Multi-robot scheduling management method and system based on cloud computing

Abstract

The invention relates to the technical field of resource management and distribution, and discloses a multi-robot scheduling management method and system based on cloud computing, and the method comprises the steps: obtaining real-time state data and a scheduling instruction of each robot in a region, and uploading the real-time state data and the scheduling instruction to a cloud through a data channel; the cloud aggregates real-time state data of the robot in the area, establishes a virtual twin digital model of the robot, and maps the real-time state of the physical robot in a map of the area; the cloud carries out calculation power transfer and distribution according to the evaluation before simulation; after distribution is completed, simulation of dispatching instructions is conducted on the robots through machine learning; and optimizing the scheduling instruction according to the simulation condition of the cloud. And the response speed of task scheduling and the resource utilization efficiency are effectively improved. Through combination of dynamic computing power distribution and a machine learning simulation technology, the problem of robot path intersection conflict is effectively solved, the waiting and delay time of the robot is remarkably shortened, and the execution efficiency and reliability of the whole system are improved.

Description

Technical Field

[0001] This invention relates to the technical field of resource management and allocation, specifically to a multi-robot scheduling and management method and system based on cloud computing. Background Technology

[0002] With the rapid development and widespread adoption of robotics technology, multi-robot collaborative systems have been increasingly applied in various fields such as industrial production, warehousing and logistics, environmental monitoring, disaster relief, and smart city construction. Compared to single-robot operations, multi-robot collaboration offers higher efficiency, flexibility, and reliability. However, multi-robot systems present complex challenges such as resource allocation, task scheduling, path planning, and coordinated control. The complexity of scheduling and management increases significantly, especially in environments with a large number of robots, dense tasks, or frequent path intersections.

[0003] Traditional scheduling methods are often based on preset rules or a single computing node, which can easily lead to unbalanced resource load, reduced task execution efficiency, path conflicts or resource competition between robots.

[0004] Furthermore, traditional static allocation strategies lack dynamic adaptability and are difficult to efficiently cope with real-time changes in task requirements and robot states. Therefore, there is an urgent need to explore and develop more advanced robot scheduling and management technologies. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a multi-robot scheduling and management method based on cloud computing, comprising:

[0007] Acquire real-time status data and scheduling instructions for each robot within the area, and upload them to the cloud via a data channel;

[0008] The cloud aggregates real-time status data of robots within a region, establishes a virtual twin digital model of the robots, and maps the real-time status of the physical robots onto a map of the region.

[0009] Based on the pre-simulation assessment, the cloud platform allocates computing power; after allocation, machine learning is used to simulate scheduling commands for the robot.

[0010] The scheduling instructions are optimized based on the simulation results in the cloud.

[0011] Once the optimization is confirmed, the solution is directly issued to the robot, and the robot directly updates its scheduling instructions.

[0012] As a preferred embodiment of the cloud computing-based multi-robot scheduling and management method described in this invention, the cloud includes multiple cloud nodes, with each region corresponding to each cloud node.

[0013] The real-time status data includes the robot's own position, speed, battery level, and energy consumption data.

[0014] As a preferred embodiment of the cloud computing-based multi-robot scheduling and management method described in this invention, the scheduling instruction includes the execution time of the scheduling task, the latest completion time, and the path planning result for the robot.

[0015] As a preferred embodiment of the cloud computing-based multi-robot scheduling and management method of the present invention, the path planning of each robot is depicted on a map according to the virtual twin digital model of the robot; in the map, attributes are marked for each coordinate position, including whether it is a drivable area and path marking;

[0016] Let x be the x-coordinate of any position on the map, and y be the y-coordinate of any position on the map; if the path of robot i passes through the coordinate position (x, y), then add a path marker to the coordinate position (x, y); different path markers can be superimposed on the same coordinate position.

