Robot scheduling method and system based on collaborative control

Abstract

The present invention relates to the field of industrial control technology, and in particular to a robot scheduling method and system based on collaborative control, comprising: obtaining task proportions and target node information in a task adjustment plan, determining the optimal task distribution path and execution order based on task priority sorting and node adaptation rules, and determining a patrol task distribution blueprint; using a low-latency scheduling framework to distribute tasks in real time based on the patrol task distribution blueprint, obtaining a dynamic distribution map of tasks between nodes and real-time distribution update data based on node response delay and inter-node communication overhead; obtaining the latest load status of each node through the dynamic distribution map and real-time distribution update data, and determining whether there is a potential risk of computational pressure overload based on distribution uniformity indicators and distribution anomaly detection. The invention aims to solve the problems of large performance differences, unbalanced loads, and low task scheduling efficiency in the process of collaborative inspection by multiple robots in a narrow space.

Description

Technical Field

[0001] The present invention relates to the field of industrial control technology, and in particular to a robot scheduling method and system based on collaborative control. Background Art

[0002] In modern industry and services, multi-robot collaborative control technology has attracted significant attention due to its enormous potential for improving efficiency, reducing costs, and handling complex tasks. This technology is considered a core pillar driving the development of intelligent manufacturing and automation, and its importance is particularly evident in scenarios where multiple robots must collaborate to complete tasks. Similar prior art includes Chinese patent application number CN118884905A, which proposes a multi-robot collaborative task allocation system comprising a task generation module, a task allocation module, a robot coordination module, a perception-environment interaction module, a path planning module, and a data stabilization module. This application calculates priorities based on task weights and deadlines, optimizes task allocation using integer programming, and ensures that high-priority tasks are executed first. It also uses Shapley value calculation to measure robot contributions, ensuring fair and reasonable task allocation and improving collaborative efficiency. It uses lidar sensors to construct occupancy grid maps and employs the A* algorithm for path planning, improving both accuracy and efficiency. It monitors task execution in real time, dynamically adjusting task allocation and execution strategies to maintain system stability. It also utilizes blockchain technology to record task allocation and execution processes, ensuring data transparency and immutability. Another similar prior art method is Chinese patent publication number CN119270795A, which proposes a collaborative inspection task scheduling and path planning method for multiple quality inspection robots. This method involves collecting production information from the production line and status information from the inspection robots. Then, based on this information, the inspection tasks are rationally assigned to each inspection robot. This ensures a balanced workload for each inspection robot, thereby improving the inspection efficiency of the entire production system. The inspection robots then proceed to the locations of the components to be inspected according to their assigned inspection tasks. The system also monitors production information from the production line and the status information of each inspection robot in real time. Based on this information, the system evaluates the priority of each inspection task, the availability of each inspection robot, and the production efficiency of the production line. Finally, based on this information, the inspection tasks assigned to each inspection robot are dynamically adjusted. This allows the production system to flexibly respond to various production changes and improves the inspection efficiency of the entire production system. While the technical solutions in these two patents address the issue of collaborative control, they fail to consider the individual performance differences of robots in resource allocation and task scheduling. Consequently, they struggle to achieve real-time response and flexible adjustments when task demands rapidly change. Summary of the Invention

[0003] The present invention provides a robot scheduling method and system based on collaborative control, the method comprising:

[0004] Obtain performance data from multiple inspection robots, determine the initial load value and computing power classification of each robot through a pre-established performance archive, and generate an initial load distribution table;

[0005] By collecting the operating data of each robot in real time, analyzing the load fluctuation, judging the load imbalance, and determining the abnormal node list;

[0006] Dynamically adjust inspection task allocation based on the abnormal node list and generate a task adjustment plan;

[0007] According to the task adjustment plan, determine the task distribution path and execution order to form a task distribution blueprint;

[0008] According to the task distribution blueprint, tasks are distributed in real time through a low-latency scheduling framework to obtain a dynamic task distribution map and real-time updated data;

[0009] The load status of each robot is obtained through the dynamic distribution map and the real-time updated data, and a risk node set is determined. Tasks are secondary distributed for the risk node set to generate an adjusted task distribution matrix.

[0010] As a preferred technical solution of the present invention, the determination of the initial load distribution includes:

[0011] The historical task processing rates and current operating parameters are obtained from a pre-established performance archive. Based on the historical task processing rates and current operating parameters, the processing capability differences of each robot are analyzed. If the processing rate of a robot is lower than the preset rate, it is marked as a low-performance node. For the low-performance node, the load distribution ratio is reduced and an initial load distribution table is generated. Based on the initial load distribution table, the load is dynamically adjusted to obtain an adjusted load distribution.

[0012] As a preferred technical solution of the present invention, obtaining the abnormal node list includes:

[0013] The load status sensing component is used to collect the operating data of each inspection robot in real time to obtain parameters related to the calculation pressure; a time series analysis algorithm is used to calculate the load fluctuation trend in the operating data to determine whether the load fluctuation exceeds the preset fluctuation; if the load fluctuation exceeds the preset fluctuation, the relevant inspection robot is marked as an abnormal node; based on the abnormal nodes and combined with historical operating data, a list of abnormal nodes is generated; for the abnormal node list, the load change trend is continuously monitored to obtain real-time status information.

[0014] As a preferred technical solution of the present invention, the acquisition of the task adjustment plan includes:

[0015] Obtain the load data in the abnormal node list and analyze the load status of each node; for nodes with unbalanced loads, use a task splitting algorithm to decompose the inspection tasks and generate a detailed task unit list; based on the task unit list and combined with the task migration cost data, calculate the resource consumption of migrating the task to the target node and determine the migration feasibility plan; if the migration feasibility plan shows that the migration cost exceeds the preset cost, re-split the task unit to obtain an adjusted task unit combination; based on the adjusted task unit combination, verify the load capacity of the target node and generate the task adjustment plan.

[0016] As a preferred technical solution of the present invention, the determination of the task distribution blueprint includes:

[0017] The task proportions and target node data in the task adjustment plan are obtained, and the task priority sorting rules are used to classify the tasks; based on the classification results, combined with the node adaptation rules, the resource status of the target node is analyzed to determine the corresponding relationship between the tasks and the nodes; based on the corresponding relationship, the path optimization algorithm is used to analyze the feasibility of the task distribution path and generate a preliminary distribution path; if the resource occupancy of the preliminary distribution path exceeds the preset value, the task proportion is dynamically adjusted to generate an optimized distribution path, otherwise no adjustment is made; based on the optimized distribution path or the preliminary distribution path that does not need to be optimized, combined with the execution order rules, the task execution order is arranged to form the task distribution blueprint.

