences increased waiting time due to busy conditions, thus increasing the total weight including waiting time, the path planning algorithm will automatically avoid that edge and choose another path with lower time cost.

[0048] In another example, considering the load differences of the vertical conveyor in different operating directions, the runtime may vary. For instance, when the vertical conveyor is fully loaded and moving upwards, its runtime might be set to 12 seconds, while when it's empty and moving downwards, the runtime might be set to 8 seconds. The system dynamically adjusts the weight of the virtual edges across floors based on the current operating status. If the current task requires upward transport, the system adds the runtime of 12 seconds to the door opening / closing time of 5 seconds, resulting in a weight of 17 seconds. If the current task requires downward transport, the system adds the runtime of 8 seconds to the door opening / closing time of 5 seconds, resulting in a weight of 13 seconds. This dynamic weight calculation method allows the path planning algorithm to flexibly adjust the selection strategy for cross-floor paths based on the real-time operating conditions of the vertical conveyor.

[0049] In this invention, by using the running time of the vertical conveyor equipment and the door opening / closing time as weights for the virtual edges across floors, the system achieves a unified quantification of spatial distance and time cost. This method allows the path planning algorithm to transcend the limitations of a two-dimensional plane and perform global optimization in a topological map combining three-dimensional space and time dimensions. The path planning algorithm no longer merely seeks the path with the shortest physical distance, but rather the optimal path with the least time consumption. This effectively avoids the problem of reduced overall transportation efficiency caused by slow operation of the vertical conveyor equipment or frequent door opening / closing. Simultaneously, treating cross-floor actions as a continuous edge with time constraints simplifies the state judgment logic during path search, reduces the computational complexity of the algorithm during node switching, and significantly improves the real-time performance and accuracy of path planning in multi-floor warehousing environments.

[0050] In one embodiment, step S20, namely the dynamic task allocation using the improved contract network protocol, further includes the following steps: S21: When the warehouse management system issues a new order, it sends a tender notice to the idle robots.

[0051] S22: After receiving the task, the robot calculates the comprehensive bidding cost function value, which is composed of the estimated distance from the robot's current position to the task starting point, the mobile energy consumption and load energy consumption required to execute the task, the reciprocal of the task's urgency coefficient, and the path prediction congestion penalty value.

[0052] S23: The robot sends the calculated comprehensive bidding cost function value to the system as the bid price.

[0053] Specifically, when the warehouse management system issues a new order, it first parses the task requirements in the order, determines the starting location, target location, and task type, and then issues a tender notice to all idle robots in the warehouse environment. Upon receiving the tender notice, the robot immediately invokes its internal path planning and status evaluation modules to calculate a comprehensive bidding cost function value based on its real-time status. This comprehensive bidding cost function value is a weighted sum of the estimated distance from the robot's current position to the task's starting point, the movement energy consumption and load energy consumption required to execute the task, the reciprocal of the task's urgency coefficient, and the estimated path congestion penalty value. The robot sends the calculated comprehensive bidding cost function value as its bid price to the warehouse management system. The warehouse management system collects the bid prices from all participating robots and, based on the principle of the lowest bidder wins, assigns the task to the robot with the lowest comprehensive bidding cost function value, thus achieving dynamic task allocation.

[0054] For example, suppose there are three idle robots in a warehouse system: Robot A, Robot B, and Robot C. When the system issues a new goods handling order, it sends a tender notice to these three robots. Robot A, upon receiving the notice, calculates the estimated distance from its current position to the task's starting point as 20 meters, the energy consumption for movement and load required to execute the task (converted to 15 joules), the task's urgency coefficient as 2, and the estimated path congestion penalty as 3. After weighted calculation, the overall bidding cost function value is 10. Robot B calculates its estimated distance as 30 meters, its energy consumption (converted to 10 joules), its urgency coefficient as 2, and its congestion penalty as 5, resulting in an overall bidding cost function value of 18. Robot C calculates its estimated distance as 15 meters, its energy consumption (converted to 20 joules), its urgency coefficient as 2, and its congestion penalty as 2, resulting in an overall bidding cost function value of 12. The system compares these three values ​​and determines that Robot A has the lowest overall bidding cost function value; therefore, the task is assigned to Robot A.