[0017] If there is only one path marker at coordinate position (x, y), the marker content is empty; if there is more than one path marker at coordinate position (x, y), the marker content is: the estimated time interval for robot i to reach coordinate position (x, y);

[0018] The pre-simulation evaluation includes analyzing the coordinate positions (x, y) where the path markers are not empty: if the expected time intervals of the coordinate positions (x, y) overlap, the overlap coefficient of the coordinate positions (x, y) is assigned a value equal to the cumulative number of overlaps: among all path markers of the coordinate positions (x, y), any two expected time intervals are selected, and if they overlap each time, the cumulative value is increased by 1; if the expected time intervals of the coordinate positions (x, y) do not overlap, the overlap coefficient of the coordinate positions (x, y) is assigned a value of 0.

[0019] As a preferred embodiment of the cloud computing-based multi-robot scheduling and management method described in this invention, the allocation of computing power includes associating coordinate positions with non-empty marked content, packaging the data according to the association results, allocating it to other cloud nodes, and simulating scheduling instructions through other cloud nodes.

[0020] The association process is as follows:

[0021] The marked coordinates serve as the link between different robot paths;

[0022] Paths with direct or indirect connections are extracted to obtain multiple sets of associated paths;

[0023] For any set of associated paths j, in the virtual twin digital model, other paths outside of the set of associated paths j are eliminated, and the eliminated virtual twin digital model is packaged to obtain packaged data about the set of associated paths j.

[0024] There is no kinetic link between different sets of associated paths.

[0025] As a preferred embodiment of the cloud computing-based multi-robot scheduling and management method described in this invention, the allocation of computing power further includes: if the total overlap coefficient in the packaged data of the associated path set j is 0, then no allocation is performed and the original scheduling instruction is maintained.

[0026] If the total overlap coefficient in the packaged data of the associated path set j is not 0, then the sum of the number of links and the overlap coefficients in the associated path set j is used as the allocation reference coefficient; and cloud nodes are allocated according to the magnitude of the allocation reference coefficient.

[0027] During allocation, if the number of available cloud nodes is greater than or equal to the number of associated path sets, the allocation reference coefficients are allocated sequentially from largest to smallest according to the remaining computing power of each cloud node.

[0028] If the number of available cloud nodes is less than the number of associated path sets, the associated path sets are combined into an optimal hybrid path set. The allocation reference coefficients of the optimal hybrid path set are allocated sequentially from largest to smallest according to the remaining computing power of each cloud node from largest to smallest.

[0029] In this process, the associated path set is randomly combined to obtain multiple mixed path sets, and the optimal mixed path set is calculated through the objective function and constraints.

[0030] The objective function is to make the ratio of remaining computing power of each available cloud node as close as possible to the ratio of the allocation reference coefficient of the hybrid path set; the constraint is that the number of hybrid path sets equals the number of available cloud nodes.

[0031] The allocation reference coefficient of the mixed path set is quantified as the sum of the allocation reference coefficients of the associated path sets in the combination.

[0032] As a preferred embodiment of the cloud computing-based multi-robot scheduling and management method of the present invention, the simulation of the scheduling instructions includes, by using machine learning, adding instructions so that each robot can wait at the position before reaching each link.

[0033] The scheduling process is simulated by adding new instructions, and the process of adding instructions and simulating the scheduling process is repeated.

[0034] Optimizing the scheduling instructions includes setting the minimum sum of waiting times in the scheduling instructions as the objective; using the latest completion time of each robot as a constraint; and selecting the optimal addition instruction.

[0035] The optimal addition instruction is added to the original scheduling instruction to complete the optimization.

[0036] A cloud-based multi-robot scheduling and management system employing the method described in this invention, wherein:

[0037] The data acquisition unit obtains real-time status data and scheduling instructions for each robot within the area and uploads them to the cloud via a data channel. The cloud aggregates the real-time status data of the robots within the area, establishes a virtual twin digital model of the robots, and maps the real-time status of the physical robots onto the map of the area.

[0038] The cloud-based allocation unit redistributes computing power based on the pre-simulation assessment; after allocation, machine learning is used to simulate scheduling commands for the robot.

[0039] The optimization unit optimizes the scheduling instructions based on the simulation results in the cloud.