[0018] As a preferred technical solution of the present invention, the acquisition of task dynamic distribution map and real-time updated data includes: for the task distribution blueprint, using the low-latency scheduling framework to preliminarily allocate tasks, and analyzing the current load status of each node; according to the current load status, combined with the node response delay and communication overhead data, calculating the task allocation ratio, and determining the dynamically adjusted distribution state; if the deviation of the distribution state exceeds the preset state deviation, fine-tuning the task allocation ratio to generate the task dynamic distribution map; for the task dynamic distribution map, obtaining real-time updated node interaction data, judging the delay risk, and determining the adjusted distribution strategy; according to the adjusted distribution strategy, continuously monitoring the task distribution state and generating the real-time updated data.

[0019] As a preferred technical solution of the present invention, the determination of the risk node set includes: obtaining the load status information of each inspection robot through the dynamic distribution map in combination with the real-time update data; using the distribution uniformity index to evaluate the load distribution of the load status information, and judging whether the distribution deviation exceeds the preset distribution deviation; if the distribution deviation exceeds the preset distribution deviation, marking the relevant node as an abnormal distribution node; according to the abnormal distribution node, combined with historical load records, analyzing the load pressure accumulation, and determining the risk node set; for the risk node set.

[0020] As a preferred technical solution of the present invention, it also includes:

[0021] According to the task distribution matrix, a real-time monitoring system is used to track the task execution process, obtain execution status feedback data, and determine the progress of inspection task completion; the performance archive is updated through the execution status feedback data to determine the resource allocation benchmark and adjustment trigger conditions for the next cycle.

[0022] As a preferred technical solution of the present invention, it also includes:

[0023] According to the task distribution matrix, a real-time monitoring system is used to track the task execution process, obtain execution status feedback data, and determine the progress of inspection task completion; the performance archive is updated through the execution status feedback data to determine the resource allocation benchmark and adjustment trigger conditions for the next cycle.

[0024] The present invention also provides a robot scheduling system based on collaborative control, for implementing the above method, the system comprising:

[0025] A generation unit is used to obtain performance data of multiple inspection robots, determine the initial load value and computing power classification of each robot through a pre-established performance archive, and generate an initial load distribution table;

[0026] The determination unit is used to collect the operating data of each robot, analyze the load fluctuation, determine the load imbalance, and determine the abnormal node list;

[0027] An allocation unit, configured to dynamically adjust inspection task allocation based on the abnormal node list, generate a task adjustment plan, and determine task distribution paths and execution sequences based on the task adjustment plan to form a task distribution blueprint;

[0028] An acquisition unit, configured to distribute tasks in real time through a low-latency scheduling framework according to the task distribution blueprint, obtain a dynamic distribution map of tasks, and update data in real time;

[0029] The allocation unit is further used to obtain the load status of each robot through the dynamic distribution map and the real-time updated data, determine the risk node set, perform task secondary allocation on the risk node set, and generate an adjusted task distribution matrix.

[0030] The present invention also provides a computer-readable storage medium having instructions stored thereon, and the above-mentioned method is implemented when the instructions are executed by a processor.

[0031] Advantages compared to existing technologies:

[0032] This invention proposes a collaborative control-based robot scheduling method and system, aiming to address the significant performance variations, load imbalance, and low task scheduling efficiency inherent in multi-robot collaborative inspections within confined spaces. This technical solution builds a robot performance archive, collects and analyzes each robot's operating data in real time, and dynamically assesses its load status and computing power, thereby enabling intelligent task allocation and scheduling. The system first classifies robots based on their historical processing rates and current operating parameters (such as CPU utilization and memory usage) and generates an initial load distribution table. Subsequently, a load status sensing component monitors load fluctuations at each node in real time, identifies abnormal nodes, and dynamically adjusts tasks to ensure load balancing. During task distribution, the system employs a low-latency scheduling framework that combines node response latency and communication overhead to generate a dynamic distribution map. The system then continuously optimizes task allocation strategies based on real-time data updates. For potentially risky nodes, the system uses a distribution uniformity indicator and anomaly detection mechanism to proactively identify computing overload risks and trigger a secondary task allocation mechanism, further improving system stability and robustness. The system also features real-time task execution tracking capabilities, dynamically adjusting task plans based on execution status feedback data and feeding historical data back to the performance archive, enabling long-term load optimization and adaptive updates of scheduling strategies. Overall, this technical solution enables efficient multi-robot collaborative operations in complex environments, significantly improving the balance, real-time performance, and stability of task execution while reducing system energy consumption and communication overhead. It holds great engineering application value and promising prospects for widespread adoption. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Flowchart of a robot scheduling method based on collaborative control in an embodiment of the present invention;

[0034] Figure 2 Schematic diagram of a method for obtaining a task adjustment plan in an embodiment of the present invention;

[0035] Figure 3 Schematic diagram of a method for determining a task distribution blueprint in an embodiment of the present invention;

[0036] Figure 4 This is a structural diagram of a robot scheduling system based on collaborative control in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] Embodiments of the present invention provide a robot scheduling method and system based on collaborative control. The terms "first", "second", "third", "fourth", etc. (if any) in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or apparatus. For ease of understanding, the specific process of an embodiment of the present invention is described below. In this embodiment, the robot scheduling method based on collaborative control, such as Figure 1 As shown, including:

[0038] Step S1: Obtain performance data of multiple inspection robots, determine the initial load value and computing power classification of each robot through a pre-established performance archive, and generate an initial load distribution table;

[0039] Specifically, by obtaining the performance data of multiple inspection robots and using a pre-established performance archive to determine the initial load value and computing power classification of each robot, an initial load distribution table is generated. This process first requires a comprehensive analysis of the robot's historical task processing rate and current operating parameters to assess the differences in its processing capabilities. For robots with processing rates lower than the preset rate, they are marked as low-performance nodes and their load distribution ratio is reduced. In this way, the system can reasonably distribute the initial load according to the actual performance of the robot, ensuring the fairness and efficiency of task allocation. The above technical solution provides a scientific and reasonable load distribution basis for subsequent task scheduling, preventing some robots from affecting the overall performance due to excessive load, and providing accurate initial data support for dynamic adjustment.