[0055] In this invention, by introducing a comprehensive bidding cost function composed of estimated distance, energy consumption, the reciprocal of the urgency coefficient, and a congestion penalty value into the bidding decision, the system achieves a comprehensive consideration of multi-dimensional factors for task allocation. This allocation mechanism avoids the problems of excessive energy consumption or path congestion that may result from assigning tasks solely based on distance. Incorporating the reciprocal of the task's urgency coefficient into the calculation allows robots to prioritize tasks with high urgency, improving the system's response speed to high-priority orders. Simultaneously, the introduction of the path estimation congestion penalty value allows robots to proactively avoid potential congested areas during the bidding stage, effectively balancing traffic flow within the warehouse environment. This dynamic allocation strategy based on comprehensive cost significantly improves the efficiency of multi-robot collaborative operations, reduces the overall operating energy consumption of the system, and enhances the robustness and flexibility of the warehousing and logistics system.

[0056] In one embodiment, step S22, where the path prediction congestion penalty value is calculated based on historical traffic flow heatmaps, further includes the following steps: S221: The system calculates the traffic frequency of each road segment within a historical time period and generates the aforementioned historical traffic flow heat map; S222: During the task allocation phase, the frequency of passage through high-heat areas along the task path is used as the input variable for the estimated congestion penalty value of the path, so that the robot can actively avoid historically high-heat congestion areas.

[0057] Specifically, the system first calculates the traffic frequency of each road segment within a historical time period and generates a historical traffic flow heatmap accordingly. This historical time period is typically set to the same time window as the current task execution period; for example, it calculates the traffic conditions of each road segment from 10:00 AM to 11:00 AM daily over the past week. Traffic frequency refers to the number of times a robot or handling equipment passes through a specific road segment within that historical time period. Based on the traffic frequency of each road segment, the system classifies the road segments within the warehouse environment into different heatmap levels; the higher the traffic frequency, the higher the heatmap level, thus forming the historical traffic flow heatmap. During the task allocation phase, when the robot plans its path from its current location to the task start point, the system extracts the road segments traversed by the task path and uses the traffic frequency of these road segments in the historical traffic flow heatmap as input variables for the path's estimated congestion penalty value. If the route traversed by the task is in a high-heat area in the historical traffic flow heatmap, a larger estimated congestion penalty value will be assigned; if the route traversed by the task is in a low-heat area, a smaller estimated congestion penalty value will be assigned.

[0058] For example, suppose there is a main aisle connecting the shelving area and the sorting area in the warehouse environment. The system analyzes the traffic frequency of this main aisle from 2 PM to 3 PM daily over the past month and finds that the traffic frequency during this period is 120 times, which is considered a high-congestion area. Meanwhile, the system finds that another detour aisle has a traffic frequency of only 20 times during the same period, which is considered a low-congestion area. When the warehouse management system issues a new order at 2:10 PM, robot A's planned path needs to pass through the main aisle. The system uses the main aisle's traffic frequency of 120 times as an input variable to calculate a higher estimated congestion penalty value. Robot B's planned path passes through the detour aisle. The system uses the detour aisle's traffic frequency of 20 times as an input variable to calculate a lower estimated congestion penalty value. In the calculation of the overall bidding cost function, robot B has a lower estimated congestion penalty value than robot A, so the system ultimately assigns the task to robot B, allowing it to perform the task through the detour aisle.

[0059] In this invention, by introducing path prediction congestion penalty values ​​based on historical traffic flow heatmaps, the system possesses proactive congestion prediction capabilities during the task allocation phase. This mechanism enables robots to proactively avoid historically high-congestion areas and select alternative routes with lower traffic frequency, thereby effectively dispersing traffic flow within the warehouse environment. This solution avoids traffic congestion and deadlock problems caused by multiple robots simultaneously entering the same popular road segment, improving the traffic efficiency of multi-robot collaborative operations. Furthermore, congestion prediction based on historical data eliminates the need for real-time sensor monitoring, reducing the system's computational complexity and hardware costs, making task allocation decisions more efficient and stable.

[0060] In one embodiment, in step S20, the urgency coefficient of the task is positively correlated with the urgency level of the task. The urgency coefficient of the task corresponding to the emergency order is greater than that of the task corresponding to the regular order, so that the comprehensive bidding cost function value corresponding to the emergency order is less than that of the comprehensive bidding cost function value corresponding to the regular order.