[0040] Once the control unit has been optimized and confirmed, the solution is directly issued to the robot, and the robot directly updates its scheduling instructions.

[0041] A computer device includes: a memory and a processor; the memory stores a computer program, wherein: when the processor executes the computer program, it implements the steps of the method described in any one of the present invention.

[0042] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in any one of the present invention.

[0043] The beneficial effects of this invention are as follows: The cloud-based multi-robot scheduling and management method provided by this invention achieves real-time and precise management and optimization of multi-robot systems through cloud computing and virtual twin technology, effectively improving the response speed and resource utilization efficiency of task scheduling. The combination of dynamic computing power allocation and machine learning simulation technology effectively solves the problem of robot path intersection conflicts, significantly reduces robot waiting and delay time, and improves the overall system execution efficiency and reliability. In addition, the path overlap analysis and node association strategy proposed in this invention makes the scheduling process more intelligent and precise, greatly reducing the risk of on-site task execution. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 The first embodiment of the present invention provides an overall flowchart of a cloud computing-based multi-robot scheduling and management method. Detailed Implementation

[0046] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0047] Example 1, referring to Figure 1 As an embodiment of the present invention, a multi-robot scheduling and management method based on cloud computing is provided, comprising:

[0048] S1: Acquire real-time status data and scheduling instructions for each robot within the area, and upload them to the cloud via the data channel.

[0049] It's important to understand that "area" here refers to a specific space or geographical area within a particular task execution environment where multiple robots collaborate to complete a task. This is typically a clearly defined work scenario or site. Examples include: factory workshops, logistics warehouses, disaster relief sites, a specific urban area, or other limited locations requiring multiple robots to collaborate on a task. Clearly defining the area facilitates management and optimizes robot scheduling, ensuring the task is carried out efficiently, coordinated, and in an orderly manner.

[0050] Furthermore, the cloud contains multiple cloud nodes, with each region corresponding to each cloud node.

[0051] The real-time status data includes the robot's position, speed, battery level, and energy consumption. The scheduling instructions include the task's execution time, latest completion time, and the robot's path planning results. The latest completion time provides a time constraint for task execution, ensuring the task is completed within the specified time. This avoids delays causing a chain reaction on subsequent tasks, effectively guaranteeing the timeliness and stability of the overall task. Path planning determines the robot's route during task execution and is a core factor affecting robot collaboration. The path planning results enable the robot to perform tasks efficiently and safely, preventing path conflicts or collisions between robots and optimizing on-site work order.

[0052] S2: The cloud aggregates real-time status data of robots within the region, establishes a virtual twin digital model of the robot, and maps the real-time status of the physical robot onto the map of the region.

[0053] Based on the virtual twin digital model of the robot, the path planning of each robot is depicted on a map; specifically, a two-dimensional digital map is constructed in the model, and the robot's movement trajectory is accurately projected onto this map. In the map, attributes are marked for each coordinate position, including whether it is a traversable area and a path marker.

[0054] Specifically, let x be the horizontal coordinate of any position on the map and y be the vertical coordinate of any position on the map; if the path of robot i passes through the coordinate position (x, y), then add a path marker to the coordinate position (x, y); different path markers can be superimposed on the same coordinate position.

[0055] Furthermore, if there is only one path marker at coordinate position (x, y), the marker content is empty; note that there is a marker, but its content is empty. If there is more than one path marker at coordinate position (x, y), the marker content is: the estimated time interval for robot i to reach coordinate position (x, y).

[0056] Minimum value of the interval (time point): .

[0057] Maximum value of the interval (time): .

[0058] in, This indicates the current moment (if the robot is not started). T1 represents the time when the task starts; if the robot has already started, then... (Indicates the current time). This represents the length of the path between the current position and the coordinate position of robot i; This represents the minimum speed of robot i; This represents the maximum speed of robot i.

