Multi-robot cloud scheduling cooperation method and system

Abstract

The invention relates to the technical field of robot scheduling, in particular to a multi-robot cloud scheduling cooperation method and system. The method comprises the following steps that building elevator group data are obtained, building elevator service groups are divided according to the building elevator group data, and building elevator service group data are obtained; acquiring target path data of the robot; performing path resource section cutting on the robot target path data and the building elevator service group data to obtain split path data; acquiring current transit floor data, and performing transit waiting estimation according to the current transit floor data and the split path data to obtain transit waiting estimation data; carrying out transfer graph construction according to the transfer waiting estimation data to obtain transfer graph data; obtaining path conflict data according to the transfer graph data; and obtaining robot scheduling data according to the path conflict data so as to carry out multi-robot cooperative scheduling operation. According to the method, the task completion efficiency and the scheduling intelligence level can be improved in a multi-robot cross-floor collaborative scene.

Description

Technical Field

[0001] The present invention relates to the field of robot scheduling technology, and in particular to a multi-robot cloud scheduling collaboration method and system. Background Art

[0002] With the widespread adoption of service robots in enclosed environments such as buildings, hospitals, and hotels, efficiently scheduling multiple robots and coordinating their use of shared resources (such as elevators) has become a core challenge. For example, in high-rise buildings, to improve elevator efficiency and zoning, the elevator system is often divided into high-rise and low-rise groups, with the service floors of these two groups non-overlapping. While this zoning approach improves operational efficiency, it also inevitably introduces the problem of "cross-elevator transfers" during the robot's mission path. For example, a robot may need to take an elevator from the low-rise area to a transfer floor (such as the 10th floor) and then wait for the high-rise elevator to complete its transfer before reaching its target high floor. Existing technologies lack effective solutions for this cross-elevator scheduling requirement. Existing methods often assume that all elevators are a unified scheduling resource, failing to fully consider the elevator service area boundaries and transfer point capacity constraints. This leads to problems such as robot congestion at transfer floors, increased transfer delays, and frequent scheduling conflicts. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a multi-robot cloud scheduling collaboration method and system to solve at least one of the above technical problems.

[0004] This application provides a multi-robot cloud scheduling and coordination method, which is applied to a cloud platform. The method includes: S1. Obtain building elevator group data, and divide building elevator service groups according to the building elevator group data to obtain building elevator service group data; S2. Obtain robot target path data; perform path resource segmentation on the robot target path data and the building elevator service group data to obtain split path data; S3. Obtain current transfer floor data, and perform transfer waiting estimation based on the current transfer floor data and the split path data to obtain transfer waiting estimation data; S4. Construct a transfer graph based on the transfer waiting estimation data to obtain transfer graph data; perform graph structure conflict calculation based on the transfer graph data to obtain path conflict data; perform scheduling generation based on the path conflict data to obtain robot scheduling data for multi-robot collaborative scheduling operations.

[0005] The present invention divides the service areas of the elevator groups in the building and cuts the path resource segments in combination with the robot's target path, which can achieve accurate identification of the cross-floor service boundaries of multiple elevator groups; through the transfer floor waiting estimation, the transfer risk and waiting time can be dynamically evaluated to avoid scheduling delays caused by the separation of elevator groups; by constructing a transfer graph and performing graph structure conflict calculation, the scheduling bottlenecks and resource competition relationships between path segments can be effectively identified; based on the conflict data, a scheduling plan is generated to achieve collaborative scheduling of multiple robots in scenarios where high- and low-floor elevators are not interconnected, thereby improving overall operating efficiency and cross-floor task completion rate.

[0006] Optionally, the building elevator service group division includes: Extract the floor elevator service relationship based on the building elevator group data to obtain the floor elevator matrix data; Obtaining elevator operation behavior data, and calculating service overlap frequency based on the elevator operation behavior data to obtain service overlap frequency data; The floor elevator connectivity graph is constructed based on the floor elevator matrix data and the service overlapping frequency data to obtain the building elevator service group data.

[0007] The present invention extracts the service relationship of floor elevators to form floor elevator matrix data, which can clearly depict the service coverage of each elevator on different floors and provide structured input for scheduling; through the elevator operation behavior data, combined with the operation trajectory and scheduling records to calculate the service overlap frequency, it is possible to quantify the degree of coordination and redundancy between different elevators in floor services, and avoid scheduling rigidity caused by static division; by fusing matrix data with service overlap frequency data to construct a floor elevator connectivity diagram, it not only improves the visibility and flexibility of cross-elevator group scheduling in a building, but also can effectively identify service fault areas or high overlap risk areas, providing data support for building a reasonable building elevator service group.

[0008] Optionally, constructing the floor elevator connectivity graph includes: Structural coverage feature extraction and elevator collaborative scheduling feature extraction are performed based on the floor elevator matrix data and the service overlapping frequency data, and structural coverage feature data and elevator collaborative scheduling feature data are obtained respectively; Constructing a coupling diagram for the floor elevator matrix data according to the structural coverage feature data and the elevator collaborative scheduling feature data to obtain coupling diagram data; Perform fault zone detection on the coupling diagram data to obtain authorization control diagram data, wherein the fault zone detection includes floor jump fault detection, floor no access record detection and authorization restriction detection; Perform time period service mode switching constraints on the authorization control graph data to obtain constraint graph data; Connected branches are extracted based on the constraint graph data to obtain the building elevator service group data.

[0009] The present invention extracts structural coverage features and elevator collaborative scheduling features respectively by fusing floor elevator matrix data and service overlapping frequency data, which can not only identify the service scope of each elevator at the spatial level, but also reflect its collaborative capability in actual operation, thereby realizing comprehensive modeling of service capability and scheduling correlation; constructing a coupling graph helps to reveal the coupling relationship and resource accessibility between elevator services; by performing fault area detection on the coupling graph data, it can accurately identify service blind spots or inaccessible areas in the building caused by factors such as physical structure, elevator grouping, and access rights, thereby enhancing the system's adaptability to complex scenarios; through the time period service mode switching constraint, the elevator authorization and path feasibility in different time periods can be dynamically adjusted to meet the needs of morning and evening peaks or nighttime strategy changes; through the constraint graph, connected branches are extracted to achieve fine division of multi-elevator service blocks, providing highly reliable structural support and scheduling foundation for cross-group path scheduling and robot cloud collaboration.

[0010] Optionally, the path resource segment cutting includes: Divide the robot target path data and the building elevator service group data into segments to obtain path division data; Perform service attribute extraction and transfer point extraction based on the path division data to obtain service attribute data and transfer point data respectively; Perform transfer constraint dependency processing on the path division data according to the service attribute data and the transfer point data to obtain path dependency data; A path dependency graph is constructed based on the path dependency data to obtain split path data.

[0011] In this paper, the robot target path data is combined with the building elevator service group data to perform path segment division, which can achieve accurate distinction between cross-elevator group and cross-floor sections in the path, thereby improving scheduling granularity; extracting the service attributes and transfer point information of the path segment helps to identify the service elevator group and its potential transfer nodes corresponding to each path segment, introducing space-resource dual semantics for path planning; transfer constraint dependency processing is performed through service attributes and transfer points, which can explicitly model the resource connection logic and time sequence constraints between path segments, which is suitable for path switching management in high- and low-area elevator separation scenarios; finally, a path dependency graph is constructed, which not only improves the structural clarity of path planning, but also provides an operational graphical model support for scheduling sequence reasoning and multi-robot collaborative allocation, significantly enhancing the path feasibility and robustness of the scheduling system under the building structure.

[0012] Optionally, the transit waiting estimation includes: The arrival time of multiple robots is superimposed and calculated based on the current transfer floor data and the split path data to obtain the arrival superposition data; Kernel density calculation is performed based on the arrival superposition data to obtain local aggregation data; Throughput calculation and light-opening door cycle calculation are performed based on the current transfer floor data to obtain throughput data and light-opening door cycle data respectively; The transfer waiting time of the local aggregated data is estimated based on the throughput data and the light gate cycle data to obtain the transfer waiting estimation data.