[0040] Step S2: Collect the operating data of each robot, analyze the load fluctuation, determine the load imbalance, and determine the abnormal node list;

[0041] Specifically, by collecting the operating data of each robot in real time, the load state perception component is used to obtain parameters related to the calculation pressure, and then analyze the load fluctuation situation to determine whether there is a load imbalance phenomenon, so as to determine the list of abnormal nodes. The system uses a time series analysis algorithm to calculate the load fluctuation trend in the operating data and sets a preset fluctuation threshold for comparison. When the load fluctuation exceeds the preset threshold, the relevant robot will be marked as an abnormal node. Combined with historical operating data, the system generates a list of abnormal nodes and continuously monitors the load change trends of these nodes to obtain real-time status information. The above technical solution can promptly detect and mark robots with abnormal loads, provide accurate target node information for subsequent task adjustments and load balancing, and ensure that the system can quickly respond to load changes and maintain overall stable operation.

[0042] Step S3: dynamically adjust inspection task allocation for the abnormal node list and generate a task adjustment plan;

[0043] Specifically, for the abnormal node list, a task adjustment plan is generated by dynamically adjusting the distribution of inspection tasks. The system first obtains the load data in the abnormal node list and deeply analyzes the load status of each node. For nodes with unbalanced loads, a task splitting algorithm is used to decompose the inspection tasks into smaller task units and generate a task unit list. Combined with the task migration cost data, the system calculates the resource consumption required to migrate the task unit to the target node to determine the feasibility of the migration plan. If the migration cost exceeds the preset cost, the system will re-split the task unit to obtain the adjusted task unit combination. Finally, based on the adjusted task unit combination, the load capacity of the target node is verified to generate the final task adjustment plan. The above technical solution effectively solves the load imbalance problem by dynamically adjusting task allocation, optimizes resource utilization, ensures the reasonable distribution of tasks among robots, and improves the overall efficiency and response speed of the system.

[0044] Step S4: According to the task adjustment plan, determine the task distribution path and execution order to form a task distribution blueprint;

[0045] Specifically, based on the task adjustment plan, the task distribution path and execution order are determined to form a task distribution blueprint. The system first obtains the task ratio and target node data in the task adjustment plan and classifies the tasks according to the task priority sorting rules. Combined with the node adaptation rules, the resource status of the target node is analyzed to determine the correspondence between tasks and nodes. Based on this correspondence, a path optimization algorithm is used to analyze the feasibility of the task distribution path and generate a preliminary distribution path. If the resource usage of the preliminary distribution path exceeds the preset value, the system dynamically adjusts the task ratio and generates an optimized distribution path; otherwise, the preliminary distribution path remains unchanged. Finally, based on the optimized distribution path or the preliminary distribution path that does not require optimization, combined with the execution order rules, the execution order of the tasks is arranged to form a complete task distribution blueprint. The above technical solution reduces the time and resource consumption of task execution by optimizing the task distribution path and execution order, improves the efficiency and reliability of task execution, and ensures that the system can efficiently complete inspection tasks.

[0046] Step S5: According to the task distribution blueprint, tasks are distributed in real time through a low-latency scheduling framework to obtain a dynamic task distribution map and real-time update data;

[0047] Specifically, according to the task distribution blueprint, tasks are distributed in real time through a low-latency scheduling framework, and a dynamic distribution map of tasks and real-time updated data are obtained. The system first uses a low-latency scheduling framework to perform a preliminary distribution of tasks and analyze the current load status of each node. Combined with the node response delay and communication overhead data, the distribution ratio of tasks among the nodes is calculated to determine the dynamically adjusted distribution status. If the deviation of the distribution status exceeds the preset state deviation, the system will fine-tune the task distribution ratio and generate a dynamic distribution map of tasks. For the dynamic distribution map of tasks, the system obtains real-time updated node interaction data, determines whether there is a delay risk, and determines the adjusted distribution strategy. Based on the adjusted distribution strategy, the system continuously monitors the distribution status of tasks and generates real-time updated data. The above technical solution ensures the balanced distribution of tasks among nodes by distributing and dynamically adjusting tasks in real time, reduces the delay in task execution, improves the real-time performance and response speed of the system, and effectively reduces the communication overhead between nodes, thereby improving the efficiency of overall collaborative work.

[0048] Step S6: Obtain the load status of each robot through the dynamic distribution map and the real-time updated data, determine the risk node set, perform secondary task allocation on the risk node set, and generate an adjusted task distribution matrix.

[0049] Specifically, through dynamic distribution maps and real-time updated data, the system obtains the load status of each robot, identifies a set of risk nodes, and then performs a secondary redistribution of tasks for these risk nodes to generate an adjusted task distribution matrix. The system first combines the dynamic distribution map and real-time updated data to obtain load status information for each inspection robot. It then uses a distribution uniformity metric to assess the load distribution and determine whether the distribution deviation exceeds a preset distribution deviation. Nodes with distribution deviations exceeding the preset distribution deviation are marked as abnormally distributed nodes by the system. Combined with historical load records, the system analyzes the accumulated pressure at these abnormally distributed nodes and identifies a set of potential risk nodes. For this set of risk nodes, the system performs a secondary redistribution of tasks, adjusts the task distribution, and generates a new task distribution matrix. This technical solution, through the timely identification of risk nodes and the secondary redistribution of tasks, effectively prevents computational overload on some robots due to excessive load, ensuring system stability and reliability while optimizing task distribution and improving overall task execution efficiency and resource utilization.

[0050] Furthermore, the determination of the initial load distribution includes:

[0051] The historical task processing rates and current operating parameters are obtained from a pre-established performance archive. Based on the historical task processing rates and current operating parameters, the processing capability differences of each robot are analyzed. If the processing rate of a robot is lower than the preset rate, it is marked as a low-performance node. For the low-performance node, the load distribution ratio is reduced and an initial load distribution table is generated. Based on the initial load distribution table, the load is dynamically adjusted to obtain an adjusted load distribution.