[0059] The pre-simulation evaluation includes analyzing the coordinate positions (x, y) where the path markers are not empty: if the expected time intervals of the coordinate positions (x, y) overlap, the overlap coefficient of the coordinate positions (x, y) is assigned a value equal to the cumulative number of overlaps: among all path markers of the coordinate positions (x, y), any two expected time intervals are selected, and if they overlap each time, the cumulative value is increased by 1; if the expected time intervals of the coordinate positions (x, y) do not overlap, the overlap coefficient of the coordinate positions (x, y) is assigned a value of 0.

[0060] Specifically: "Cumulative Value" The calculation formula is: .

[0061] Where N represents the number of tags, and n and m are both tag indices, with m and n being unequal. An indicator function for determining the overlap of two intervals. Defined as:

[0062]

[0063] in, This represents the minimum value in the interval among the nth markers; This represents the maximum value in the interval of the nth marker; This represents the minimum value in the interval among the m-th markers; This represents the maximum value in the interval of the m-th marker.

[0064] It's important to note that while multiple robots may pass through the same coordinate point, resource competition only arises when their traversal times overlap. This design precisely distinguishes between the concepts of "marker overlap" and "time conflict," giving path markings stronger temporal semantics, which facilitates subsequent simulation and scheduling analysis. The overlap coefficient is calculated by traversing all path time interval combinations and using a judgment function to determine whether each group overlaps. The result is a measurable conflict level value, facilitating further resource allocation.

[0065] S3: The cloud allocates computing power based on the pre-simulation assessment; after allocation, machine learning is used to simulate scheduling instructions for the robot.

[0066] The allocation of computing power includes associating coordinates where the marked content is not empty, packaging the data according to the association results, and distributing it to other cloud nodes, simulating scheduling instructions through these other cloud nodes. Each node does not participate in the instruction adjustment for its corresponding region; this is because if a node itself participates in the adjustment of its jurisdiction, data packets will be preferentially allocated to the node itself, meaning it has no additional resources for real-time monitoring, real-time data updates, or analysis of updated data. Allocating all resources outwards effectively avoids these problems; furthermore, it avoids contention for multiple types of data on a single node. Similarly, the node's remaining computing power will also participate in the computation of data packets from other nodes.

[0067] The association process is as follows:

[0068] The marked coordinates are used as a link between different robot paths.

[0069] Paths with direct or indirect connections are extracted to obtain multiple sets of associated paths. Direct connections exist when two robot paths P1 and P2 share at least one coordinate on the map. Indirect connections exist when paths P1 and P2 may not share coordinates, but another path P3 is connected to both P1 and P2. Indirect connections also exhibit transitivity.

[0070] For any set of associated paths j, in the virtual twin digital model, all paths outside of set j are eliminated. The eliminated virtual twin digital model is then packaged to obtain packaged data about set j. There are no interrelationships between different sets of associated paths.

[0071] Furthermore, if the total overlap coefficient in the packaged data of the associated path set j is 0, no allocation will be made, and the original scheduling instruction will be maintained.

[0072] If the total overlap coefficient in the packaged data of the associated path set j is not 0, then the number of links in the associated path set j multiplied by the sum of the overlap coefficients (multiplying the sum of the overlap coefficients at each link by the number of links; in this way, the complexity of the instruction adjustment is expressed) is used as the allocation reference coefficient; and cloud nodes are allocated according to the magnitude of the allocation reference coefficient.

[0073] During allocation, if the number of available cloud nodes is greater than or equal to the number of associated path sets, the allocation reference coefficients are applied sequentially from largest to smallest based on the remaining computing power of each cloud node. In practice, the goal of this scheme is to allocate path sets with more severe conflicts and higher computational demands to cloud nodes with more sufficient computing power, ensuring that high-complexity tasks are prioritized for processing by high-performance nodes, thus improving overall simulation efficiency, when the number of cloud nodes is greater than or equal to the number of associated path sets. The node list C is sorted in descending order (from strongest to weakest computing power); the path set list R is also sorted in descending order (from highest to lowest demand); the order of the two lists corresponds sequentially from front to back as follows:

[0074]

[0075] This represents the nth element in the path set list; This represents the nth element in the list of nodes.