[0013] The present invention performs a superposition calculation of the arrival time of multiple robots on the current transfer floor data and the split path data, which can capture the potential path concentration trend and accurately simulate the probability of multiple robots arriving at the transfer point at the same time; kernel density calculation is used to smooth the arrival distribution to generate local aggregation data, which can effectively measure the congestion intensity of the transfer node within a specific time window; combined with the actual throughput of the transfer floor and the elevator door opening and closing cycle data, the processing capacity and traffic efficiency of the transfer elevator at the actual physical operation level are dynamically reflected; the above-mentioned throughput and cycle parameters are introduced into the local aggregation data for weighted estimation, which can quantify the waiting delay of the current transfer node under different scheduling schemes, improve the congestion perception and response capability of the scheduling system on the cross-elevator group scheduling path, provide data support for path conflict avoidance and task rescheduling, and effectively improve the timeliness and stability of multi-robot collaborative scheduling.

[0014] Optionally, the transfer map construction includes: Construct a path segment transfer map based on the current transfer floor data and the split path data to obtain the path segment transfer map data; Assign edge weights to the path segment transfer graph data based on the transfer waiting estimation data to obtain weighted graph data; Directed scheduling is performed on the weighted graph data according to the weighted graph data to obtain transfer graph data.

[0015] In the present invention, by constructing a path segment transfer graph based on transfer floor data and split path data, the transfer relationship between robots in different service segments can be abstracted into a structured graph model, which clearly represents the connection structure between path segments and transfer nodes; by introducing transfer waiting estimation data to assign edge weights to the graph structure, dynamic factors such as transfer time and congestion level can be integrated into the graph model, making the graph structure not only connected but also schedulable; by performing directed scheduling on weighted graph data, transfer graph data oriented to execution order is generated to ensure that the transfer sequence between path segments meets the constraints of time dependency and resource accessibility; the system's structural analysis capability for multiple tasks in a path across elevator groups is improved, so that the scheduling engine has the data processing capability of "path-transfer-scheduling cost".

[0016] Optionally, the graph structure conflict calculation includes: Perform path occupancy mapping based on the transfer map data to obtain path occupancy data; Conflict pairs are extracted according to the path occupancy data to obtain conflict pair data; The path overlap tension is calculated based on the conflict data to obtain the path conflict data.

[0017] In the present invention, by performing path occupancy mapping on transfer map data, the spatial and temporal occupancy relationship of multiple robots on key resources such as shared channels and elevator nodes can be explicitly expressed, forming a high-precision resource scheduling view; based on the extraction of conflict pair information based on path occupancy data, potential scheduling conflict areas and key path intersections can be quickly located, providing basic support for conflict intervention; by introducing a path overlap tension calculation method, combined with the spatial overlap and temporal concurrency of path segments, the degree of competition of different paths for shared resources is quantified, and then path conflict data is generated, making the severity of conflicts measurable and sortable; the present invention enhances the ability to identify and evaluate potential bottlenecks and congestion points between multi-robot paths, provides an accurate conflict prediction and preprocessing mechanism for the scheduling system, effectively avoids task delays and system deadlocks, and significantly improves the coordination and scheduling success rate of multi-robot scheduling.

[0018] Optionally, the path overlap tension calculation includes: According to the conflict, the non-rigid geometric overlap and time occupancy intersection of the data are calculated to obtain fuzzy boundary overlap data and time occupancy intersection data respectively; According to the fuzzy boundary overlap data and the time occupancy intersection data, the shared resource node competition process is performed to obtain the node competition data; Propagate resource occupancy tension based on node competition data to obtain occupancy propagation matrix data; Based on the occupied propagation matrix data, gravity control is performed to schedule high-priority tasks and obtain path conflict data.

[0019] By calculating the non-rigid geometric overlap and temporal occupancy intersection of conflict pair data, the present invention can more accurately capture the potential overlap between robot paths in fuzzy boundary areas in high-rise building environments with diverse indoor corridor styles. Combined with the overlap of task execution time windows, it outputs more fine-grained fuzzy conflict features. The present invention also processes shared resource node competition, identifying the resource contention trends of multiple robots at bottleneck nodes such as elevator doors and corridor corners. By introducing a resource occupancy tension propagation mechanism, the local competition relationship is extended to the overall path structure, generating propagation matrix data that reflects the global scheduling pressure diffusion effect caused by local competition between paths. Through a high-priority task scheduling gravity control strategy, the attraction of high-priority tasks to resources is dynamically balanced with the overall system tension, improving the priority guarantee capability of task execution. The overall process enhances the modeling accuracy and response flexibility of complex scheduling conflicts, possesses strong predictive and feedforward control capabilities, and effectively improves the coordination efficiency and service quality of multi-robot scheduling systems in high-concurrency, resource-constrained scenarios.

[0020] Optionally, the schedule generation includes: Identify conflict core groups based on path conflict data to obtain conflict core group data; Perform path-level scheduling sorting based on conflict core group data to obtain scheduling sorting data; Generate robot scheduling data based on scheduling sorting to perform multi-robot collaborative scheduling operations.

[0021] The present invention uses path conflict data to identify conflict core groups, which can accurately capture the robot path set that forms a high-density competitive relationship in a specific time window or spatial node, and identify the key conflict sources that affect the overall scheduling bottleneck; on this basis, path-level scheduling sorting is performed, and sorting is performed according to multi-dimensional factors such as conflict intensity, task priority, and resource occupancy time, so that high-priority tasks can avoid low-priority tasks while taking into account the overall resource utilization efficiency; specific robot scheduling data is generated according to the sorting results, which can realize path staggered execution, active conflict avoidance and dynamic resource reuse, thereby improving the stability of multi-robot collaborative scheduling operations and system throughput.

[0022] Optionally, the present application further provides a multi-robot cloud scheduling collaborative system for executing the multi-robot cloud scheduling collaborative method described above, wherein the multi-robot cloud scheduling collaborative system includes: An elevator service group division module is used to obtain building elevator group data and divide building elevator service groups according to the building elevator group data to obtain building elevator service group data; The path segment structure modeling module is used to obtain the robot's target path data; the robot's target path data and the building elevator service group data are cut into path resource segments to obtain split path data; The transfer waiting prediction module is used to obtain the current transfer floor data and perform transfer waiting estimation based on the current transfer floor data and the split path data to obtain transfer waiting estimation data; The conflict-aware scheduling generation module is used to construct a transfer graph based on the transfer waiting estimation data to obtain transfer graph data; perform graph structure conflict calculations based on the transfer graph data to obtain path conflict data; and perform scheduling generation based on the path conflict data to obtain robot scheduling data for multi-robot collaborative scheduling operations.