[0052] Specifically, through the pre-built archive data, the historical task processing rate and current parameters of each micro-inspection robot are obtained. The current operating parameters include CPU occupancy, memory usage and current task processing rate. For the operating state of the preset space, that is, the narrow space, the performance differences of each robot are determined. According to the obtained historical task processing rate and current parameters, the processing capability of each robot in the collaborative inspection scenario is analyzed to judge the specific manifestations of the performance differences. If the processing rate of a robot is lower than the preset rate, it will be marked as a low-performance node, that is, the performance grading result. The performance grading result is used, combined with the collaborative inspection requirements in the preset space, that is, the narrow space, to calculate the initial load value of each robot, reduce the load distribution ratio for low-performance nodes, and determine the initial load distribution plan. The initial load distribution plan analyzes the degree of matching between the robots' computing capabilities. If a node's load exceeds its computing capacity, its load is dynamically adjusted to obtain an adjusted load distribution table. Based on this adjusted load distribution table, the operating status data of each robot in a confined space is obtained. For nodes with large performance differences, load prediction is performed using a support vector machine algorithm. This support vector machine model is trained with historical load data (including at least CPU usage, memory utilization, and task processing rate). The current load data is input into this support vector machine model to predict the load for future time periods and determine a predicted load balancing plan. Based on this predicted load balancing plan, combined with historical processing data and current parameters, the stability of each robot during collaborative inspection is analyzed. If the stability falls below a preset threshold, the load of the relevant node is fine-tuned again to obtain the final load optimization result. The final load optimization result is obtained. Based on the optimized load distribution, the collaborative inspection efficiency of each robot in the confined space is analyzed to determine whether the operating status of each node is balanced, and a complete performance optimization distribution table is obtained.Here is an example: In a scenario where multiple robots work together in a narrow space, first, the historical task processing rate and current operating parameters of each micro inspection robot are obtained through the pre-established robot performance archive; for example, suppose there are three robots, numbered R1, R2, and R3, with historical task processing rates of 5.2 tasks / minute, 4.8 tasks / minute, and 3.5 tasks / minute, respectively. The current operating parameters include CPU occupancy of 60%, 75%, and 50%, and memory usage of 40%, 55%, and 30%, respectively. These data are automatically extracted through the database query interface and stored as a structured table. Then, for performance difference analysis, a weighted scoring algorithm is used to classify the historical task processing rates into corresponding values. The service processing rate has a weight of 0.6, the current CPU usage has a weight of 0.3, and the memory usage has a weight of 0.1. The comprehensive performance score is calculated. For example, the score of R1 is 5.2×0.6+(100-60)×0.3+(100-40)×0.1=3.12+12+6=21.12, the score of R2 is 4.8×0.6+(100-75)×0.3+(100-55)×0.1=2.88+7.5+4.5=14.88, and the score of R3 is 3.5×0.6+(100-50)×0.3+(100-30)×0.1=2.1+15+7=24.1. The analysis shows that R3 has the best performance, followed by R1, and R2 has the lowest performance. Subsequently, the initial load value and computing power classification of each node are determined. Assuming a total of 100 tasks, the load is distributed based on the score ratio: R3 accounts for 24.1 / (21.12+14.88+24.1)=40%, or 40 tasks; R1 accounts for 35%, or 35 tasks; and R2 accounts for 25%, or 25 tasks. Computing power is also classified into high (R3), medium (R1), and low (R2). The classification labels are automatically generated using an algorithm. Finally, an initial load distribution table is generated. The system outputs the load values ​​and classification results in a predefined format, such as {"R1":{"Load":35, "Level":"Medium"}, "R2":{"Load":25, "Level":"Low"}, and "R3":{"Load":40, "Level":"High"}}. This table is then stored in the task scheduling database, providing a basis for subsequent collaborative inspection task allocation. The above process drives load distribution through performance data, forming a complete logical chain from data extraction to result output, ensuring the maximum efficiency of multi-robot collaboration in narrow spaces.

[0053] Furthermore, obtaining the abnormal node list includes:

[0054] The load status sensing component is used to collect the operating data of each inspection robot in real time to obtain parameters related to the calculation pressure; a time series analysis algorithm is used to calculate the load fluctuation trend in the operating data to determine whether the load fluctuation exceeds the preset fluctuation; if the load fluctuation exceeds the preset fluctuation, the relevant inspection robot is marked as an abnormal node; based on the abnormal node and combined with historical operating data, a list of abnormal nodes is generated.

[0055] Specifically, through the load status perception component, the operation data of each inspection robot is collected in real time, the change trend of the calculation pressure and the load fluctuation are analyzed and calculated, and it is determined whether there is an imbalance. A list of abnormal nodes is obtained. According to the list of abnormal nodes, the historical operation data and current status information of the nodes with large load fluctuations are obtained, the specific manifestations of the change trend are analyzed, and the range of nodes that need to be adjusted is determined. The support vector machine algorithm is used to predict the future load pressure distribution for the range of nodes that need to be adjusted, combined with the change trend and load fluctuation data, to obtain the basis for load adjustment after prediction. Based on the load adjustment after prediction, for abnormal nodes, analyze the matching degree between their operating data and the prediction results. If the matching degree is lower than the preset matching degree, make a preliminary distribution adjustment on their load to obtain a temporary load distribution plan. According to the temporary load distribution plan, obtain the real-time operating data of each inspection robot. For the load distribution after preliminary adjustment, analyze whether there is a new imbalance phenomenon and determine the node list for secondary adjustment. Through the node list for secondary adjustment, for new imbalance phenomenon, combine the latest data collected by the state perception component. If the load fluctuation of a node exceeds the preset range, fine-tune it to obtain the final load balancing plan. According to the final load balancing plan, continuously monitor the load distribution and calculation pressure changing trend for the operating data of each inspection robot, and judge The stability of the overall system can be judged and a basis for continuous optimization can be obtained. For example, in the process of realizing load status perception and abnormal node detection, the operating data of each micro inspection robot is first collected in real time through the load status perception component. Assume that there are 5 robots, numbered R1 to R5. The collected data include CPU usage, memory occupancy and task processing delay. For example, the CPU usage of R1 is 75%, the memory usage is 60%, and the delay is 200ms; the CPU usage of R2 is 30%, the memory usage is 25%, and the delay is 100ms; the CPU usage of R3 is 85%, the memory usage is 80%, and the delay is 300ms; the data of R4 and R5 are 40%, 35%, 120ms and 50%, 45%, 150ms respectively. Next, based on the changing trends in node computing pressure, a time series analysis algorithm was used to calculate the average and standard deviation of each robot's CPU usage over the past 10 minutes. For example, R1's average was 70% with a standard deviation of 5%, while R3's average was 82% with a standard deviation of 3%. Using the standard deviation to determine load fluctuations, we found that R3's standard deviation was below the 2% threshold, indicating that its load was consistently high and potentially stressful. Furthermore, to identify load imbalances, we set a load balancing threshold. Assuming a normal range for average CPU usage of 10%, and calculating the average CPU usage for all robots to be 56%, the normal range is 46% to 66%. Analysis revealed that the CPU usage of R1 and R3 exceeded the upper limit, while that of R2 was below the lower limit, indicating a load imbalance.Finally, based on this analysis, we determined a list of nodes with abnormal loads. The algorithm is as follows: if CPU usage exceeds the normal range and the standard deviation is less than the 2% threshold, the node is classified as abnormal. Consequently, R3 is added to the abnormal list. Meanwhile, R1, due to its large standard deviation, is marked as temporarily overloaded and requires further monitoring. This technical solution forms a complete logical chain from data collection to abnormality identification, ensuring that the system automatically identifies load issues and providing a basis for subsequent load balancing adjustments.

