Scheduling method for abnormal position complementing of multiple robots in industrial scene

Abstract

The invention discloses a multi-robot abnormal position complementing scheduling method in an industrial scene, and relates to the technical field of industrial robot scheduling, and the scheduling method comprises the following steps: S1, configuring a multi-source sensing unit for each industrial robot, and constructing a multi-dimensional abnormal sensing model to achieve the real-time monitoring and precise recognition of the abnormal state of the robot; s2, designing a layered position covering scheduling framework to generate an initial position covering scheme; s3, an improved self-adaptive scheduling algorithm is adopted to optimize a position complementing scheme; s4, establishing a dynamic adjustment mechanism to realize real-time optimization of the position covering process; s5, state feedback and knowledge base updating are carried out after position complementing is completed; multi-dimensional data such as the running state, the environment and the task progress of the robot are collected through the multi-source sensing unit, feature fusion is achieved through a weighted fusion algorithm, accurate judgment is conducted in combination with an anomaly recognition knowledge base, the anomaly misjudgment rate and the anomaly missed judgment rate are effectively reduced, the anomaly recognition accuracy is effectively improved, and a reliable basis is provided for follow-up position covering scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial robot scheduling, in particular to a scheduling method for abnormal position compensation of multiple robots in an industrial scene. BACKGROUND

[0002] In modern industrial production, multi-robot collaborative operation has become a core mode to improve production efficiency and ensure production continuity, and is widely used in key fields such as automobile manufacturing, electronic assembly, and logistics and warehousing. A multi-robot system completes complex production tasks through division of labor and cooperation, and its running stability directly determines the production capacity and product quality of the entire production line. However, the industrial scene environment is complex and changeable, and robots are prone to running abnormally during operation due to mechanical failure, sensor abnormality, task load mutation, energy depletion, or external interference, etc., which causes the robot to be unable to complete the task as planned, and if effective position compensation scheduling cannot be performed in time, it will cause task accumulation and production interruption, causing serious economic losses.

[0003] In the existing multi-robot abnormal position compensation scheduling method, most of them rely on preset rules or single optimization algorithm to realize position compensation decision. For example, some methods allocate position compensation robots based on fixed priority, and when a high-priority robot is in a full-load state, it is still forced to undertake a position compensation task, resulting in low position compensation efficiency and easy to cause new task delays; some methods use single intelligent algorithms such as genetic algorithm and particle swarm algorithm for scheduling optimization, but such algorithms have the problems of slow convergence speed and easy to fall into local optimal solution when facing dynamic changes in task demand, robot state and environment information in the industrial scene, and cannot quickly generate a position compensation scheme with strong adaptability; at the same time, the existing methods rely on single sensor data for robot abnormal state perception, and the perception accuracy is low, which is easy to cause misjudgment or omission, further affecting the timeliness and accuracy of position compensation scheduling.

[0004] In view of the related technologies in the above, a solution is proposed. SUMMARY

[0005] The purpose of the present application is to provide a scheduling method for abnormal position compensation of multiple robots in an industrial scene to solve the technical problems that the existing monitoring system cannot adapt to low-frequency working conditions, the fault recognition accuracy is low, and there is no targeted early warning mechanism.

[0006] The scheduling method for abnormal position compensation of multiple robots in an industrial scene provided by the present application adopts the following technical scheme: the scheduling method for abnormal position compensation of multiple robots in an industrial scene, the scheduling method comprises the following steps: S1: Configure a multi-source perception unit for each industrial robot, and build a multi-dimensional abnormal perception model to realize real-time monitoring and accurate identification of robot abnormal states, wherein the multi-dimensional abnormal perception model comprises a multi-source perception unit, an abnormal feature fusion rule, and an abnormal identification knowledge base. The robot running state, working environment, and task progress data are collected by the multi-source perception unit, and after feature fusion, the abnormal identification knowledge base is compared to complete abnormal identification; S2: Design a hierarchical replacement scheduling framework to generate an initial replacement scheme. First, quantitatively evaluate and sort normal robots through a robot capability evaluation model, and then filter candidate replacement robots in combination with replacement task parameters to form an initial scheme according to the load margin. S3: Optimize the replacement scheme using an improved adaptive scheduling algorithm. The algorithm is a particle swarm-genetic fusion algorithm, and the initial replacement scheme is used as the initial population. The global optimal scheme is searched by combining a multi-objective optimization function and constraint conditions. S4: Establish a dynamic adjustment mechanism to realize real-time optimization of the replacement process. Real-time replacement execution data is collected, and when the preset trigger condition is met, the evaluation model and task parameters are updated and the scheme is re-optimized. S5: After the replacement is completed, state feedback and knowledge base updating are performed. The task completion is analyzed by the scheduling effect evaluation index, and the abnormal identification knowledge base content is supplemented and the robot capability evaluation model parameters are adjusted accordingly.