[0023] The present invention aims to: S1. By dividing building elevator group data into service groups, introducing floor-elevator connectivity and service overlap frequency modeling, effectively identify elevator service boundaries and support the rational division of subsequent path structures; S2. By segmenting the robot's target path into resource segments based on the elevator service range, and modeling the path's service attributes and transfer dependencies, structured split path data is generated; S3. By analyzing the convergence behavior of transfer floors between different path segments, local aggregation density, throughput, and elevator cycle response are calculated to accurately estimate transfer waiting delays; S4. The robot task path is divided into path segment units and an initial transfer graph is constructed using these units as nodes and transfer floors as connecting edges. Edge weights in the graph are assigned based on transfer waiting estimates to form a weighted directed graph reflecting scheduling overhead. Within this graph structure, the system identifies path pairs with time window overlap and spatial path intersection as potential conflicting pairs; calculates the overlap tension function between conflicting paths to assess the severity of their scheduling conflicts; and finally, extracts the core conflict groups and performs local sorting optimization based on path priority and resource intensity, thereby achieving coordinated scheduling and control of multiple robot tasks at transfer points. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 A flowchart showing the steps of a multi-robot cloud scheduling collaboration method according to one embodiment is shown; Figure 2 A flowchart showing the steps of a method for dividing building elevator service groups according to an embodiment is shown; Figure 3 A flowchart showing the steps of a method for cutting a path resource segment according to an embodiment is shown; Figure 4 A flowchart showing the steps of a transit waiting estimation method according to an embodiment is shown; Figure 5A flowchart showing the steps of a method for constructing a transfer map according to an embodiment is shown; The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0025] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0026] Furthermore, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0027] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0028] See also Figures 1 to 5 , the present application provides a multi-robot cloud scheduling collaboration method, applied to a cloud platform, the method comprising: S1. Obtain building elevator group data, and divide building elevator service groups according to the building elevator group data to obtain building elevator service group data; Specifically, the basic information of all elevators in the building is obtained, including the range of floors served by each elevator, the controller number to which the elevator belongs, and the group control logic parameters; a floor-elevator service relationship matrix is ​​constructed, where the horizontal axis represents the floor number and the vertical axis represents the elevator number; let the matrix be M, and its dimension be n m, where n is the total number of floors and m is the number of elevators. If the jth elevator serves the ith floor, then =1, otherwise =0; Based on the historical elevator operation log, count the number of times any two elevators serve the same floor at the same time, and use this to calculate the service overlap frequency, and generate the overlap matrix R between elevators, where Indicates the With the The frequency index of elevators having service overlap in all service floors is obtained; the floors are regarded as nodes in the graph model, and the relationship between two floors with elevator service overlap is regarded as an edge, and an elevator-floor coupling graph model is constructed; "fault areas" are identified in the graph structure, that is, areas where the following situations exist: (1) floor skipping service, that is, some floors cannot be directly reached by any single elevator; (2) isolated floors without elevator coverage service; (3) partially restricted access floors under authority control; based on the constructed coupling graph, a graph traversal algorithm (such as depth-first traversal, union-find, etc.) is used to extract all connected subgraphs. Each connected subgraph corresponds to a group of physically reachable and dispatchable elevator service areas, and finally multiple elevator service groups are formed. The elevators in each service group can communicate with each other to serve all floors to which they belong, and the service groups are connected through floor transfer; each elevator service group is numbered and identified, and its key attribute information is output, including: the set of floor numbers covered, the set of elevator numbers to which it belongs, and possible transfer floors (that is, floors that belong to two or more service groups at the same time).

[0029] S2. Obtain robot target path data; perform path resource segmentation on the robot target path data and the building elevator service group data to obtain split path data; Specifically, the current task path data for each robot to be scheduled is obtained, including the coordinates of the task's starting and ending points. This path is then mapped to the building's spatial structure, clearly identifying the floors and physical locations of the starting and ending points, thereby forming a complete navigation path for the robot within the building. Based on the building's elevator service group data, the robot's path is determined to determine whether the floors it traverses span multiple elevator service groups. Potential transfer floors are identified at the path and elevator service boundaries, serving as path cut points. If the path is continuously within the same elevator service group, the path segment is not cut and remains as a single segment. If the path spans different elevator service groups, a transfer segment is inserted at the service boundary and marked as a connecting path requiring transfer. Each path segment is numbered by starting and ending floor, and the elevator group number on which the path depends is indicated. Service attribute extraction: For each segmented path segment, service attribute information is extracted, including the elevator group number to which it belongs; the starting and ending floor numbers; the estimated path segment's operating time window, such as the expected time range for entering and exiting the path segment; and resource type tags, such as "elevator segment" or "corridor segment." Establish scheduling dependencies in the path segment graph. If the execution of a path segment requires the completion of the previous segment as a prerequisite (for example, the previous segment reaches the transfer floor by elevator), a directed dependency edge is established between them to represent the scheduling constraints of the order.

[0030] S3. Obtain current transfer floor data, and perform transfer waiting estimation based on the current transfer floor data and the split path data to obtain transfer waiting estimation data; Specifically, based on the scheduled execution order and planned start and end times of each robot path segment, the expected arrival time of each robot at a specific transfer floor is calculated, forming a time series set. Based on the expected arrival times of all robots at that transfer floor, the arrival frequency distribution per unit time is statistically calculated to form a preliminary arrival density function. This arrival density function is smoothed using a kernel density estimation algorithm to generate a continuous clustering intensity function, representing the expected arrival density of robots at that transfer floor at any given moment, denoted as the clustering intensity function. For the elevators serving the target transfer floor, typical operating cycle parameters are extracted, including door opening time, passenger start time, operating time, arrival dwell time, and door closing time. The average time required for each complete elevator service cycle is calculated, and the maximum number of robot services that can be completed per unit time is calculated based on this clustering intensity and the elevator's maximum handling capacity. A queuing model is evaluated at any given moment based on the clustering intensity and the elevator's throughput capacity. If the clustering intensity at a given moment exceeds the elevator's maximum throughput capacity, the queue backlog is calculated, and the corresponding waiting time is estimated accordingly. The estimation function of waiting time is: divide the number of robots that exceed the throughput limit by the elevator processing capacity and multiply it by the time required for each service cycle to obtain the estimated transfer waiting time of the robot at that moment, that is, ,in To estimate the transit waiting time, For the moment The aggregation strength value, is the maximum throughput capacity of the elevator, The total time required for a complete elevator service cycle is calculated. Based on the transfer floors and scheduling time windows involved in each segment in the path resource segment graph, the estimated transfer waiting time is assigned to the corresponding path segment.

[0031] S4. Construct a transfer graph based on the transfer waiting estimation data to obtain transfer graph data; perform graph structure conflict calculation based on the transfer graph data to obtain path conflict data; perform scheduling generation based on the path conflict data to obtain robot scheduling data for multi-robot collaborative scheduling operations.

[0032] Specifically, each path segment is treated as a node in the graph. If an elevator transfer relationship exists between two path segments (i.e., there is a logical forward and backward dependency in the path and involves different service groups), a directed edge is introduced between the corresponding nodes. The edge weight is defined as the estimated transfer wait time between the two path segments, and is integrated with resource congestion indicators (such as the resource utilization rate of the target transfer node during the time period) to form the scheduling cost. Additional attribute information is added to each node, including the elevator service group number to which the path segment belongs, the elevator ID, and the estimated usage time window. Furthermore, taking into account the multi-entry and multi-exit structure of the building, it supports the construction of asymmetric graph structures with multiple in-degree and out-degree nodes.

[0033] For each path segment, a resource occupancy model in the space-time dimension is established. The model includes the location segments corresponding to the path segments in the floor space (such as corridors, elevator waiting areas, inside the elevator car, etc.), as well as the time windows of the path segments within the execution cycle, forming a three-dimensional space-time resource occupancy representation. For any two path segments, determine whether they have an intersection in the time window, and whether they overlap or are adjacent in the spatial area (for example, sharing elevator doors or passing through the same floor corridor). If both are true, they are determined to be a group of conflicting pairs. For each group of conflicting path pairs, the spatial overlap and temporal intersection are calculated separately, and a conflict tension function is established accordingly. This function can be expressed as a tension value equal to the spatial overlap multiplied by the weight coefficient α, plus the temporal intersection multiplied by the weight coefficient β, that is, ,in is the path conflict tension value, is the spatial overlap weight coefficient, is the spatial overlap, indicating the path segment and The degree of overlap in spatial location, such as the shared proportion of corridor sections and elevator waiting areas, is the time intersection weight coefficient, is the time intersection, representing the path segment and The degree of overlap in execution time windows, Number the path segments, Number the other path segment. The tension values ​​of all conflicting pairs are combined to form a path overlap tension matrix. Using a graph-based clustering algorithm (such as density clustering or spectral clustering), a conflict-intensive subgraph is extracted from the path overlap tension matrix. This identifies sets of path segments with high conflict coupling, forming several core conflict groups.