[0076] If the number of available cloud nodes is less than the number of associated path sets, the associated path sets are combined into an optimal hybrid path set. The allocation reference coefficients of the optimal hybrid path set are then allocated sequentially from largest to smallest according to the remaining computing power of each cloud node (the allocation principle is the same as the corresponding principle mentioned above).

[0077] In this process, the associated path set is randomly combined to obtain multiple mixed path sets, and the optimal mixed path set is calculated through the objective function and constraints.

[0078] The objective function is to make the ratio of the remaining computing power of each available cloud node as close as possible to the ratio of the allocation reference coefficient of the hybrid path set; the constraint is that the number of hybrid path sets equals the number of available cloud nodes.

[0079] The allocation reference coefficient of the mixed path set is quantified as: the sum of the allocation reference coefficients of the associated path sets in the combination: Let there be a total of There are 1 set of associated paths, respectively Each path set The reference coefficient is ;existing One available cloud node, , needs to be generated A set of mixed paths; the set of mixed paths is denoted as . (It can contain only one set of associated paths), where Each And each Belongs to only one For the first A set of mixed paths Its allocation reference coefficient The calculation is as follows:

[0080]

[0081] in: Represents a set of mixed paths The allocation reference coefficient; Represents the set of associated paths Reference coefficients; Represents a set of mixed paths The set of paths contained therein.

[0082] The objective function used to select the "optimal mixed force scheme" can be expressed as:

[0083]

[0084] in: Indicates the first The remaining computing power of each cloud node; Indicates the reference coefficients for the corresponding mixed path set; This represents the total computing power of all available cloud nodes; This represents the total reference coefficient for all mixed path sets.

[0085] The constraint is: generate M mixed path sets, i.e., the value of e ranges from 1 to M.

[0086] It's important to note that in multi-robot scheduling systems, path conflicts between robots often exhibit a highly uneven distribution: some path sets are densely conflicted and computationally demanding, while others are relatively simple. The available computing power of cloud nodes also exhibits volatility and heterogeneity. Therefore, if the number of cloud nodes is less than the number of associated path sets, allocating resources using a "single path set → single node" approach will lead to resource misallocation, wasting computing power and potentially preventing timely optimization of highly conflicting tasks. By merging multiple associated path sets into "hybrid path sets" using a specific combination strategy, each combined unit has controllable scheduling complexity and resource requirements, thus better adapting to the upper limit of limited cloud resources. The ratio difference between the reference coefficient of each hybrid path set and the computing power of cloud nodes is minimized. In other words, the closer the "task demand ratio" is to the "resource supply ratio," the better, improving system scheduling efficiency and execution stability from a global perspective.

[0087] If a simple allocation method is used directly, some nodes may be overloaded, delaying the simulation completion and affecting the overall system efficiency. This design, however, optimizes the combination of path sets to achieve a relatively balanced load on each cloud node, thereby significantly improving parallel computing efficiency and avoiding bottleneck nodes. Furthermore, by randomly generating multiple combination schemes and selecting the optimal combination based on the objective function and constraints, this method is applicable to different numbers of path sets, different types of conflict distributions, and dynamically changing cloud node resources, exhibiting high versatility and scalability.

[0088] S4: Optimize the scheduling instructions based on the simulation results in the cloud.

[0089] The simulation of scheduling instructions includes using machine learning to add instructions (which are actually wait instructions) so that each robot can wait at its position before reaching each link. It's important to note that wait instructions are used to mitigate path conflicts; that is, if two or more robots are expected to arrive at the same coordinate point (link) simultaneously, a potential collision or resource contention may occur. To avoid such conflicts, the system can manually "make a robot wait."

[0090] In this invention, machine learning is an intelligent decision-making model for path conflict scheduling optimization. Its main function is to identify the path locations requiring additional waiting instructions during scheduling simulations and automatically determine the duration and strategy of such additions to achieve conflict resolution, path smoothing, and overall efficiency optimization. The core objective of this machine learning system is as follows: to gradually reduce path conflicts by intelligently adding waiting instructions until all overlap coefficients are zero. Instruction addition location: the "previous position" of all nodes with non-zero overlap coefficients in each path. Simulation evaluation criteria: after each addition of a waiting instruction, the entire path execution process is re-simulated, and the overlap coefficients of all path marker coordinates are calculated. Convergence condition: the overlap coefficient of all marker points is zero.