[0061] Specifically, in the improved contract network protocol, the comprehensive bidding cost function is a weighted sum of estimated distance, energy consumption, the reciprocal of the task's urgency coefficient, and the estimated path congestion penalty. The task's urgency coefficient is positively correlated with its urgency level; that is, the higher the task's urgency level, the larger its assigned urgency coefficient. When calculating the comprehensive bidding cost function, the system uses the reciprocal of the task's urgency coefficient as a calculation factor. According to mathematical principles, the larger the task's urgency coefficient, the smaller its reciprocal, resulting in a smaller final comprehensive bidding cost function value. Since the system identifies the robot with the smallest comprehensive bidding cost function value as the winning bidder, this design ensures that the comprehensive bidding cost function value for urgent orders is lower than that for regular orders, thus providing a significant advantage in the bidding process.

[0062] For example, the warehouse management system receives two types of orders simultaneously. The first type is urgent orders, which the system sets to the highest urgency level and assigns a large urgency coefficient. The second type is regular orders, which the system sets to a normal urgency level and assigns a smaller urgency coefficient. When the system issues a tender notice to idle robots in the warehouse environment, participating robots calculate the comprehensive bid cost function value for each type of order. For urgent orders, because the urgency coefficient is large, its reciprocal is small, resulting in a lower overall comprehensive bid cost function value including this reciprocal. For regular orders, because the urgency coefficient is small, its reciprocal is large, leading to a relatively high comprehensive bid cost function value. After the system aggregates the bids from all robots, robots with the same distance and energy consumption conditions will inevitably have a lower comprehensive bid cost function value when executing urgent orders than when executing regular orders. Therefore, the system will prioritize assigning urgent orders to robots with execution capabilities.

[0063] In this invention, a dynamic priority scheduling mechanism based on a mathematical model is established by incorporating the reciprocal of the task urgency coefficient into the comprehensive bidding cost function. This mechanism eliminates the need for complex rule-based judgments during the path planning phase, directly converting time sensitivity into numerical weights within the cost function. This design ensures that urgent orders receive system-level priority in task allocation, effectively preventing delays in urgent tasks caused by backlogs of regular orders. Simultaneously, this solution maintains overall system fairness, allowing non-urgent tasks to be allocated according to optimal distance and energy consumption principles. This ensures timely execution of high-priority tasks while maintaining the overall operational efficiency and resource utilization of the warehousing and logistics system.

[0064] In one embodiment, such as Figure 2As shown, in step S30, namely the path planning and deadlock prevention based on spatiotemporal right-of-way auction, the following steps are also included: S31: The robot predicts its trajectory in real time within a preset time period.

[0065] S32: When the predicted trajectories of two or more robots are detected to overlap in space and time, the right-of-way auction mechanism is triggered.

[0066] S33: Each robot participating in the auction calculates its own right-of-way bidding function value, which is a weighted sum of the time delay cost caused by avoidance and the task priority.

[0067] S34: The system compares the right-of-way bidding function values ​​of each party in the conflict area, allocates the exclusive right-of-way in the overlapping area to the robot with the largest right-of-way bidding function value, and the remaining robots execute avoidance strategies such as slowing down and waiting or planning detour paths.

[0068] Specifically, during the handling process, the robot predicts its trajectory in real time for a predetermined time period. This prediction process is based on the robot's current speed, acceleration, and a planned global path, generating a sequence of predicted trajectories containing spatial coordinates and timestamps. When the system detects that the predicted trajectories of two or more robots overlap in space and time, a path conflict is identified, and a right-of-way auction mechanism is immediately triggered. Each participating robot calculates its own right-of-way bidding function value, which is a weighted sum of the time delay cost caused by avoidance and the task priority. The time delay cost caused by avoidance refers to the additional time expected to be consumed if the robot performs an avoidance action. This time is positively correlated with the robot's current load status; a fully loaded robot has a greater time delay cost due to its larger inertia and longer braking distance than an unloaded robot. The task priority is determined by the urgency of the current task. The system compares the right-of-way bidding function values ​​of each party within the conflict area, allocating the exclusive right-of-way in the overlapping area to the robot with the highest right-of-way bidding function value. The remaining robots then implement avoidance strategies such as slowing down and waiting or planning detours.