[0034] For all path segments to be scheduled, a scheduling scoring function is constructed based on factors such as their task priority index, the total tension value of the conflict group to which they belong, and the required waiting time. According to the scoring function, the path segments are topologically sorted, and the execution order of each path segment is determined while ensuring the correctness of the dependency relationship. Based on the sorting results and the transfer graph structure, a scheduling time window accurate to the path segment level is generated for each robot, and an elevator resource scheduling plan is simultaneously constructed. This plan is sent down to the floor intelligent terminal, elevator control system, and robot execution unit through the cloud platform to ensure the consistency of multi-agent behavior. During the actual scheduling execution process, if there are emergencies such as temporary elevator failure, robot deviation from the path, or changes in the floor environment, the system automatically triggers the dynamic scheduling recalculation mechanism and adjusts the scheduling plan in real time to restore scheduling stability and coordination.

[0035] Optionally, the building elevator service group division includes: S11. Extracting floor elevator service relationships based on building elevator group data to obtain floor elevator matrix data; Specifically, the system extracts elevator group configuration information from the building control platform or its associated device information sources. This information can come from the Building Information Modeling (BIM) system, building automation control system logs, property elevator operation configuration databases, or control logic parameters provided by the elevator manufacturer. Key attributes of each elevator include, but are not limited to, the elevator number, which uniquely identifies the elevator; the group control number, which indicates whether the elevator belongs to a specific control group; the service floor list, which lists the floors covered by the elevator (e.g., serving floors 1 to 10, or only floors 1, 3, 5, and 7); and the service time period, which, if a time-limited service rule exists, records the elevator's service range during different time periods (e.g., direct service during rush hour). Suppose a building contains L floors, numbered 1 to L, and there are E elevators, numbered 1 to E. A floor-elevator service matrix can be constructed. For floor i and elevator j, if elevator j serves floor i, the corresponding position is filled with "1" to indicate a service relationship; if elevator j does not serve floor i, the corresponding position is filled with "0" to indicate no service relationship. The matrix has L rows and E columns, with each row representing the service status of a floor and each column representing the service coverage of an elevator. Additional label information is added to each service cell in the matrix, including but not limited to the following: a service priority label, which marks the priority of the elevator for that floor (e.g., "high priority" indicates frequent stops or is the primary service elevator); an operating status label, which records the current or historical operating status, such as "available," "under maintenance," or "busy"; and a control logic label, which indicates whether the service mode is "direct elevator," "normal stop elevator," or "restricted stop elevator."

[0036] S12. Obtain elevator operation behavior data, and calculate service overlap frequency based on the elevator operation behavior data to obtain service overlap frequency data; Specifically, the system extracts historical operation log data from the building's elevator control platform (e.g., a commercially available brand of elevator control system). The time period for log records can be set based on actual business needs (e.g., the past day, the past week, or the past month). Each log record includes, but is not limited to, the elevator number, indicating which elevator performed the operation; the operation timestamp, which specifies the specific stop time, accurate to the second or minute; the direction of elevator operation, such as up, down, or empty return; the stop floor number, which specifies the floor reached during the operation; the full load mark, indicating whether the stop was fully loaded, indicating whether there was a transport task; and the request type, such as triggered by a passenger's outbound call (floor button) or an inbound call (car button). The service overlap frequency is the frequency with which any two elevators stop at the same floor at approximately the same time. For any pair of elevators numbered Elevator 1 and Elevator 2, the number of simultaneous stops on each floor is counted within a set observation period (e.g., 24 hours). A simultaneous stop is defined as the difference in the time between the two elevators stopping on a given floor not exceeding a set threshold (e.g., 60 seconds). The statistical results for all floors are summed and divided by the total observation time (in minutes or hours). The service overlap frequency between Elevator 1 and Elevator 2 is the ratio of the total number of near-simultaneous stops of any two elevators on all floors to the total observation time. After completing this overlap frequency statistics, the system constructs a service overlap frequency matrix between elevator pairs. This matrix is ​​symmetric, with the number of rows and columns equal to the total number of elevators. The value in the i-th row and j-th column of the matrix represents the service overlap frequency between the i-th elevator and the j-th elevator. The system sets an overlap frequency threshold (e.g., 0.002 times / minute) to filter out elevator pairs with extremely low service overlap. The retained elevator pairs with high overlap serve as key connecting edges in the subsequent construction of the service connectivity graph.

[0037] S13. Construct a floor elevator connectivity graph based on the floor elevator matrix data and the service overlapping frequency data to obtain building elevator service group data.

[0038] Specifically, an undirected weighted graph is constructed based on the floor elevator service matrix and the service overlap frequency matrix. Each elevator corresponds to a node in the graph, and the node's basic attributes include elevator number, group control number, and service floor range. Two elevators have a common service record on at least one floor, which is marked as "1" in the floor elevator matrix for that floor. The service overlap frequency of the two elevators exceeds a preset threshold (for example, 0.001 times / minute). Based on the basic edge weights, the following feature attributes are further extracted and added to each edge, such as the number of shared floors (the total number of floors served by the two elevators); the degree of service area overlap (by calculating the intersection and difference of the service floors of the two elevators); the trajectory similarity score (by calculating the cosine similarity based on the embedded vectors of the trajectory or stop pattern); and the group control logic consistency flag (by determining whether the two elevators are dispatched by the same group control system), which can be represented by a Boolean value flag (yes / no). When there are consecutive non-overlapping floors between elevator groups (for example, one group serves floors 1-10, and the other serves floors 12-20), this is defined as a service fault. If two elevators belong to different group control systems and their service areas are discontinuous, fault determination is enhanced. The service connectivity graph is segmented, dividing areas with weak connectivity and obvious faults into different subgraphs.

[0039] Optionally, constructing the floor elevator connectivity graph includes: Structural coverage feature extraction and elevator collaborative scheduling feature extraction are performed based on the floor elevator matrix data and the service overlapping frequency data, and structural coverage feature data and elevator collaborative scheduling feature data are obtained respectively; Specifically, the structural coverage features include , extract its service floor set The coverage interval width (i.e., coverage span) is defined as ,in is the maximum value function, For elevators The service floor collection, is the minimum function; calculate the elevator Floor coverage density , defined as the ratio of the number of service floors to the coverage area: ,in For the floor coverage density, For elevators The service floor collection, is the span of the service interval; determine whether each elevator serves the first floor (for example, the 1st floor) or the top floor (for example, the Lth floor) to identify the key vertical boundary coverage capacity; at the same time, identify whether there is a cross-fault structure in the service area, that is, whether there are continuous vacant intervals on the service floors, as a potential indicator of structural faults.

[0040] Elevator collaborative scheduling feature extraction includes for any pair of elevators , extract the intersection of its service floor sets , calculate the number of service overlapping floors ;for The floor position in the statistics is calculated, and its standard deviation is used to calculate whether the overlap is concentrated in a certain floor segment (such as the middle or lower part). The smaller the standard deviation, the more concentrated the overlap. and The frequency of simultaneous service requests on a shared floor. Determines whether the elevator pair belongs to the same group control system (e.g., shared scheduling logic or the same controller), represented by a Boolean feature (yes / no).

[0041] Constructing a coupling diagram for the floor elevator matrix data according to the structural coverage feature data and the elevator collaborative scheduling feature data to obtain coupling diagram data; Specifically, consider each elevator as a node in the graph, and for any elevator pair , if there is an intersection between their service floor sets (i.e. the number of overlapping floors ), then at the corresponding node and Create an undirected edge between . Assign an association weight to this edge , the weight value is composed of the following three weighted combinations: ,in is the edge weight between elevator nodes, is the weight data of the number of overlapping floors, is the number of overlapping floors, For elevators Service floor gathering, For elevators Service floor gathering, For collaborative scheduling frequency weight data, is the coordinated scheduling frequency, is the control structure consistency weight coefficient, is the control structure consistency index, is the elevator node sequence item, is another elevator node sequence item. The following attribute information is attached to each graph node, such as service coverage density, service area span, boundary coverage status, and fault identification auxiliary information.