[0007] Further, the multi-source perception unit in S1 comprises a running state sensor, an environment perception sensor, and a task progress acquisition module. The parameters collected by the running state sensor include joint temperature, motor speed, power supply voltage, load torque, and motion error. The parameters collected by the environment perception sensor include obstacle distance, environment temperature, and light intensity. The parameters obtained by the task progress acquisition module include task completion progress, remaining task amount, and task deadline.

[0008] Further, the abnormal feature fusion rule in S1 uses a weighted fusion algorithm, and in the weighted fusion algorithm, the running state parameter weight accounts for 60-70%, the task progress data weight accounts for 20-30%, and the environment data weight accounts for 10-20%. The abnormal identification knowledge base stores the feature thresholds and determination rules of common abnormal types such as mechanical failure, energy shortage, sensor failure, and task overload.

[0009] Further, the implementation process of the S1 abnormality identification is: the data collected by the multi-source perception unit is preprocessed by cleaning, denoising and standardization, a unified scale feature value is generated by a weighted fusion algorithm, the feature value is compared with the feature threshold in the abnormality identification knowledge base, and the abnormal robot number, abnormal type, occurrence time and influence range are output combined with the determination rule.

[0010] Further, the evaluation indexes of the robot capability evaluation model in S2 include current load rate, job efficiency, motion accuracy, energy remaining amount and distance from the abnormal robot working area; the weights of each index are determined by the analytic hierarchy process, wherein the weight proportion of the current load rate is 30-40%, the distance weight proportion is 20-30%, the job efficiency weight proportion is 15-20%, the energy remaining amount and motion accuracy weights each account for 5-10%, and the weight calculation result is subjected to consistency test.

[0011] Further, the screening conditions of the candidate replacement robot in S2 are: the comprehensive capability score is not less than the preset threshold, the current load rate is not more than 50%, and the distance from the abnormal working area is not greater than the preset distance. The initial replacement scheme generation rule is: when the candidate set contains only one robot, all tasks are allocated, and when it contains multiple robots, tasks are split according to the load margin to ensure that the load rate of each robot after replacement is not more than 70%.

[0012] Further, the optimization objectives of the multi-objective optimization function in S3 include the shortest replacement task completion time, the lowest total robot energy consumption and the highest task completion accuracy. The constraint conditions include that the robot motion speed does not exceed the maximum limit, the load rate does not exceed the upper limit, the energy consumption does not exceed the remaining energy capacity, and the working position is within the preset regional boundary.

[0013] Further, the optimization process of the improved adaptive scheduling algorithm in S3 is: taking the initial replacement scheme as the initial population, first updating the particle position and velocity by the particle swarm algorithm to realize global search and select the global optimal candidate scheme; then performing genetic operation on the candidate scheme for local optimization, and repeating the above process until the optimal fitness variation of the population in the last 5 generations is less than , and output the optimal scheme.

[0014] Further, the dynamic adjustment trigger conditions in S4 include: the replacement robot appears a new abnormality, the task completion progress and the plan deviation is not less than 10%, the energy consumption speed exceeds the preset threshold, and a new obstacle appears in the working area. The adjustment process is: updating the robot capability evaluation model and the replacement task parameters, calling the improved adaptive scheduling algorithm in step S3 to regenerate the replacement scheme and issuing it.

[0015] Further, the scheduling effect evaluation index in S5 includes task delay rate, precision compliance rate and energy consumption optimization rate. The abnormality recognition knowledge base update content includes newly emerging abnormality types and corresponding feature parameters, and the robot capability evaluation model update content includes weights and scoring standards of each evaluation index.

[0016] To sum up, the present application has at least one of the following beneficial technical effects: High accuracy of abnormality perception: the present application collects multi-dimensional data such as robot running state, environment and task progress through the multi-source perception unit, realizes feature fusion through the weighted fusion algorithm, accurately judges in combination with the abnormality recognition knowledge base, effectively reduces the abnormality misjudgment rate and the missed judgment rate, and effectively improves the accuracy of abnormality recognition, thereby providing a reliable basis for subsequent position filling scheduling.