[0091] Input features: For each conflicting node (x, y) in the robot's current path, the model will consider: the current path structure (path length, node order); node overlap coefficient; the expected arrival time interval of the node; the position where instructions can be inserted (usually the previous node); the current time and the remaining scheduling window (latest completion time).

[0092] Output target: Location of the added instruction: i.e., which predecessor node; Waiting time: Recommended waiting time Δt, in seconds or time step.

[0093] The loop process is as follows:

[0094] Initial detection: Traverse all coordinate positions in the virtual twin map. Identify nodes with an overlap coefficient > 0. For each such node, record its corresponding path number and path location.

[0095] Additional instruction candidate generation: For each position with an overlap coefficient > 0, select the preceding position in its path. Use this preceding position as a candidate point for adding a waiting instruction.

[0096] The machine learning model evaluates these locations and outputs a recommended wait time.

[0097] Update scheduling instructions: Insert a waiting operation at the candidate position (e.g., "wait 4s at (x1,y1)"). Generate a new set of scheduling instructions.

[0098] Resimulate: Resimulate the entire path using the new instructions. Recalculate the overlap coefficients at all locations.

[0099] Termination judgment: If all overlap coefficients are 0 → end, output the final scheduling scheme; otherwise → continue the loop (return to step 1).

[0100] The scheduling process is simulated using added instructions. The simulation of adding instructions and scheduling is repeated cyclically. Upon completion and termination, a "strategy for adding instructions" (referring to all added instructions) is output. Through multiple repetitions, each "strategy for adding instructions" is ensured to be different from the previously generated strategy, resulting in multiple strategies (generally with a preset maximum number of repetitions).

[0101] Optimizing the scheduling instructions includes: setting the minimum sum of waiting times in the scheduling instructions as the objective; using the latest completion time of each robot as a constraint; selecting the optimal addition instruction; and adding the optimal addition instruction to the original scheduling instructions to complete the optimization.

[0102] In multi-robot scheduling systems, conflicts caused by intersecting robot paths are a major cause of task delays, increased collision risks, and decreased resource utilization. To avoid such conflicts, the system can add "waiting instructions" at key points along the path, achieving safe and orderly scheduling through staggered travel times.

[0103] However, while waiting can resolve conflicts, it can also lead to decreased system efficiency, especially in large-scale robot collaborative operations. Inappropriately set waiting times can severely impact overall task completion time and scheduling quality. This paper sets the "sum of waiting times" as the objective function, aiming to minimize the overall system's "waiting cost" while achieving scheduling feasibility, reducing ineffective robot waiting time, and improving overall execution efficiency. By using each robot's "latest completion time" as a constraint, it avoids situations where the optimized solution minimizes conflicts but becomes unexecutable due to timeouts, ensuring that each task is completed within its time window and meeting system stability requirements. Compared to traditional methods of adjusting and manually setting waiting times one by one, this optimization design introduces mathematical modeling and machine learning / heuristic strategies to comprehensively evaluate and select the overall optimal combination from all possible waiting instructions, exhibiting globality, intelligence, and high robustness.

[0104] S5: After optimization and confirmation, the solution is directly issued to the robot, and the robot directly updates the scheduling instructions.

[0105] On the other hand, this embodiment also provides a cloud-based multi-robot scheduling and management system, which includes:

[0106] The data acquisition unit obtains real-time status data and scheduling instructions for each robot within the area and uploads them to the cloud via a data channel. The cloud aggregates the real-time status data of the robots within the area, establishes a virtual twin digital model of the robots, and maps the real-time status of the physical robots onto the map of the area.

[0107] The cloud-based computing power allocation unit distributes computing power based on the pre-simulation assessment. After allocation, machine learning is used to simulate scheduling instructions for the robot.

[0108] The optimization unit optimizes the scheduling instructions based on the simulation results in the cloud.