[0042] Perform fault zone detection on the coupling diagram data to obtain authorization control diagram data, wherein the fault zone detection includes floor jump fault detection, floor no access record detection and authorization restriction detection; Specifically, for each elevator Corresponding service floor set , if there is any pair of adjacent floors satisfy , then the elevator is considered to have a floor jumping fault problem; if there is an edge in the graph that is emitted by the elevator node, and the elevator corresponding to the target node also has jumping behavior, then the edge weight is attenuated: ,in is the edge weight after jump layer attenuation, is the jump penalty coefficient, ranging from 0.5 to 0.8. is the original edge weight.

[0043] For all floors in the building , count whether it is covered by at least one elevator. It does not appear in the service set of all elevators, that is, , For elevators If the service floor set is larger than 1, it is determined that the floor has a "physical fault" phenomenon; for all adjacent floors serving the fault (such as or ) to restrict its participation in the service group clustering process of the graph and prevent incorrect connectivity judgment due to local service loss.

[0044] Extract the access control configuration for building elevators, including but not limited to floor access permissions for different roles, such as staff elevators, freight elevators, and visitor elevators. This data can come from the building control system configuration table or the elevator controller parameter set. For floors with restricted access permissions, mark the corresponding elevator-floor service pairs as "logical faults," allowing only robot tasks with specific permissions to traverse them. This structure forms "semi-fault" separations in the graph, requiring additional policy judgment during path generation. The identified fault regions are added to the original coupling graph as an attribute layer, marked as unreachable edges, nodes that do not participate in clustering, or logical boundaries requiring special policy processing.

[0045] Perform time period service mode switching constraints on the authorization control graph data to obtain constraint graph data; Specifically, according to the actual building operation rules and robot operation density, the scheduling time periods are divided, such as the morning peak period: 8:00 to 10:00 in the morning; the off-peak period: 10:00 to 16:00 in the afternoon; the evening peak period: 16:00 to 18:30 in the afternoon; the night low-frequency period: 19:00 to 6:00 the next day. During different time periods, some elevators execute different operation modes, including limited direction (such as only upward during the morning peak and only downward during the evening peak); suspension of service (such as closing some non-main elevators late at night for energy saving); and adjustment of service priority (such as giving priority to freight or visitor requests). Each edge in , define the corresponding active time period set ;like , indicating that the path is available during peak hours in the morning and evening; if an edge is reachable throughout the day, then If a time period is unreachable or disabled, the edge will be blocked or direction-restricted during that time period. If a time period only supports one-way traffic (for example, an elevator only goes up), the direction of the edge in the corresponding diagram must be set to "one-way edge", which means that only traffic from arrive If a path suspends service during a certain period of time, the edge is set as a "conditionally unreachable edge" or completely removed during that period; the "period dependency label" is recorded in the edge attributes.

[0046] Connected branches are extracted based on the constraint graph data to obtain the building elevator service group data.

[0047] Specifically, based on the time-labeled and directional-restricted graph structure generated in the previous step (referred to as the time-constrained graph), connectivity analysis is performed for each preset scheduling time period (e.g., morning peak, off-peak, and evening peak). Edges in the graph that are "valid" only within the current time period are considered available edges. Within this time period, all strongly connected subgraphs (i.e., any two nodes have a bidirectional reachable path) are extracted from the graph. Each strongly connected subgraph represents an elevator service unit with complete reachability within that time period. For each extracted connected subgraph, its service overlap index (e.g., number of shared floors, number of elevators, etc.) is calculated. If the service overlap of a subgraph falls below a preset threshold, it is determined to be a redundant service group or auxiliary path. Redundant subgraphs can be merged into adjacent primary service groups or marked as "auxiliary service groups" for low-frequency scheduling or abnormal backup path selection. During merging, time period consistency and direction consistency are maintained to avoid introducing disconnected structures across time periods.

[0048] Optionally, the path resource segment cutting includes: S21, dividing the robot target path data and the building elevator service group data into path segments to obtain path segmentation data; Specifically, it represents the complete path sequence of the robot from the starting floor to the target floor, which is recorded as the floor node sequence; the building elevator service group data represents a set composed of multiple elevator service units; each node in the traversal path , determine the service group to which it belongs ; If the floor node Belongs to the service group The set of service floors (i.e. ), the node is classified into The corresponding path segment. For the adjacent node pairs in the path ( ) for analysis, if they belong to different service groups (i.e. , ,and ), the position between the node pairs is used as the transfer breakpoint. The entire path is divided into several segments according to the transfer breakpoints. It includes the starting and ending floors; the elevator service group number to which this route belongs; such as the route direction (up / down), route length (expressed as floor difference or spatial distance), and estimated time window.

[0049] S22, extracting service attributes and transfer points based on the path division data to obtain service attribute data and transfer point data respectively; Specifically, for each path segment in the path division data, the following service attribute features are extracted from the building elevator management system, historical operation records or static configuration data to construct service attribute data, including the elevator group identification: recording the elevator service group ID on which this path segment depends; elevator type labeling: including classification information such as freight elevator, passenger elevator, high-speed elevator, low-speed elevator, etc.; control logic information: distinguish whether the elevator is in group control scheduling or independent operation mode; passage performance indicators: statistical historical data of the average passage time, peak waiting time, failure rate and other time-related indicators of this elevator group; robot accessibility: indicating whether the robot is allowed to pass through this path segment, for example, some elevators are limited to manual elevator riding, or are restricted by load, sensor limitations, etc. The boundary between path segments is a potential transfer point. The system extracts transfer information from the connection points of adjacent path segments and generates a transfer point dataset, including the location identifier and the floor number corresponding to the transfer point; service group connection pair: represents the two elevator service group pairs connected by the transfer point, which are used to identify the source and destination of the transfer path; transfer condition evaluation indicators, such as whether there is physical connectivity, such as whether they are on the same transfer platform, and whether there is an inter-floor passage; transfer channel length (meters), minimum transfer area width (meters); historical transfer congestion index (which can be defined as the number of pedestrians / robots passing per unit time); whether continuous robot path recognition is supported (such as barrier-free platform, no gate obstruction, etc.).

[0050] S23, performing transfer constraint dependency processing on the path division data according to the service attribute data and the transfer point data to obtain path dependency data; Specifically, dependency relationships are established, including time dependency marking processing, spatial mutual exclusion dependency marking processing (if the physical area where the transfer point is located is a narrow structure (such as a single channel or a narrow platform), the system prohibits two or more robots from occupying it at the same time, which is exclusive; the judgment basis includes the transfer point space width index, transfer congestion index, etc.), and resource interlocking dependency processing (when two path segments need to share a set of resources (such as elevators, access control, elevator scheduling slots, etc.), they need to be executed serially in a mutually exclusive scheduling manner). For each type of valid dependency relationship, the system constructs directed dependency edges between path segments.

[0051] S24. Construct a path dependency graph based on the path dependency data to obtain split path data.

[0052] Specifically, according to the path dependency relationship , build a line from point to directed edges; each edge is accompanied by a dependency type label: such as time dependency, resource lock dependency, spatial conflict, etc.; scheduling cost: such as waiting time estimation, resource competition risk; dynamic waiting valuation: if there is a congestion probability at the transfer point, the expected waiting time is estimated based on the current robot density and scheduling load; when there is a path segment going back and forth between high and low floors (such as high → low → high), the system performs graph traversal detection to identify whether a resource occupation deadlock loop is formed; for the detected potential deadlock path, the conflict loop path is recorded and a penalty weight is added or path avoidance processing is performed.