[0017] Fast position filling decision speed: the hierarchical position filling scheduling framework quickly generates an initial position filling scheme through robot capability evaluation and candidate screening, and the improved particle swarm-genetic fusion algorithm combines the advantages of the two algorithms, thereby greatly improving the convergence speed of scheme optimization, significantly shortening the position filling scheme generation time compared with the traditional single algorithm, and ensuring the rapid start of the position filling task.

[0018] Good optimization of scheduling scheme: the multi-objective optimization function simultaneously considers position filling efficiency, energy consumption and precision, and the improved adaptive scheduling algorithm can search for a global optimal solution, thereby shortening the position filling task completion time, reducing the total energy consumption of the robot, improving the task completion precision compliance rate, and effectively improving the work efficiency and resource utilization rate of the multi-robot system.

[0019] Strong dynamic adaptability: the dynamic adjustment mechanism can respond to various changes in the position filling process in real time, avoid new task delays or abnormalities through real-time optimization of the scheme, and make the multi-robot system adapt to the dynamic changes of the industrial scene, thereby improving the anti-interference ability and production continuity of the system. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flowchart of the scheduling method of the present application. DETAILED DESCRIPTION

[0021] The following will be described in detail in combination with the accompanying drawings. Figure 1 The present application will be further described in detail.

[0022] The embodiments of the present application disclose a scheduling method for multi-robot abnormal position filling in an industrial scene.

[0023] The scheduling method for multi-robot abnormal position filling in an industrial scene includes the following steps: S1: configure a multi-source perception unit for each industrial robot, and construct a multi-dimensional abnormal perception model to realize real-time monitoring and accurate identification of the abnormal state of the robot, wherein the multi-dimensional abnormal perception model comprises a multi-source perception unit, an abnormal feature fusion rule and an abnormal identification knowledge base, the robot running state, the working environment and the task progress data are collected through the multi-source perception unit, and the abnormal identification is completed by comparing the feature fusion with the abnormal identification knowledge base; The multi-source perception unit in S1 comprises a running state sensor, an environment perception sensor and a task progress acquisition module. The parameters collected by the running state sensor include joint temperature, motor speed, power supply voltage, load torque and motion error. The parameters collected by the environment perception sensor include obstacle distance, environment temperature and illumination intensity. The parameters obtained by the task progress acquisition module include task completion progress, remaining task quantity and task deadline.

[0024] The abnormal feature fusion rule in S1 adopts a weighted fusion algorithm, and in the weighted fusion algorithm, the weight proportion of the running state parameter is 60%-70%, the weight proportion of the task progress data is 20%-30%, and the weight proportion of the environmental data is 10%-20%. The implementation process of the abnormal identification in S1 is as follows: the data collected by the multi-source perception unit is preprocessed by cleaning, denoising and standardization, a unified scale feature value is generated through the weighted fusion algorithm, the feature value is compared with the feature threshold in the abnormal identification knowledge base, and the abnormal robot number, the abnormal type, the occurrence time and the influence range are output combined with the determination rule.

[0025] The fusion formula in the abnormal feature fusion rule is: wherein, is the feature value after multi-dimensional data fusion; is the weight of the th data, and satisfies ; is the value of the th data after standardization processing; is the total number of data types, and in this method, , respectively corresponding to the running state, the task progress and the environmental data; The min-max standardization method is adopted for data standardization to convert multi-source data of different dimensions into feature values of a unified scale, thereby providing standardized input for subsequent abnormal identification.

[0026] The abnormal identification knowledge base stores the feature threshold and determination rule of common abnormal types such as mechanical failure, energy shortage, sensor failure and task overload. The anomaly recognition knowledge base contains feature thresholds and judgment rules for common anomaly types of industrial robots. Common anomaly types include mechanical failure, insufficient energy, sensor failure, and task overload. The fused feature data is compared with the feature thresholds in the anomaly recognition knowledge base, and combined with the judgment rules, the abnormal state of the robot is identified in real time, and the abnormal robot number, anomaly type, anomaly occurrence time, and impact range are output. The rule for determining the anomaly type uses a threshold comparison method, and the specific formula is as follows: ,in, The first Lower and upper threshold values ​​for the characteristic values ​​of anomalies (provided by the anomaly identification knowledge base, such as the threshold range for mechanical faults determined by statistical analysis of historical fault data); This represents the feature value range for normal robot operation. When fused feature values... When the robot falls within a certain threshold range for anomalies, it is determined that the robot has encountered the corresponding type of anomaly; if it falls within the normal range, it is determined that the robot is operating normally.