[0056] Furthermore, the task adjustment plan is obtained, such as Figure 2 As shown, including:

[0057] Obtain the load data in the abnormal node list and analyze the load status of each node; for nodes with unbalanced loads, use a task splitting algorithm to decompose the inspection tasks and generate a detailed task unit list; based on the task unit list and combined with the task migration cost data, calculate the resource consumption of migrating the task to the target node and determine the migration feasibility plan; if the migration feasibility plan shows that the migration cost exceeds the preset cost, re-split the task unit to obtain an adjusted task unit combination; based on the adjusted task unit combination, verify the load capacity of the target node and generate the task adjustment plan.

[0058] Specifically, the data in the load abnormality node list is obtained. In response to the imbalance phenomenon, the load status and inspection task distribution of each node are analyzed to determine the priority node for task adjustment. Based on the task adjustment priority node, the composition of the current inspection task is decomposed according to the task segmentation granularity to obtain a refined task unit list. Through the refined task unit list, combined with the migration cost data, the resource consumption of each task unit when migrating to the target node is analyzed to determine the feasibility of the migration plan. If the migration feasibility plan shows that the migration cost of a task unit exceeds the preset cost, it is re-segmented to obtain the adjusted task unit combination. Based on the adjusted task unit combination, the feasibility of the allocation ratio is analyzed according to the load capacity of the target node to obtain a preliminary task allocation plan. Through the preliminary task allocation plan, combined with the dynamic adjustment mechanism, the real-time load status of the target node is verified. If new imbalances are found, the allocation ratio is fine-tuned to determine the final task adjustment plan. Based on the final task adjustment plan, the system continuously monitors the status of inspection tasks on target nodes, obtains real-time feedback data for load anomalies, and determines the overall system balance. For example, when dealing with load imbalances, the system dynamically adjusts inspection tasks, developing a task adjustment plan based on the granularity of task segmentation and the cost of task migration. Consider four micro-inspection robots, numbered M1 through M4. M1's task load accounts for 40% of the total task load, M2 for 15%, M3 for 35%, and M4 for 10%. The system detects that M1 and M3 have high loads and needs to reallocate tasks. First, the system uses a task-splitting algorithm to break down tasks into minimal execution units, setting the computational requirements of each unit to a baseline of five load points. For example, if M1 currently has eight load points to migrate, this is calculated as 40% minus the target balancing value of 25% multiplied by the corresponding load points for the total number of tasks. Next, a cost assessment is performed to assess the task migration cost, taking into account network transmission delay and task switching overhead. The cost of migrating one load point to M2 is calculated to be three time units, while migrating one load point to M4 is two time units. These time units can be set based on actual conditions. Analysis shows that migrating to M4 is more optimal. The system then determines a task ratio adjustment plan, migrating six load points from M1 and four load points from M3 to M4. This increases M4's load share to 20%, reduces M1 to 28%, reduces M3 to 27%, and leaves M2 unchanged. This entire process is accomplished through an automated scheduling algorithm. Based on the load point distribution and the migration cost matrix, the system generates a final task adjustment plan to ensure balanced load across nodes while keeping the migration cost within the minimum range of 10 time units, forming a complete logical chain from load detection to task allocation.For further optimization, the system also associates the inspection area coverage data. If the M4 coverage drops below 80% after migration, a secondary fine-tuning will be triggered to migrate some lightweight tasks back to M2 to ensure business continuity. The above technical solution can continuously monitor node load changes, obtain real-time task execution status, and ensure the continuity of dynamic task adjustment.

[0059] Furthermore, the task distribution blueprint is determined as follows: Figure 3 As shown, including:

[0060] The task proportions and target node data in the task adjustment plan are obtained, and the task priority sorting rules are used to classify the tasks; based on the classification results, combined with the node adaptation rules, the resource status of the target node is analyzed to determine the corresponding relationship between the tasks and the nodes; based on the corresponding relationship, the path optimization algorithm is used to analyze the feasibility of the task distribution path and generate a preliminary distribution path; if the resource occupancy of the preliminary distribution path exceeds the preset value, the task proportion is dynamically adjusted to generate an optimized distribution path, otherwise no adjustment is made; based on the optimized distribution path or the preliminary distribution path that does not need to be optimized, combined with the execution order rules, the task execution order is arranged to form the task distribution blueprint.

[0061] Specifically, the task ratio and target node data are obtained from the task adjustment plan. Based on the priority sorting rules, the inspection tasks are preliminarily classified to obtain a task priority list. Based on the task priority list and combined with the node adaptation rules, the resource status and matching degree of the target node are analyzed. A pre-established adaptation model is used to determine the correspondence between each task and the target node, and a preliminary distribution path is obtained. Based on the preliminary distribution path, the feasibility of the path is analyzed based on the resource occupancy during the task distribution process. If the resource occupancy exceeds the preset value, the task ratio is dynamically adjusted to obtain an optimized distribution path. Otherwise, it is not necessary. Based on the optimized distribution path or the preliminary distribution path that does not require optimization, the time dependency and node processing capacity of the inspection tasks are analyzed in combination with the execution order rules. The task execution order is arranged using a scheduling tool to determine the final execution plan. Through the final execution plan, dynamic feedback information is obtained based on the real-time status data of the target node. If the feedback information shows that the node load is uneven, the distribution path is partially corrected to obtain an adjusted execution plan. According to the adjusted execution plan, the running status of the inspection tasks on the target nodes is continuously monitored, and information processing tools are used to extract abnormal data, determine the rationality of the overall system distribution, and obtain real-time monitoring results. Based on the real-time monitoring results, the long-term stability of task distribution is analyzed, the matching degree of path analysis and execution order is analyzed, and the optimization direction of the subsequent task distribution blueprint is determined. For example, the system automatically obtains the task ratio and target node information in the task adjustment plan, and based on the specific requirements of the inspection task, combines task priority sorting and node adaptation rules to generate the optimal task distribution path and execution order, and finally forms the inspection task distribution blueprint. For example, assuming there are 5 inspection devices, numbered P1 to P5, and the current task ratio is P1 accounting for 30%, P2 accounting for 10%, P3 accounting for 25%, P4 accounting for 20%, and P5 accounting for 15%, the system first uses the priority evaluation algorithm to classify tasks into three categories according to their urgency: high, medium, and low. Among them, high-priority tasks account for 40% and require the shortest path distribution. Next, the system analyzes the historical execution efficiency and current power status of each device based on the node adaptation rules, and calculates that P2 and P5 have the highest adaptability, at 85% and 80% respectively. However, P1 has only 30% power left, so its adaptability drops to 50%. Subsequently, the system uses a path optimization algorithm, combined with the geographical distribution of tasks, to calculate the path delay for high-priority tasks assigned to P2 as 5 seconds, and to P5 as 7 seconds. Based on the comprehensive adaptability and delay analysis, P2 is determined to be the primary target node, and 30% of high-priority tasks are assigned to P2, while the remaining 10% are assigned to P5. For medium and low priority tasks, the system sets the task execution time window to the preset time by executing a sequential planning model, so that the execution order of P3 and P4 can be adjusted to improve overall efficiency.The entire process forms a logically closed loop. The system also links to equipment maintenance cycle data. If the power level after the P2 task is executed is expected to be lower than the preset power level, the backup task buffer mechanism is automatically triggered to temporarily store some non-urgent tasks in the cloud scheduling pool. The above technical solution can ensure the stability of the task distribution blueprint and the sustainability of equipment operation, and ultimately generate a complete task allocation and execution plan.