[0053] Optionally, the transit waiting estimation includes: S31. Perform a multi-robot arrival time superposition calculation based on the current transfer floor data and the split path data to obtain arrival superposition data; Specifically, the statistical time window length is set to (e.g. 5 seconds), the time axis is discretized to form a time segment sequence ; For each transfer floor , statistics in the time period The number of robots expected to arrive at the floor within the specified time period is defined as ,in For the moment to Arrival floor during The total number of robots, Number the variables for the robot, is an indicator function. If the condition in the expression is met (i.e. the arrival time of robot r falls within the current time window), the value of this item is 1, otherwise it is 0. is the robot arrival timestamp, is the transfer floor number, is the starting point of the current time slice, is the time window length. The above statistical operation is performed on all time segments t to form the multi-robot arrival density sequence of each floor, which is recorded as , the sequence reflects the transfer floors in different time periods Dynamic evolution of robot arrival density.

[0054] S32, performing kernel density calculation based on the arrival superposition data to obtain local aggregation data; Specifically, given a transfer floor Arrival time series , and its kernel density function is defined as follows: ,in For the moment When the robot reaches the transfer floor The local aggregation amount, is the total number of robots arriving at the floor, is the bandwidth parameter of the kernel density estimate, which represents the smoothing window size (e.g. h = 5 seconds), is the robot sequence term, is the kernel function, , is the circumference constant, is the base of natural logarithm, is the normalized time deviation, Estimate the time for the target, For: Robot Arrival floor For each transfer floor , output its local aggregation function in continuous time.

[0055] S33. Calculate the throughput and the light-opening door cycle based on the current transfer floor data to obtain throughput data and light-opening door cycle data respectively; Specifically, the transfer throughput refers to the maximum number of entities (such as robots) that can pass through a transfer floor under ideal scheduling conditions per unit time, which is limited by the carrying capacity and service cycle of the elevator or access control system connected to the floor. , if it is caused by (number of elevators) elevators or other transfer devices provide services, and the maximum load capacity of each device is , the average time it takes for an elevator to complete a service cycle is , then the maximum throughput of this floor It can be expressed as ,in is the maximum throughput of the floor, The maximum number of people that a single elevator can carry at one time (can be the number of people or robots). To serve the floor The number of elevators, The duration of a complete service cycle, including door opening, passenger entry / exit, door closing, and travel to the next landing. The door opening / closing cycle is estimated as the time it takes for an elevator to perform a complete door opening / closing operation. This is defined as the sum of the door opening time, passenger (or robot) entry / exit time, and door closing time. This parameter can be set empirically, obtained from historical operation logs, or real-time sensor data.

[0056] S34. Estimating the transfer waiting time of the local accumulation data according to the throughput data and the light gate cycle data to obtain transfer waiting estimation data.

[0057] Specifically, when the current gathering density exceeds the elevator's throughput capacity, the robot needs to queue up for multiple service cycles, resulting in a nonlinear increase in waiting time. At the moment The estimated waiting time is ,in For floor At the moment Estimated waiting time, For adjustment factors, for example, during peak hours (such as 08:00–09:00), the maximum value is set to 1.5, and during off-peak hours, it is set to 0.7. is the kernel density aggregation value in the current time window, is the maximum number of robots that can be processed per unit time, The time required to open, enter, exit, and close the door. , indicating that the current floor is in an overloaded state; in order to avoid short-term peak interference, the continuous time window Perform sliding average or Gaussian weighted smoothing; the output is floor A sequence of waiting time distribution estimates over future scheduling periods.

[0058] Optionally, the transfer map construction includes: Construct a path segment transfer map based on the current transfer floor data and the split path data to obtain the path segment transfer map data; Specifically, a heterogeneous two-layer graph structure is constructed, which consists of two types of nodes, including path segment nodes. Abstract as a path node ; Transfer resource node, transfer elevator on each floor Abstract as a transit node For each path segment , if its path requires passing through an elevator Realize from the floor arrive , then the following edges are established in the graph, such as pointing from the path segment node to the dependent elevator node , indicating that the path segment needs to dispatch an elevator ; From the elevator node to the next path segment node , indicating that the elevator transfer resource is available for use by the next route segment. If the transfer requires the coordination of multiple elevators (such as inter-zone transfers), a multi-hop dependency chain is constructed to achieve cascade connections between elevator nodes. The graph has heterogeneous node types, and edges have directional and functional meanings (dependencies, transfers). It is expanded into an attributed graph structure, with edges carrying additional information such as estimated time, floor displacement distance, and congestion weight. The graph structure represents the initial scheduling topology for the interaction between task path dependencies and transfer resources.

[0059] Assign edge weights to the path segment transfer graph data based on the transfer waiting estimation data to obtain weighted graph data; Specifically, for each directed edge in the graph, identify whether it belongs to one of the following two categories: the edge from the path segment node to the transfer node, indicating that the path segment depends on a certain elevator service; the edge from the transfer node to the next path segment node, indicating the continuation of the path after the transfer resource completes the service. For each identified edge, calculate its scheduling cost as the edge weight. The edge weight includes a static time-consuming part, including the elevator door opening time, the time for passengers to get on and off the elevator, the time for the elevator door to close and prepare for departure, etc., collectively referred to as the basic transfer time, the value of which can be obtained from historical operation data or preset standards; the dynamic waiting part, which comes from the estimated waiting time of the current transfer floor, indicating the queuing waiting time caused by the current transfer elevator being busy; the delay adjustment factor, to reflect the delay amplification effect of high-load or high-congestion floors, introduces an adjustment coefficient (between 1 and 2). The scheduling cost (i.e., edge weight) of each edge is expressed as: ,in is the scheduling cost of the edge (i.e., edge weight), is the static time-consuming part, is the delay adjustment factor, The dynamic waiting portion. All edge weights are positive real numbers; larger values ​​indicate higher expected scheduling costs. If a transfer floor is underloaded during the current period, its corresponding edge weight approaches the basic static time cost. If the edge connects to a highly congested floor or a single-channel transfer node, the dynamic portion increases significantly, indicating a potential scheduling bottleneck. The output is a weighted path segment transfer graph.

[0060] Directed scheduling is performed on the weighted graph data according to the weighted graph data to obtain transfer graph data.

[0061] Specifically, for all the “path segment node → transfer node → path segment node” structure patterns in the graph, the scheduling order dependency between the path segments is extracted; if there is a path segment With path segments All rely on a transit resource , and the scheduling logic requires Must precede Execution (for example, due to time windows, resource mutual exclusion or priority differences), a directed edge is constructed. Represents its execution constraints. All transit nodes (such as elevator nodes) are logically collapsed into edge constraints between path segments. Instead of retaining them as explicit nodes, their scheduling impact is reflected in the edge weights and dependencies between path segments. The scheduling delays and queuing effects introduced by transit resources are retained. The system performs graph simplification, such as merging continuous and uninterrupted path segments into a single node. It also identifies and marks possible dependency loops in the graph to prevent deadlocks in task scheduling. It also identifies all strongly connected subgraphs as independently schedulable units.

[0062] Optionally, the graph structure conflict calculation includes: Perform path occupancy mapping based on the transfer map data to obtain path occupancy data; Specifically, for each path segment, construct its resource occupancy description tuple, recorded as , where each Represents a path segment The occupancy of the jth resource node during execution is in the form of a triple: ,in Identifiers of occupied resource nodes, including but not limited to elevator shaft numbers, door area buffer identification, floor corridor numbers, etc. The time point when the path segment starts to occupy the resource; The time point at which the resource is released for the path segment. The path segment includes the elevator car resource, which represents the space occupied by the robot or passenger when moving vertically in the elevator; the elevator door buffer zone, which represents the area where the robot briefly stops before and after entering and exiting the elevator door area; and the floor corridor, which represents the path area occupied by the robot when moving horizontally between floors. It can be obtained by calculating the estimated start and end time of the scheduling path; if waiting or congestion delay is included, the occupancy window of the corresponding resource should include the estimated waiting time period.