[0027] S2: Design a hierarchical replacement scheduling framework to generate an initial replacement scheme. First, use the robot capability assessment model to quantitatively evaluate and rank normal robots. Then, combine the replacement task parameters to screen candidate replacement robots and allocate tasks according to the load margin to form an initial scheme. The evaluation indicators of the robot capability assessment model described in S2 include current load rate, operation efficiency, motion accuracy, remaining energy, and distance to the abnormal robot operation area. The weights of each indicator are determined by the analytic hierarchy process, with the current load rate accounting for 30%-40%, the distance accounting for 20%-30%, the operation efficiency accounting for 15%-20%, and the remaining energy and motion accuracy each accounting for 5%-10%. The weight calculation results pass the consistency test.

[0028] The robot capability assessment model quantifies the capabilities of all normally operating robots in an industrial setting; in the analytic hierarchy process (AHP), a judgment matrix is ​​constructed. Determine the weights. To assess the number of indicators, Indicates the first The first indicator is relative to the first The importance of each indicator is assigned using a 1-9 scale; weight calculations must pass a consistency check, and consistency indicators... ,in, To determine the largest eigenvalue of a matrix; Random Consistency Ratio ,in As the average random consistency index, according to The value is determined. hour ,when The matrix is ​​judged to meet the consistency requirement.

[0029] The selection criteria for candidate replacement robots in S2 are: comprehensive capability score not lower than the preset threshold, current load rate not exceeding 50%, and distance from abnormal operation area not greater than the preset distance. Based on the evaluation indicators and weights, a comprehensive capability score is calculated for each normal robot, forming a robot capability ranking table. The comprehensive capability score formula is: ,in, To score the robot's overall capabilities, For the first The weights of each evaluation indicator (determined by the analytic hierarchy process and satisfying the following conditions) ); For the first The standardized values ​​of each evaluation indicator; for negative indicators such as the current load rate, the standardization process uses... Perform a reverse conversion to ensure consistency in the direction of the indicators.

[0030] Based on the scope of the anomaly's impact and the task requirements, the parameters for the replacement task are determined. These parameters include the type of replacement task, the workload, the accuracy requirements, and the completion time limit. A set of candidate replacement robots that meet the requirements is selected from the robot capability ranking table using the following screening criteria: comprehensive capability score ≥ preset threshold, current load rate ≤ 50%, and distance from the abnormal operation area ≤ preset distance. The initial replacement scheme generation rule is as follows: when the candidate set contains only 1 robot, all tasks are assigned; when it contains multiple robots, tasks are split according to the load margin to ensure that the load rate of each robot does not exceed 70% after replacement. Specifically: when the candidate replacement robot set contains only one robot, that robot is directly identified as the replacement execution entity and all replacement tasks are assigned; when the candidate replacement robot set contains multiple robots, based on the workload of the replacement tasks and the load margin of each candidate robot, a task splitting algorithm is used to reasonably allocate the replacement tasks to multiple candidate robots, ensuring that the load rate of each robot after replacement is ≤70%, and an initial replacement scheme is generated. The initial replacement scheme includes the replacement robot number, task allocation amount, operation sequence and start time.

[0031] S3: An improved adaptive scheduling algorithm is used to optimize the replacement scheme. The algorithm is a particle swarm-genetic fusion algorithm. The initial replacement scheme is used as the initial population. The global optimal scheme search is completed by combining the multi-objective optimization function and constraints. The optimization objectives of the multi-objective optimization function described in S3 include the shortest completion time for the replacement task, the lowest total energy consumption of the robot, and the highest task completion accuracy. The multi-objective optimization function is transformed into a single-objective function using a weighted summation method, as shown in the following formula: ,in, The variables for the replacement scheme include the replacement robot number and the amount of task allocation. To fill in the task completion time; This represents the upper limit of the task completion time. Total energy consumption of the robot; This is the sum of the rated energy consumption of all replacement robots; For the accuracy of task completion; This is the lower limit of the required accuracy for the task. , , To optimize the target weights (satisfying) Adjustments can be made based on production needs, such as when efficiency is prioritized.