[0069] For example, suppose there is an intersection in the warehouse environment, and the predicted trajectories of robot A and robot B overlap in time and space at this intersection. Robot A is currently fully loaded and executing a high-priority urgent order. The system calculates that the time delay cost caused by its avoidance is high, and its task priority is extremely high. The weighted average of these two factors results in a right-of-way bidding function value of 100. Robot B is currently unloaded and executing a regular order. The system calculates that the time delay cost caused by its avoidance is low, and its task priority is low. The weighted average of these two factors results in a right-of-way bidding function value of 30. The system compares the right-of-way bidding function values ​​of the two robots and determines that the right-of-way bidding function value of robot A is greater than that of robot B. Therefore, the system allocates exclusive right-of-way in the overlapping area to robot A. Robot A maintains its current speed and passes through the intersection, while robot B, after receiving the avoidance instruction, implements a deceleration and waiting strategy at a preset safe distance from the intersection. Robot B resumes its journey only after robot A has completely passed through the overlapping area.

[0070] In this invention, an auction mechanism based on right-of-way bidding function values ​​is introduced to dynamically resolve conflicts and prevent deadlock among multiple robots. This mechanism comprehensively quantifies avoidance costs and task priorities, ensuring that high-value tasks and fully loaded robots receive priority passage, effectively preventing a decline in overall system efficiency caused by blind avoidance. Simultaneously, based on the prediction of spatiotemporal overlap areas and right-of-way allocation, robots can proactively implement avoidance strategies before conflicts occur, eliminating deadlock caused by multiple robots refusing to yield to each other at the source. This dynamic right-of-way allocation method significantly improves the smoothness and response speed of the multi-robot collaborative operation system, ensuring the efficient operation of the warehousing and logistics system.

[0071] In one embodiment, in step S30, the time delay cost caused by the avoidance is positively correlated with the robot's current load status and the priority of the currently executed task. The time delay cost caused by the avoidance of a fully loaded robot or a high-priority task robot is greater than the time delay cost caused by the avoidance of an empty robot or a non-urgent task robot.

[0072] Specifically, the time delay cost caused by obstacle avoidance is used to quantify the system efficiency loss incurred by the robot due to obstacle avoidance actions. This cost is positively correlated with the robot's current load status and the priority of the current task, where positive correlation refers to the degree of negative impact of the obstacle avoidance action on the robot's current state. Specifically, the time delay cost caused by obstacle avoidance for a fully loaded robot or a high-priority task robot is greater than that for an unloaded robot or a non-urgent task robot. The system determines the robot's current load status by collecting load sensor data and obtains the priority of the current task by combining it with instructions from the task scheduling system. For a fully loaded robot, due to its large load mass, long braking distance, and high energy consumption required for restarting and accelerating, the time and energy consumption required to recover to the original driving state significantly increase once an obstacle avoidance action is performed. For a high-priority task robot, its task completion time directly affects the timeliness of the overall logistics chain; any deceleration or detour will result in a significant time penalty. Therefore, the system quantifies the above physical losses and time penalties as the time delay cost caused by obstacle avoidance, ensuring that high-load and high-priority robots gain a numerical advantage in right-of-way competition.

[0073] For example, suppose robot C and robot D are about to have a path conflict in a warehouse aisle. Robot C is currently fully loaded with 500 kg of cargo and is performing an emergency replenishment task. Robot D is currently unloaded with 0 kg of cargo and is performing a routine inventory check. When the system triggers the right-of-way auction mechanism, it calculates the time delay cost caused by the avoidance for each robot. For robot C, the system determines that if it decelerates to avoid the collision, it will consume a large amount of energy to re-accelerate due to its large inertia, and it will delay the emergency replenishment task. Therefore, it is assigned a higher time delay cost for avoiding the collision, set to 80. For robot D, the system determines that it is unloaded, can start and stop flexibly, and its routine inventory check task is not time-sensitive, so avoiding the collision has a smaller impact on it. Therefore, it is assigned a lower time delay cost for avoiding the collision, set to 20. When calculating the right-of-way bidding function value, robot C, with its higher time delay cost for avoiding the collision, has a significantly higher right-of-way bidding function value than robot D, thus successfully winning the right-of-way.