[0062] Furthermore, the acquisition of task dynamic distribution map and real-time updated data includes: for the task distribution blueprint, using the low-latency scheduling framework to preliminarily allocate tasks and analyze the current load status of each node; according to the current load status, combined with the node response delay and communication overhead data, calculating the task allocation ratio and determining the dynamically adjusted distribution status; if the deviation of the distribution status exceeds the preset status deviation, fine-tuning the task allocation ratio and generating the task dynamic distribution map; for the task dynamic distribution map, obtaining real-time updated node interaction data, judging the delay risk, and determining the adjusted distribution strategy; according to the adjusted distribution strategy, continuously monitoring the task distribution status and generating the real-time updated data.

[0063] Specifically, to meet the real-time distribution requirements of inspection tasks, a low-latency scheduling framework is used to perform preliminary task allocation. Combined with the basic data of node responses, the current load status of each node is analyzed to obtain initial distribution status information. Based on this initial distribution status information, a pre-established analysis model is used to calculate the task allocation ratio between nodes based on response delay and communication overhead data, and the dynamically adjusted distribution status is determined. The dynamically adjusted distribution status is combined with real-time updated node data to analyze whether the distribution of tasks between nodes is balanced. If the distribution deviation exceeds the preset distribution deviation, the task allocation ratio is fine-tuned using information processing tools to obtain an optimized distribution map. Based on the optimized distribution map, interaction data between nodes is obtained based on real-time changes in communication overhead, key bottlenecks are extracted, and it is determined whether there is a potential delay risk to obtain a risk assessment result. Based on the risk assessment results, the scheduling framework is used to locally optimize the task distribution path for the bottleneck points in the dynamic map. Combined with the real-time updated data stream, the adjusted distribution strategy is determined. Based on the adjusted distribution strategy, task execution feedback is obtained based on the latest node response status. Information processing tools are used to analyze anomalies in the feedback. If anomalies exceed the preset anomaly threshold, the distribution strategy is revised again to obtain the final real-time distribution plan. This final real-time distribution plan, combined with updated data, continuously monitors the running status of tasks across nodes. Data collection tools are used to record changes in distribution status, determine the overall operational stability of the system, and obtain long-term monitoring data. For example, based on the inspection task distribution blueprint, the system uses a low-latency scheduling framework to distribute tasks in real time, and generates a dynamic distribution map of tasks between nodes and real-time updated data based on the node response delay and inter-node communication overhead. For example, assuming there are 4 inspection nodes, numbered N1 to N4, the system first collects the response delay data of each node in real time and obtains the average delay of N1 to be 3.2 seconds, N2 to be 4.5 seconds, N3 to be 2.8 seconds, and N4 to be 5.1 seconds. At the same time, the communication overhead between nodes is analyzed and the communication bandwidth occupancy rate between N1 and N3 is calculated to be 12%, while the occupancy rate between N2 and N4 is as high as 28%, indicating that the former has higher communication efficiency. Next, the system uses a dynamic load balancing algorithm to weight the total workload based on node latency and communication overhead. The latency weight is set to 0.6, and the communication overhead weight is set to 0.4. The calculated overall score for N3 is 0.82, resulting in a 35% priority allocation to N3. N1 receives 25%, and N2 and N4 receive 20% and 15%, respectively. Subsequently, the system updates node status every 30 seconds based on a real-time data stream update mechanism. If N3's workload is nearing saturation and its latency rises to 3.5 seconds, the task migration model is automatically triggered, shifting 5% of the workload from N3 to N1. The dynamic distribution map is also updated, generating a visual heat map showing the changing trends in the task distribution ratios of each node.Furthermore, to ensure a closed-loop logic, the system also incorporates network coverage data within the inspection area. If the network signal strength of a node, such as N4, falls below 60%, it automatically reduces its assigned task ratio to 10% and redistributes the remaining task load to other nodes, ensuring stable task distribution. Through this process, the system continuously monitors and adjusts, generating real-time updated task distribution data that provides a basis for the dynamic scheduling of subsequent inspection tasks. Based on the inspection task distribution blueprint, a low-latency scheduling framework is used to distribute tasks in real time. A dynamic distribution map of tasks across nodes is generated based on node response latency and communication overhead. Based on this dynamic distribution map, node response latency data is extracted to generate an inter-node latency distribution matrix. This latency distribution matrix is ​​used to determine task load balancing parameters for each node and generate real-time distribution update data. This real-time distribution update data is used to determine whether inter-node communication overhead exceeds a preset threshold. If so, the task distribution strategy is adjusted to generate an optimized task allocation plan. Based on this optimized task allocation plan, communication path data between nodes is obtained to determine the shortest path. Based on the node response latency data, the node load balancing coefficient is calculated and load adjustment instructions are output. The load adjustment instruction is used to update the task distribution path and generate a new dynamic distribution map. The above technical solution extracts communication overhead data through the real-time updated dynamic distribution map to achieve optimized task scheduling.

[0064] Furthermore, the determination of the risk node set includes: obtaining the load status information of each inspection robot through the dynamic distribution map in combination with the real-time updated data; using the distribution uniformity index to evaluate the load distribution of the load status information, and judging whether the distribution deviation exceeds the preset distribution deviation; if the distribution deviation exceeds the preset distribution deviation, marking the relevant node as an abnormal distribution node; based on the abnormal distribution node, combined with historical load records, analyzing the load pressure accumulation, and determining the risk node set; for the risk node set.