[0109] Once the control unit has been optimized and confirmed, the solution is directly issued to the robot, and the robot directly updates its scheduling instructions.

[0110] If the above functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0111] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0112] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0113] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0114] Example 2 is an embodiment of the present invention, which provides a multi-robot scheduling and management method based on cloud computing. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiment.

[0115] Experimental preparation and process:

[0116] This embodiment selects a warehouse environment as the multi-robot scheduling task scenario. The experimental area is 200 square meters, and 12 delivery robots are deployed. The system divides the area into 4 computing nodes, each connected to the cloud. The scheduling tasks include a total of 120 material handling tasks, with each robot needing to execute an average of 10 paths.

[0117] Before the experiment, two schemes were constructed: a traditional scheduling system and the system of the present invention, and the same task list, robot start and end positions and scheduling time window were used.

[0118] Traditional systems employ fixed priority path rules and implement a "first-come, first-served" strategy in conflict areas. The system of this invention, however, uses a virtual twin model to map the robot's state in real time, calculates the path conflict intensity based on the overlap coefficient, and simulates different combinations of waiting instructions through a machine learning model. Under the premise of meeting the completion time limit, it selects the combination scheme with the shortest total waiting time to add instructions.

[0119] The experimental results, with average task completion time, number of path conflicts, and total robot waiting time as core evaluation indicators, show that the proposed solution significantly reduces the number of path conflicts by approximately 84%, reflecting that the overlap coefficient assessment and waiting instruction simulation mechanisms can identify potential risks in advance and resolve conflicts caused by multiple robots simultaneously passing through the same area with minimal waiting costs. Regarding average task completion time, the proposed system completes the same amount of tasks with an improvement of approximately 17.5%, demonstrating its global optimization capability for scheduling efficiency. Furthermore, the total waiting time is reduced from 108 seconds in the traditional solution to 39 seconds, showing a significant advantage in the objective function of scheduling optimization.

[0120] Therefore, this method demonstrates superior scheduling intelligence and resource allocation capabilities in complex multi-path, high-density operation scenarios. It not only effectively solves the problems of slow response and numerous conflicts in existing scheduling strategies, but also provides a more intelligent solution for balancing execution efficiency and energy consumption. It has high novelty and industrial application promotion value.

[0121] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-robot scheduling and management method based on cloud computing, characterized in that, include: Acquire real-time status data and scheduling instructions for each robot within the area, and upload them to the cloud via a data channel; The cloud aggregates real-time status data of robots within a region, establishes a virtual twin digital model of the robots, and maps the real-time status of the physical robots onto a map of the region. Based on the pre-simulation assessment, the cloud platform allocates computing power; after allocation, machine learning is used to simulate scheduling commands for the robot. The scheduling instructions are optimized based on the simulation results in the cloud. Once the optimization is confirmed, the solution is directly issued to the robot, and the robot directly updates its scheduling instructions.

2. The cloud computing-based multi-robot scheduling and management method as described in claim 1, characterized in that: The cloud contains multiple cloud nodes, with each region corresponding to each cloud node in a one-to-one manner. The real-time status data includes the robot's own position, speed, battery level, and energy consumption data.

3. The cloud computing-based multi-robot scheduling and management method as described in claim 2, characterized in that: The scheduling instructions include the execution time of the scheduling task, the latest completion time, and the path planning results for the robot.

4. The cloud computing-based multi-robot scheduling and management method as described in claim 3, characterized in that: Based on the virtual twin digital model of the robot, the path planning of each robot is depicted on the map; In the map, each coordinate location is labeled with attributes, including: whether it is a walkable area and a path marker; Let x be the x-coordinate of any position on the map, and y be the y-coordinate of any position on the map; if the path of robot i passes through the coordinate position (x, y), then add a path marker to the coordinate position (x, y); different path markers can be superimposed on the same coordinate position. If there is only one path marker at coordinate position (x, y), the marker content is empty; if there is more than one path marker at coordinate position (x, y), the marker content is: the estimated time interval for robot i to reach coordinate position (x, y); The pre-simulation evaluation includes analyzing the coordinate positions (x, y) where the path markers are not empty: if the expected time intervals of the coordinate positions (x, y) overlap, the overlap coefficient of the coordinate positions (x, y) is assigned a value equal to the cumulative number of overlaps: among all path markers of the coordinate positions (x, y), any two expected time intervals are selected, and if they overlap each time, the cumulative value is increased by 1; if the expected time intervals of the coordinate positions (x, y) do not overlap, the overlap coefficient of the coordinate positions (x, y) is assigned a value of 0.