[0074] In this invention, by establishing a mechanism linking the time delay cost caused by obstacle avoidance with load status and task priority, the system achieves intelligent right-of-way allocation based on physical characteristics and business value. This mechanism avoids frequent braking and starting of fully loaded robots, effectively reducing mechanical wear and energy consumption. Simultaneously, this solution ensures that high-priority tasks can pass first when encountering conflicts, avoiding overall business process delays caused by avoiding low-priority tasks. This cost-aware obstacle avoidance strategy enables multi-robot systems to automatically prioritize protecting high-value, high-load transportation tasks when facing complex road conditions, thereby significantly improving the overall operational efficiency and economic benefits of the warehousing and logistics system.

[0075] In one embodiment, such as Figure 3 As shown, in step S40, i.e., the handshake control for performing vertical flow, the following steps are also included: S41: When the robot arrives at the trigger node corresponding to the vertical conveying equipment, it sends the reservation request to the vertical conveying equipment control system.

[0076] S42: The vertical conveying equipment control system calculates and feeds back the optimal connection time window based on the current car position, running direction and load.

[0077] S43: The robot receives the optimal connection time window and dynamically adjusts its own speed according to the difference between the optimal connection time window and the current time to ensure that it arrives at the vertical conveying equipment within the optimal connection time window.

[0078] Specifically, after receiving a cross-floor transport task, the robot travels along a planned path towards the vertical conveyor equipment. When the robot reaches the trigger node corresponding to the vertical conveyor equipment, this trigger node is set as a physical location at a preset safe distance from the vertical interaction node. The robot sends a reservation request to the vertical conveyor equipment control system via a wireless communication network. This reservation request includes the robot's identification, the target floor, and the current task type. Upon receiving the reservation request, the vertical conveyor equipment control system obtains its own operating status parameters in real time, including the current car position, direction of travel, and load status. The system calculates the travel time required to reach the target floor based on the current car position, determines whether a change of direction is necessary based on the current direction of travel, and determines the duration of the door opening and closing action based on the load status. It then calculates and provides the optimal connection time window. This optimal connection time window represents the time interval during which the vertical conveyor equipment car is open on the target floor, allowing the robot to safely enter. After receiving the optimal connection time window, the robot extracts the start time of this time window and calculates the difference between the optimal connection time window and the current time. This difference is the target travel time for the robot to reach the vertical interaction node. The robot dynamically adjusts its speed based on the target travel time and the remaining distance to ensure that it arrives at the vertical conveyor within the optimal connection time window.

[0079] For example, suppose the robot is currently at a trigger node 30 meters away from the vertical interaction node, and the current time is 10 minutes and 0 seconds. The robot sends a reservation request to the vertical conveyor control system, requesting to enter the vertical interaction node on the second floor. The vertical conveyor control system determines that the elevator car is currently on the first floor, traveling upwards, and is unloaded. Based on the current position of the elevator car, the system calculates that it will take 8 seconds to reach the second floor, and after arrival, the door will open and remain open for 5 seconds. Therefore, the optimal connection time window is calculated to be between 10 minutes and 8 seconds and 10 minutes and 13 seconds. After receiving this time window, the robot calculates that the difference between the starting time of the optimal connection time window of 10 minutes and 8 seconds and the current time of 10 minutes and 0 seconds is 8 seconds. Based on the remaining distance of 30 meters and the target travel time of 8 seconds, the robot adjusts its travel speed to 3.75 meters per second and travels towards the vertical interaction node at a constant speed, ensuring that it arrives and enters the elevator car precisely at 10 minutes and 8 seconds.

[0080] In this invention, by implementing an interaction mechanism between the reservation request and the optimal pick-up time window at the trigger node, the system achieves proactive collaboration between the robot and the vertical conveyor equipment, rather than passive waiting. The vertical conveyor equipment control system calculates the optimal pick-up time window based on its real-time status, ensuring the accuracy and executability of the time window. The robot dynamically adjusts its speed according to the difference between the time window and the current time, achieving precise matching between the arrival time and the pick-up window, effectively eliminating the robot's invalid waiting time at the vertical interaction node. This solution significantly improves the connection efficiency of cross-floor transfers, avoids robot queuing congestion caused by the vertical conveyor equipment not being ready, and ensures the efficient and continuous operation of the multi-floor warehousing system.