[0065] Specifically, through the dynamic distribution map, combined with the real-time updated data stream, the load status information of each node is obtained, and the load status is preliminarily classified using information processing tools to obtain the distribution characteristics of the node load. Based on the distribution characteristics of the node load, the distribution of each node is evaluated using a pre-established analysis model with respect to the distribution uniformity index. If the distribution deviation exceeds the preset distribution deviation, the abnormal distribution node is marked and the abnormal distribution set is determined. For the abnormal distribution set, the historical load records of each node are obtained in combination with the relevant data of the calculated pressure. The data comparison tool is used to analyze whether there is a continuous pressure accumulation and to determine the potential overload risk nodes. Based on the potential overload risk nodes, combined with the real-time updated status data, the interaction frequency between nodes is extracted to determine whether there is a resource competition bottleneck. If the interaction frequency exceeds the preset frequency, the high-risk node set is determined. For the high-risk node set, combined with the distribution uniformity goal, the scheduling tool is used to locally adjust the task allocation path, record the load changes after adjustment, and obtain new distribution status information. Based on the new distribution status information and real-time updated node data, information processing tools are used to analyze whether the adjusted distribution meets the uniformity index. If abnormal distribution still exists, the allocation path is fine-tuned again to determine the optimized distribution plan. Through the optimized distribution plan, combined with long-term data on node load status, data collection tools are used to continuously monitor the operating dynamics of each node, determine the overall distribution stability of the system, and obtain long-term monitoring records. For example, the latest load status of each node is obtained through dynamic distribution maps and real-time distribution update data. Based on the distribution uniformity index and distribution anomaly detection, it is determined whether there is a potential computing load pressure overload risk, and finally the risk node set is determined. For example, suppose we have a distributed system containing multiple computing nodes, that is, multiple robots. First, the dynamic distribution map generation tool is used to collect the CPU usage, memory utilization, and network bandwidth data of each node in real time. Specific values ​​are as follows: Node 1 has a CPU utilization of 85%, a memory utilization of 70%, and a network bandwidth utilization of 60%; Node 2 has a CPU utilization of 40%, a memory utilization of 50%, and a network bandwidth utilization of 30%. The data of other nodes is similar. Next, using a real-time distribution update mechanism, the data is refreshed every 5 minutes to ensure data timeliness. Load distribution analysis is then performed based on this data. For the distribution uniformity metric, a standard deviation algorithm is used to calculate the load distribution of each node. Assuming the standard deviation of CPU utilization is 15.2, exceeding the preset threshold of 10.0, indicating uneven load distribution. Further, using distribution anomaly detection algorithms, such as Z-score-based outlier detection, to calculate the comprehensive load score for each node, the team found that Node A's Z-score was higher than the set threshold, indicating a risk of abnormally high load.Subsequently, combined with historical data analysis, it was confirmed that the CPU usage of node A had been continuously higher than the preset value in the past preset time period, further verifying its risk status. Finally, the uniformity index and anomaly detection results were combined to determine the set of risk nodes, and the resource reallocation mechanism was triggered based on the results. Through the above technical solution, a complete logical chain can be formed from data collection to risk identification, ensuring the accurate positioning of risk nodes, and at the same time associated with the load balancing business to improve the overall efficiency of the system.

[0066] Furthermore, after step S6, the method further includes:

[0067] According to the task distribution matrix, a real-time monitoring system is used to track the task execution process, obtain execution status feedback data, and determine the progress of inspection task completion; the performance archive is updated through the execution status feedback data to determine the resource allocation benchmark and adjustment trigger conditions for the next cycle.

[0068] Specifically, based on the generated task distribution matrix, a real-time monitoring system is used to track the execution of inspection tasks. By collecting real-time execution status feedback data from each inspection robot, such as the number of completed tasks, execution time, and energy consumption, the system can accurately assess task completion progress and the robot's real-time performance. This data is not only used to monitor the execution status of the current task but also to update a pre-established robot performance archive. Based on this execution status feedback data, the system analyzes the performance changes of the robots during actual operation, identifies robots with degraded performance or uneven load, and adjusts the resource allocation benchmark for the next cycle accordingly. Furthermore, the system sets adjustment trigger conditions. For example, if a robot's energy consumption continuously exceeds a preset threshold or its task completion rate falls below expectations, task reallocation or load adjustment mechanisms are automatically triggered to ensure the system maintains efficient and stable operation. This technical solution, through the real-time monitoring system, tracks the execution of tasks, enabling the system to obtain detailed task execution status in a timely manner, ensuring that inspection tasks proceed smoothly as planned. Furthermore, using execution status feedback data to update the performance archive makes the system's assessment of robot performance more accurate and timely, providing a reliable basis for subsequent resource allocation. Furthermore, by setting adjustment trigger conditions, the system can automatically respond to performance changes that occur during robot execution, proactively preventing potential overload risks and preventing the overall efficiency of a task from being impacted by performance degradation of individual robots. This dynamic adjustment mechanism not only improves system stability and robustness, but also optimizes resource utilization, extends the lifespan of robots, and reduces maintenance costs, significantly enhancing the overall performance and engineering application value of the multi-robot collaborative inspection system.

[0069] The present invention also provides a robot scheduling system based on collaborative control, for implementing the above method, the system comprising:

[0070] A generation unit is used to obtain performance data of multiple inspection robots, determine the initial load value and computing power classification of each robot through a pre-established performance archive, and generate an initial load distribution table;

[0071] The determination unit is used to collect the operating data of each robot, analyze the load fluctuation, determine the load imbalance, and determine the abnormal node list;

[0072] An allocation unit, configured to dynamically adjust inspection task allocation based on the abnormal node list, generate a task adjustment plan, and determine task distribution paths and execution sequences based on the task adjustment plan to form a task distribution blueprint;

[0073] An acquisition unit, configured to distribute tasks in real time through a low-latency scheduling framework according to the task distribution blueprint, obtain a dynamic distribution map of tasks, and update data in real time;

[0074] The allocation unit is further used to obtain the load status of each robot through the dynamic distribution map and the real-time updated data, determine the risk node set, perform task secondary allocation on the risk node set, and generate an adjusted task distribution matrix.

[0075] The present invention also provides a computer-readable storage medium having instructions stored thereon, and the above-mentioned method is implemented when the instructions are executed by a processor.