[0063] Conflict pairs are extracted according to the path occupancy data to obtain conflict pair data; Specifically, for all path segments ,in , perform pair-by-pair comparison and extract the resource node sets and corresponding time windows occupied by each of them. , if all the following conflict judgment conditions are met, a conflict pair is formed: there is a resource node l, so that the node appears at the same time and In the resource usage description, that is: , The resource collection occupied by the path segment. is the occupation of resource nodes by the path segment; the occupation time on the same resource node has an intersection, that is, ,in For path segments The start and end time on the resource, For path segments The start and end time on the resource; based on the resource's shareability and scheduling strategy, the conflict type is classified and labeled. If the conflicting resource is a strictly mutually exclusive resource (such as an elevator car), it is marked as "strong conflict"; if the conflicting resource is a partially shared resource (such as a floor corridor), it is marked as "weak conflict"; other types of resources can be expanded and defined according to actual system rules. For path segment pairs that meet the above conditions, a conflict pair tuple is constructed. ,in Conflict type label.

[0064] The path overlap tension is calculated based on the conflict data to obtain the path conflict data.

[0065] Specifically, for each set of conflicting pairs , based on its shared resource collection , define the path conflict tension function as follows: ,in is the path conflict tension data, For shared resource nodes, For path segments and A collection of shared resources, is the overlap ratio weight coefficient, For resource nodes The overlap ratio of the path segment space occupied, for example, the ratio of the overlapping length to the total length of the corridor resource, is the weight coefficient of the overlap ratio, For resource nodes The overlap ratio of the occupied time intervals is calculated as the ratio of the time intersection length of the two path segments at the resource node to the length of any segment. For resources with fuzzy physical boundaries or buffer areas, the Gaussian kernel function is used to model the conflict impact radius; spatial overlap and temporal overlap are embedded in the corresponding kernel function for smoothing. The tension values ​​of all path segment pairs constitute the conflict tension matrix , each element Represents a path segment and The intensity of the conflict between them.

[0066] Optionally, the path overlap tension calculation includes: According to the conflict, the non-rigid geometric overlap and time occupancy intersection of the data are calculated to obtain fuzzy boundary overlap data and time occupancy intersection data respectively; Specifically, in a three-dimensional environment, the path segment 、 The occupied space is constructed as a fuzzy envelope area, and a fuzzy function (such as Gaussian function) is used to define its spatial influence range. In the corridor segment and elevator shaft, the spatial influence function of each path segment is constructed: ,in, Represents a spatial position point, and the function value represents the intensity of the influence of the path segment on the position, which is a function of the distance from the center trajectory. The overlap integral is the non-rigid geometric overlap of two path segments on the resource node, which is the integral of the product of their fuzzy occupancy functions in the spatial domain: ,in is the fuzzy boundary overlap, is the spatial position point, For shared resource space domain, For path segments The spatial influence function of , is the natural exponential function, For path segments The center trajectory position of is the kernel function bandwidth parameter, which indicates the path influence radius, For path segments The spatial influence function is similar to the previous one. This value represents the degree of spatial overlap or interference between two path segments. A larger value indicates more severe spatial overlap. The time occupancy intersection is the ratio of the intersection length of two time windows to their respective shortest durations: ,in is the time occupancy intersection, is the maximum value function, is the minimum function, For path segments The starting occupation time, For path segments The end of the occupation time, For path segments The starting occupation time, For path segments The result is 0 if the time intervals do not overlap, and 1 if they completely overlap.

[0067] According to the fuzzy boundary overlap data and the time occupancy intersection data, the shared resource node competition process is performed to obtain the node competition data; Specifically, for each conflicting pair Shared resource nodes in , calculate the unit resource competition density on the node, which is defined as follows: ,in is the node resource unit competition intensity, For path segments and In Resources The degree of spatial overlap of the fuzzy boundaries on For path segments and In Resources The time occupancy intersection on Resource Node The type weight is (e.g., elevator car is 1.0, elevator door buffer is 0.7, floor corridor is 0.5, elevator door opening and closing mechanism is 0.6, transfer intersection is 0.4, and general waiting area is 0.3). The competition intensity of the path segment on all its shared resource nodes is accumulated to obtain its node competition score: ,in For path segment pairs The total resource competition pressure score, For a single resource node, is a collection of shared resource nodes. is the unit competition intensity of node resources.

[0068] Propagate resource occupancy tension based on node competition data to obtain occupancy propagation matrix data; Specifically, the path segment and resource nodes Establish a bidirectional connection relationship; if the path segment is Shared resource nodes exist , then treat it as an edge in the bipartite graph; each edge The edge weight is set as the local competition intensity , that is, the path segment is at the resource node The unit competition on the graph is affected. Introducing the graph propagation function Perform weight transformation on conflict intensity to control propagation attenuation (e.g. ,in is the attenuation factor); for each path segment , count its differences from all other path segments The cumulative tension value of propagation between is defined as follows: ,in For path segments The resource occupation tension value, is the path order item , is the path order item , is the result of the propagation function of competition intensity, which indicates the contribution of conflict intensity to tension. To indicate a path segment and Whether there is an association in the scheduling graph (such as shared resources or execution order dependencies), if so, it is 1, otherwise it is 0, is the global path segment conflict strength. The output is the path segment tension propagation matrix , each element represents the cumulative resource conflict pressure that the path segment bears in the current scheduling structure.

[0069] Based on the occupied propagation matrix data, gravity control is performed to schedule high-priority tasks and obtain path conflict data.

[0070] Specifically, for each path segment , the system reads the scheduling priority value of its corresponding task The value can come from the task definition library, manual configuration or intelligent classification results; the higher the priority value, the more important the task is, and the more weight it should receive in resource occupation of the scheduling path. Perform priority weighted correction to obtain the adjusted tension value ,in For path segments Conflicting tension after considering task priorities, is the path segment conflict tension, is the priority gravity adjustment coefficient, which controls the sensitivity of the scheduling strategy to the task level. , then the priority 3 task will get an additional 60% scheduling tension gravity based on the tension propagation. is the priority level value of the task corresponding to the path segment. The corresponding adjusted tension value Combine and build a path conflict data set ,This set will serve as the basic input for subsequent path segment sorting priority determination, conflict avoidance strategy adjustment and multi-task scheduling optimization.

[0071] Optionally, the schedule generation includes: S41, identifying conflict core groups based on the path conflict data to obtain conflict core group data; Specifically, based on the path conflict data , build a conflict graph , where the node set is all path segments involved in scheduling; edge set is a pair of paths with conflicting relationships; each edge The edge weight is the corresponding tension value . For the conflict graph that has been constructed , a graph clustering algorithm is used to divide the conflict coupling paths into groups and identify local tension-intensive areas; each node constitutes a community separately; try to move the current node into the community where the neighbor is located, and merge it if it can maximize the local modularity gain; the modularity function is based on edge weights and automatically clusters high-tension paths into the same conflict community; terminate when the modularity no longer increases significantly, and output several conflict core groups.

[0072] S42. Perform path-level scheduling sorting based on the conflicting core group data to obtain scheduling sorting data; Specifically, for each path segment , calculate its scheduling priority , specifically ,in To schedule priority data, is the priority score weight coefficient, which is set to 0.5. is the priority score of the task corresponding to the path (for example, the priority of the drug delivery task is set to 5, and the priority of the routine inspection task is set to 2). is the total conflict tension weight coefficient, which is 0.3. is the total collision tension of the path segment with other paths in the current group, is the weight coefficient of the estimated cumulative waiting time, which is set to 0.5. is the estimated cumulative waiting time of the path segment in the current transfer structure. Sort from high to low. If the scores are similar (the difference is below a set threshold), additional sorting criteria can be introduced, such as total path length, whether the path crosses floors (cross-floor paths are prioritized), task initiation time (first come, first served), etc. Each conflict core group is then matched with a set of sorted path scheduling sequences.