[0081] In one embodiment, step S43, the dynamic adjustment of its own driving speed, further includes the following steps: S431: When the robot predicts that the arrival time is earlier than the start time of the optimal docking time window, it reduces its driving speed and follows at a constant speed.

[0082] S432: When the robot predicts that the arrival time is later than the end of the optimal pick-up time window, it increases its speed to catch up.

[0083] S433: When the robot predicts that the arrival time is within the optimal pick-up time window, maintain the current driving speed.

[0084] Specifically, after receiving the optimal docking time window, the robot calculates its predicted arrival time in real time based on its current position and the distance between itself and the vertical interaction node. The system compares this predicted arrival time with the optimal docking time window and executes corresponding speed control strategies based on the comparison results. When the robot's predicted arrival time is earlier than the start time of the optimal docking time window, it indicates that the robot is traveling too fast. Maintaining the current speed would result in arriving prematurely and incurring unnecessary waiting at the vertical interaction node. In this case, the robot reduces its speed to maintain a constant pace, extending the travel time to smooth the arrival rhythm and ensure arrival just as the optimal docking time window opens. When the robot's predicted arrival time is later than the end time of the optimal docking time window, it indicates that the robot is traveling too slowly. Maintaining the current speed would cause it to miss the docking window. In this case, the robot increases its speed to catch up, shortening the remaining travel time within the safe speed limit to try to enter the vertical interaction node before the optimal docking time window closes. When the robot's predicted arrival time is within the optimal docking time window, it indicates that the current travel rhythm perfectly matches the operating status of the vertical conveyor equipment. The robot maintains its current speed until it smoothly enters the vertical interaction node and completes the docking.

[0085] For example, assume the optimal docking time window is from 10:00 to 10:30. The robot assesses its speed 50 meters from the vertical interaction node. If the system calculates the predicted arrival time at the current speed to be 9:45, which is earlier than the initial time of 10:00, the robot controller immediately outputs a deceleration command, reducing its speed to maintain a constant following speed, thus delaying the predicted arrival time to after 10:00. If the system calculates the predicted arrival time to be 10:45, which is later than the final time of 10:30, the robot controller immediately outputs an acceleration command, increasing its speed to catch up, thus advancing the predicted arrival time to before 10:30. If the system calculates the predicted arrival time to be 10:15, which falls within the range of 10:00 to 10:30, the robot maintains its current speed and proceeds directly to the vertical interaction node.

[0086] In this invention, by comparing the predicted arrival time with the optimal pick-up time window in real time and dynamically adjusting the travel speed, the system achieves precise spatiotemporal coordination between the robot and the vertical conveying equipment. This dynamic speed adjustment mechanism effectively eliminates the robot's ineffective waiting time at vertical interaction nodes, while also avoiding secondary queuing and task replanning caused by missing the pick-up window. This solution significantly improves the continuity and smoothness of cross-floor flow, ensuring the efficient operation of the warehousing and logistics system under high-frequency cross-floor operations, and significantly increasing the overall logistics throughput.

[0087] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0088] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for unmanned warehousing based on multi-robot collaboration for cross-floor material transfer, characterized in that, include: A multi-level dynamic spatiotemporal topology map is constructed, in which the entrances and exits of the vertical transport equipment on each floor are defined as vertical interaction nodes with time window attributes, and the operating status of the vertical transport equipment is transformed into a dynamic spatiotemporal resource window to generate a topology structure that includes static road network, static obstacles and dynamic vertical transport resources. Based on the multi-level dynamic spatiotemporal topology map, dynamic task allocation is performed using an improved contract network protocol. In this process, the robot calculates the bid price based on a comprehensive bid cost function that includes estimated distance, energy consumption, urgency coefficient, and congestion penalty value, and the robot with the lowest bid price is determined as the successful bidder to perform the handling task. During the process of the winning robot performing the handling task, path planning and deadlock prevention are carried out based on spatiotemporal right-of-way auction. When the predicted trajectories of multiple robots overlap in spatiotemporal regions, the exclusive right-of-way for the overlapping regions is allocated according to the right-of-way bidding function that includes time delay costs and task priorities. When the robot arrives at the vertical interaction node while performing a cross-floor task, it performs a handshake control for vertical flow. The robot sends a reservation request containing the target floor and the estimated arrival time to the vertical conveyor and adjusts its speed according to the feedback of the optimal connection time window to pass through the vertical interaction node.

2. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 1, characterized in that, The construction of a multi-level dynamic spatiotemporal topology map includes: Establish a warehouse environment model containing multiple floors. The warehouse environment model consists of a set of nodes including shelf points, workstations and vertical interaction points, and a set of edges including intra-floor aisle edges and cross-floor virtual edges. The operating parameters of the vertical conveying equipment are acquired in real time, and the car position and door opening / closing time of the vertical conveying equipment are mapped to the time window attribute of the vertical interaction node. The connectivity condition for the cross-floor virtual edge is determined to be that it is currently within the time window interval corresponding to each vertical interaction node.

3. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 1, characterized in that, The method of using the improved contract network protocol for dynamic task allocation includes: When the warehouse management system issues a new order, it sends out a tender notice to the idle robots; After receiving the task, the robot calculates the comprehensive bidding cost function value, which is composed of the estimated distance from the robot's current position to the task starting point, the movement energy consumption and load energy consumption required to execute the task, the reciprocal of the task's urgency coefficient, and the weighted sum of the estimated path congestion penalty value. The robot sends the calculated comprehensive bidding cost function value to the system as the bid price.

4. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 3, characterized in that, The urgency coefficient of the task is positively correlated with the urgency level of the task. The urgency coefficient of the task corresponding to the emergency order is greater than that of the task corresponding to the regular order, which makes the comprehensive bidding cost function value corresponding to the emergency order less than that of the comprehensive bidding cost function value corresponding to the regular order.

5. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 1, characterized in that, The path planning and deadlock prevention based on spatiotemporal right-of-way auction includes: The robot can predict its trajectory within a preset time period in real time. When the predicted trajectories of two or more robots are detected to overlap in space and time, the right-of-way auction mechanism is triggered. Each robot participating in the auction calculates its own right-of-way bidding function value, which is a weighted sum of the time delay cost caused by avoidance and the task priority. The system compares the right-of-way bidding function values ​​of each party in the conflict area, and assigns the exclusive right-of-way in the overlapping area to the robot with the highest right-of-way bidding function value. The remaining robots execute avoidance strategies such as slowing down and waiting or planning detours.

6. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 5, characterized in that, The time delay cost caused by avoidance is positively correlated with the robot's current load status and the priority of the current task being performed. The time delay cost caused by avoidance for a fully loaded robot or a robot performing a high-priority task is greater than the time delay cost caused by avoidance for an unloaded robot or a robot performing a non-urgent task.

7. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 1, characterized in that, The handshake control for executing vertical flow includes: When the robot reaches the trigger node corresponding to the vertical conveying equipment, it sends the reservation request to the vertical conveying equipment control system. The vertical conveying equipment control system calculates and feeds back the optimal connection time window based on the current car position, running direction, and load. The robot receives the optimal docking time window and dynamically adjusts its speed based on the difference between the optimal docking time window and the current time to ensure that it arrives at the vertical conveyor within the optimal docking time window.

8. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 1, characterized in that, The dynamic adjustment of its own driving speed includes: When the robot predicts that the arrival time is earlier than the start time of the optimal pick-up time window, it reduces its driving speed and follows at a constant speed. When the robot predicts that its arrival time will be later than the end of the optimal pick-up time window, it will increase its speed to catch up. When the robot predicts that the arrival time is within the optimal pick-up time window, it maintains its current travel speed.

9. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration as described in claim 3, characterized in that, The path-based estimated congestion penalty value is calculated based on historical traffic flow heatmaps, including: The system calculates the traffic frequency of each road segment within a historical time period and generates the historical traffic flow heat map. During the task allocation phase, the frequency of passage through high-heat areas along the task path is used as an input variable for the estimated congestion penalty value of the path, enabling the robot to actively avoid historically high-heat and congested areas.

10. The unmanned warehousing method for cross-floor material transfer based on multi-robot collaboration according to claim 2, characterized in that, The weight of the virtual edge across floors in the multi-level dynamic spatiotemporal topology map is composed of the running time of the vertical conveying equipment and the opening and closing time of the door, so that the path planning algorithm regards the cross-floor action as a continuous edge with time constraints when searching the global path.