[0076] In summary, the present invention obtains robot performance data through a pre-established performance archive, generates an initial load distribution table, provides basic data for subsequent task allocation, ensures that tasks are reasonably allocated according to the processing capacity of the robot, avoids excessive load concentration, collects operating data in real time, analyzes load fluctuations, determines whether there is a load imbalance, promptly discovers load imbalance, ensures that the system can actively adjust task allocation, and avoids some robots from being overloaded or idle; dynamically adjusts task allocation for abnormal nodes, and generates a task adjustment plan. Adjust task allocation in real time to ensure a more balanced distribution of tasks, reduce task execution delays, and improve efficiency. According to the task adjustment plan, determine the task distribution path and execution order, optimize the task distribution path, reduce execution time and communication overhead, and make task execution more efficient. According to the task distribution blueprint, distribute tasks in real time through a low-latency scheduling framework, update the task dynamic distribution map, and update data in real time. Update task distribution in real time, adjust the order of task execution, ensure the adaptation of task loads at each node, avoid delays and load imbalances, detect risk nodes and perform secondary task allocation through dynamic distribution maps and real-time data updates, avoid computational overload in the system through secondary allocation of risk nodes, and improve the stability and reliability of overall scheduling. The collaborative work and dynamic adjustment mechanism of these steps improves the scheduling efficiency and execution speed of multi-robot collaborative tasks, reduces communication overhead and delays, improves real-time performance, and ensures the stable operation of the system under load fluctuations and abnormal conditions, reducing the risk of overload.

[0077] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A robot scheduling method based on collaborative control, characterized in that: include: Acquiring performance data of multiple inspection robots, determining the initial load value and computing power classification of each robot through a pre-established performance archive, and generating an initial load distribution table, including: obtaining historical task processing rates and current operating parameters from the pre-established performance archive, analyzing the processing power differences of each robot based on the historical task processing rates and current operating parameters, marking a robot as a low-performance node if its processing rate is lower than a preset rate, reducing the load distribution ratio for the low-performance node, generating an initial load distribution table, and dynamically adjusting the load based on the initial load distribution table to obtain an adjusted load distribution; By collecting the operating data of each robot in real time, analyzing the load fluctuation, judging the load imbalance, and determining the abnormal node list; Dynamically adjust inspection task allocation based on the abnormal node list and generate a task adjustment plan; According to the task adjustment plan, the task distribution path and execution order are determined to form a task distribution blueprint, including: obtaining the task ratio and target node data in the task adjustment plan, and classifying them according to the task priority sorting rule; based on the classification result, combined with the node adaptation rule, analyzing the resource status of the target node, and determining the corresponding relationship between the task and the node; based on the corresponding relationship, using a path optimization algorithm to analyze the feasibility of the task distribution path and generate a preliminary distribution path; if the resource occupancy of the preliminary distribution path exceeds a preset value, dynamically adjusting the task ratio to generate an optimized distribution path, otherwise no adjustment is made; based on the optimized distribution path or the preliminary distribution path that does not need to be optimized, combined with the execution order rule, arranging the task execution order to form the task distribution blueprint; According to the task distribution blueprint, tasks are distributed in real time through a low-latency scheduling framework, and a task dynamic distribution map and real-time updated data are obtained, including: with respect to the task distribution blueprint, the low-latency scheduling framework is used to preliminarily distribute tasks, and the current load status of each node is analyzed; according to the current load status, the task distribution ratio is calculated in combination with the node response delay and communication overhead data, and the distribution status after dynamic adjustment is determined; if the deviation of the distribution status exceeds the preset status deviation, the task distribution ratio is fine-tuned to generate the task dynamic distribution map; with respect to the task dynamic distribution map, real-time updated node interaction data is obtained, the delay risk is judged, and the adjusted distribution strategy is determined; according to the adjusted distribution strategy, the task distribution status is continuously monitored to generate real-time updated data; Through dynamic distribution maps and real-time updated data, the load status of each robot is obtained, the risk node set is determined, and tasks are secondary allocated for the risk node set to generate an adjusted task distribution matrix.

2. The robot scheduling method based on collaborative control according to claim 1, characterized in that: Acquiring the abnormal node list includes: The load status sensing component is used to collect the operating data of each inspection robot in real time to obtain parameters related to the calculation pressure; a time series analysis algorithm is used to calculate the load fluctuation trend in the operating data to determine whether the load fluctuation exceeds the preset fluctuation; if the load fluctuation exceeds the preset fluctuation, the relevant inspection robot is marked as an abnormal node; based on the abnormal nodes and combined with historical operating data, a list of abnormal nodes is generated; for the abnormal node list, the load change trend is continuously monitored to obtain real-time status information.

3. The robot scheduling method based on collaborative control according to claim 1, characterized in that: The acquisition of the task adjustment plan includes: Obtain the load data in the abnormal node list and analyze the load status of each node; for nodes with unbalanced loads, use a task splitting algorithm to decompose the inspection tasks and generate a detailed task unit list; based on the task unit list and combined with the task migration cost data, calculate the resource consumption of migrating the task to the target node and determine the migration feasibility plan; if the migration feasibility plan shows that the migration cost exceeds the preset cost, re-split the task unit to obtain an adjusted task unit combination; based on the adjusted task unit combination, verify the load capacity of the target node and generate the task adjustment plan.

4. The robot scheduling method based on cooperative control according to claim 1, characterized in that: Determining the risk node set includes: Through the dynamic distribution map, combined with the real-time updated data, the load status information of each inspection robot is obtained; based on the load status information, the distribution uniformity index is used to evaluate the load distribution, and it is determined whether the distribution deviation exceeds the preset distribution deviation; if the distribution deviation exceeds the preset distribution deviation, the relevant node is marked as an abnormal distribution node; based on the abnormal distribution node, combined with the historical load records, the load pressure accumulation is analyzed to determine the risk node set; for the risk node set.

5. The robot scheduling method based on cooperative control according to claim 1, characterized in that: Also includes: According to the task distribution matrix, a real-time monitoring system is used to track the task execution process, obtain execution status feedback data, and determine the progress of inspection task completion; The performance archive is updated through the execution status feedback data to determine the resource allocation benchmark and adjustment triggering conditions for the next cycle.

6. A robot scheduling system based on collaborative control, used to implement the method according to any one of claims 1 to 5, characterized in that: The system comprises: A generation unit is used to obtain performance data of multiple inspection robots, determine the initial load value and computing power classification of each robot through a pre-established performance archive, and generate an initial load distribution table; The determination unit is used to collect the operating data of each robot, analyze the load fluctuation, determine the load imbalance, and determine the abnormal node list; An allocation unit, configured to dynamically adjust inspection task allocation based on the abnormal node list, generate a task adjustment plan, and determine task distribution paths and execution sequences based on the task adjustment plan to form a task distribution blueprint; An acquisition unit, configured to distribute tasks in real time through a low-latency scheduling framework according to the task distribution blueprint, obtain a dynamic distribution map of tasks, and update data in real time; The allocation unit is further used to obtain the load status of each robot through the dynamic distribution map and the real-time updated data, determine the risk node set, perform task secondary allocation on the risk node set, and generate an adjusted task distribution matrix.

7. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images