[0073] S43. Generate robot scheduling data according to the scheduling sequence to perform multi-robot collaborative scheduling operations.

[0074] Specifically, each path segment is assigned a planned execution time window that meets the following conditions: It does not overlap with other path segments in the same group in terms of resource usage (to avoid occupancy conflicts); the robot's state before assignment is considered (such as whether it is still executing the previous task or is in an elevator); if the path segment involves elevator transfers, it is necessary to confirm that the elevator resources are dispatchable within the estimated time window to avoid repeated elevator occupancy; and reasonable allocation is made based on the estimated path time (such as floor access time, waiting time, and elevator usage delay). For each robot, a scheduling execution plan is generated based on its assigned path segment and corresponding time window. This plan should include a structured set of operation instructions, including instruction types such as START: Start execution of the scheduled task; MOVE_TO: Move to the target location; WAIT: Wait due to unavailable resources (such as elevators or corridors); ENTER_ELEVATOR / EXIT_ELEVATOR: Elevator entry and exit actions; and FINISH: Complete the path segment task. Each instruction should include the operation type; start and end time; location identifier or resource ID; supplementary reason (such as waiting for resource occupation to end) or additional information (such as target floor, robot number, etc.).

[0075] Optionally, the present application further provides a multi-robot cloud scheduling collaborative system for executing the multi-robot cloud scheduling collaborative method described above, wherein the multi-robot cloud scheduling collaborative system includes: An elevator service group division module is used to obtain building elevator group data and divide building elevator service groups according to the building elevator group data to obtain building elevator service group data; The path segment structure modeling module is used to obtain the robot's target path data; the robot's target path data and the building elevator service group data are cut into path resource segments to obtain split path data; The transfer waiting prediction module is used to obtain the current transfer floor data and perform transfer waiting estimation based on the current transfer floor data and the split path data to obtain transfer waiting estimation data; The conflict-aware scheduling generation module is used to construct a transfer graph based on the transfer waiting estimation data to obtain transfer graph data; perform graph structure conflict calculations based on the transfer graph data to obtain path conflict data; and perform scheduling generation based on the path conflict data to obtain robot scheduling data for multi-robot collaborative scheduling operations.

[0076] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes that fall within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0077] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A multi-robot cloud scheduling collaborative method, characterized in that: Applied to a cloud platform, the method includes: S1. Obtain building elevator group data, and divide building elevator service groups according to the building elevator group data to obtain building elevator service group data; S2. Obtain robot target path data; perform path resource segmentation on the robot target path data and the building elevator service group data to obtain split path data; S3. Obtain current transfer floor data, and perform transfer waiting estimation based on the current transfer floor data and the split path data to obtain transfer waiting estimation data; S4. Construct a transfer graph based on the transfer waiting estimation data to obtain transfer graph data; perform graph structure conflict calculation based on the transfer graph data to obtain path conflict data; perform scheduling generation based on the path conflict data to obtain robot scheduling data for multi-robot collaborative scheduling operations.

2. The method according to claim 1, characterized in that The building elevator service group division includes: Extract the floor elevator service relationship based on the building elevator group data to obtain the floor elevator matrix data; Obtaining elevator operation behavior data, and calculating service overlap frequency based on the elevator operation behavior data to obtain service overlap frequency data; The floor elevator connectivity graph is constructed based on the floor elevator matrix data and the service overlapping frequency data to obtain the building elevator service group data.

3. The method according to claim 2, characterized in that The construction of the floor elevator connectivity graph includes: Structural coverage feature extraction and elevator collaborative scheduling feature extraction are performed based on the floor elevator matrix data and the service overlapping frequency data, and structural coverage feature data and elevator collaborative scheduling feature data are obtained respectively; Constructing a coupling diagram for the floor elevator matrix data according to the structural coverage feature data and the elevator collaborative scheduling feature data to obtain coupling diagram data; Perform fault zone detection on the coupling diagram data to obtain authorization control diagram data, wherein the fault zone detection includes floor jump fault detection, floor no access record detection and authorization restriction detection; Perform time period service mode switching constraints on the authorization control graph data to obtain constraint graph data; Connected branches are extracted based on the constraint graph data to obtain the building elevator service group data.

4. The method according to claim 1, wherein The path resource segment cutting includes: Divide the robot target path data and the building elevator service group data into segments to obtain path division data; Perform service attribute extraction and transfer point extraction based on the path division data to obtain service attribute data and transfer point data respectively; Perform transfer constraint dependency processing on the path division data according to the service attribute data and the transfer point data to obtain path dependency data; A path dependency graph is constructed based on the path dependency data to obtain split path data.

5. The method according to claim 1, wherein The transit wait estimate includes: The arrival time of multiple robots is superimposed and calculated based on the current transfer floor data and the split path data to obtain the arrival superposition data; Kernel density calculation is performed based on the arrival superposition data to obtain local aggregation data; Throughput calculation and light-opening door cycle calculation are performed based on the current transfer floor data to obtain throughput data and light-opening door cycle data respectively; The transfer waiting time of the local aggregated data is estimated based on the throughput data and the light gate cycle data to obtain the transfer waiting estimation data.

6. The method according to claim 1, characterized in that The transfer map construction includes: Construct a path segment transfer map based on the current transfer floor data and the split path data to obtain the path segment transfer map data; Assign edge weights to the path segment transfer graph data based on the transfer waiting estimation data to obtain weighted graph data; Directed scheduling is performed on the weighted graph data according to the weighted graph data to obtain transfer graph data.

7. The method according to claim 1, characterized in that The graph structure conflict calculation includes: Perform path occupancy mapping based on the transfer map data to obtain path occupancy data; Conflict pairs are extracted according to the path occupancy data to obtain conflict pair data; The path overlap tension is calculated based on the conflict data to obtain the path conflict data.

8. The method according to claim 7, characterized in that The path overlap tension calculation includes: According to the conflict, the non-rigid geometric overlap and time occupancy intersection of the data are calculated to obtain fuzzy boundary overlap data and time occupancy intersection data respectively; According to the fuzzy boundary overlap data and the time occupancy intersection data, the shared resource node competition process is performed to obtain the node competition data; Propagate resource occupancy tension based on node competition data to obtain occupancy propagation matrix data; Based on the occupied propagation matrix data, gravity control is performed to schedule high-priority tasks and obtain path conflict data.

9. The method according to claim 1, characterized in that The schedule generation includes: Identify conflict core groups based on path conflict data to obtain conflict core group data; Perform path-level scheduling sorting based on conflict core group data to obtain scheduling sorting data; Generate robot scheduling data based on scheduling sorting to perform multi-robot collaborative scheduling operations.

10. A multi-robot cloud scheduling collaborative system, characterized in that: For executing the multi-robot cloud scheduling collaborative method according to claim 1, the multi-robot cloud scheduling collaborative system comprises: An elevator service group division module is used to obtain building elevator group data and divide building elevator service groups according to the building elevator group data to obtain building elevator service group data; The path segment structure modeling module is used to obtain the robot's target path data; the robot's target path data and the building elevator service group data are cut into path resource segments to obtain split path data; The transfer waiting prediction module is used to obtain the current transfer floor data and perform transfer waiting estimation based on the current transfer floor data and the split path data to obtain transfer waiting estimation data; The conflict-aware scheduling generation module is used to construct a transfer graph based on the transfer waiting estimation data to obtain transfer graph data; perform graph structure conflict calculations based on the transfer graph data to obtain path conflict data; and perform scheduling generation based on the path conflict data to obtain robot scheduling data for multi-robot collaborative scheduling operations.

Images