5. The cloud computing-based multi-robot scheduling and management method as described in claim 4, characterized in that: The allocation of computing power includes associating coordinate positions where the marked content is not empty, packaging the data according to the association results, allocating it to other cloud nodes, and simulating scheduling instructions through other cloud nodes; The association process is as follows: The marked coordinates serve as the link between different robot paths; Paths with direct or indirect connections are extracted to obtain multiple sets of associated paths; For any set of associated paths j, in the virtual twin digital model, other paths outside of the set of associated paths j are eliminated, and the eliminated virtual twin digital model is packaged to obtain packaged data about the set of associated paths j. There is no kinetic link between different sets of associated paths.

6. The cloud computing-based multi-robot scheduling and management method as described in claim 5, characterized in that: The allocation of computing power also includes the following: if the total overlap coefficient in the packaged data of the associated path set j is 0, then no allocation is performed and the original scheduling instruction is maintained. If the total overlap coefficient in the packaged data of the associated path set j is not 0, then the sum of the number of links and the overlap coefficients in the associated path set j will be used as the allocation reference coefficient. Cloud nodes are allocated according to the magnitude of the allocation reference coefficient; During allocation, if the number of available cloud nodes is greater than or equal to the number of associated path sets, the allocation reference coefficients are allocated sequentially from largest to smallest according to the remaining computing power of each cloud node. If the number of available cloud nodes is less than the number of associated path sets, the associated path sets are combined into an optimal hybrid path set. The allocation reference coefficients of the optimal hybrid path set are allocated sequentially from largest to smallest according to the remaining computing power of each cloud node from largest to smallest. In this process, the associated path set is randomly combined to obtain multiple mixed path sets, and the optimal mixed path set is calculated through the objective function and constraints. The objective function is to make the ratio of remaining computing power of each available cloud node as close as possible to the ratio of the allocation reference coefficient of the hybrid path set; the constraint is that the number of hybrid path sets equals the number of available cloud nodes. The allocation reference coefficient of the mixed path set is quantified as the sum of the allocation reference coefficients of the associated path sets in the combination.

7. The cloud computing-based multi-robot scheduling and management method as described in claim 6, characterized in that: The simulation of the scheduling instructions includes using machine learning to add instructions so that each robot can wait at the position before reaching each link; The scheduling process is simulated by adding new instructions, and the process of adding instructions and simulating the scheduling process is repeated. Optimizing the scheduling instructions includes setting the minimum value of the sum of waiting times in the scheduling instructions as the objective. The latest completion time of each robot is used as a constraint; The optimal addition instruction is obtained through selection. The optimal addition instruction is added to the original scheduling instruction to complete the optimization.

8. A cloud computing-based multi-robot scheduling and management system employing the method described in any one of claims 1-7, characterized in that: The data acquisition unit obtains real-time status data and scheduling instructions for each robot within the area and uploads them to the cloud via a data channel. The cloud aggregates the real-time status data of the robots within the area, establishes a virtual twin digital model of the robots, and maps the real-time status of the physical robots onto the map of the area. The cloud-based allocation unit redistributes computing power based on the pre-simulation assessment; after allocation, machine learning is used to simulate scheduling commands for the robot. The optimization unit optimizes the scheduling instructions based on the simulation results in the cloud. Once the control unit has been optimized and confirmed, the solution is directly issued to the robot, and the robot directly updates its scheduling instructions.

9. A computer device, comprising: A memory and a processor; the memory stores a computer program, characterized in that: when the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-7.

Images