[0032] , , ); The constraints include that the robot's movement speed does not exceed the maximum limit, the load rate does not exceed the upper limit, the energy consumption does not exceed the remaining energy capacity, and the working position is within the boundary of the preset area; The expression is: ,in, , The first The actual movement speed and maximum speed limit of the robot in each part; , The first The load rate and load limit of each part of the robot; , The first Energy consumption and remaining energy of each part of the robot; For the first The coordinates of the robot's working position in each part; This refers to the boundary range of the work area; The optimization process of the improved adaptive scheduling algorithm described in S3 is as follows: using the initial replacement scheme as the initial population, firstly, the particle position and velocity are updated through the particle swarm optimization algorithm to achieve a global search and screen for globally better candidate schemes; then, genetic operations are performed on the candidate schemes to perform local optimization, and the above process is repeated until the change in the optimal fitness of the population is less than 5 times for 5 consecutive generations. Output the optimal solution; The specific strategy of the improved adaptive scheduling algorithm is to combine the global search capability of the particle swarm optimization algorithm with the local optimization capability of the genetic algorithm. In the particle swarm optimization part, the particle velocity and position update formulas are: ,in, This represents the number of iterations. , The first The velocity and position of the substitute particle (corresponding to the replacement scheme variables) ); As inertia weights, a linear decreasing strategy is adopted. ( 4 represents the initial and final inertia weights. (Maximum number of iterations); For learning factors; A random number within the interval [0,1]. The optimal position for each individual particle; The optimal position for the entire population; First, a global search for the task allocation scheme is performed using the particle swarm optimization algorithm to update the particle positions and velocities and select a set of globally superior candidate schemes. Then, genetic operations are performed on the globally optimal candidate solutions, including selection, crossover, and mutation; the selection operation uses the roulette wheel selection method, where the individual selection probability is... ,in, The fitness function (derived from the optimization objective function); The number of candidate solutions; Crossover operations use single-point crossover, with crossover probability... for: ,in, The maximum and minimum crossover probabilities; The maximum and minimum fitness of the current population; The mutation operation uses basic bit mutation, with a mutation probability of... ,in, To maximize and minimize the mutation probability, an adaptive crossover and mutation probability mechanism is used to avoid the algorithm getting trapped in local optima. This process is repeated until the algorithm converges, i.e., the change in the optimal fitness of the population is less than [value missing] for five consecutive generations. Output the optimal solution.

[0033] S4: Establish a dynamic adjustment mechanism to achieve real-time optimization of the fill-in process, collect fill-in execution data in real time, and update the evaluation model and task parameters and re-optimize the scheme when the preset trigger conditions are met. The specific strategy is to collect dynamic data in real time during the replacement process, including the real-time operating status of the replacement robot, task completion progress, energy consumption rate, and environmental change data of the work area. The dynamic adjustment triggering conditions described in S4 include: a new anomaly occurs in the replacement robot, the deviation between the task completion progress and the plan is not less than 10%, the energy consumption rate exceeds the preset threshold, and a new obstacle appears in the work area; The formulas for calculating the task completion schedule deviation and the energy consumption rate deviation are as follows: ,in, Deviation in task completion progress; This represents the actual progress completed. To the planned completion schedule; This is due to deviations in the rate of energy consumption; This represents the actual rate of energy consumption. To plan the rate of energy consumption; when or For example, when it reaches 20%, dynamic adjustment is triggered; The dynamic adjustment process is as follows: update the robot capability assessment model and the replacement task parameters, call the improved adaptive scheduling algorithm in step S3 to regenerate the replacement scheme and issue it.

[0034] S5: After the replacement is completed, status feedback and knowledge base update are performed. The task completion status is analyzed through scheduling effect evaluation indicators. Based on this, the anomaly identification knowledge base content is supplemented and the robot capability evaluation model parameters are adjusted.

[0035] The specific strategy is as follows: After the completion of the replacement task, collect data on the task completion status of the replacement robot, including actual completion time, task completion accuracy, total energy consumption, and operating status; compare and analyze the actual completion status with the optimization target, and calculate the replacement scheduling effect evaluation index. The replacement scheduling effect evaluation index mentioned in S5 includes task delay rate, accuracy achievement rate, and energy consumption optimization rate. The specific formula is as follows: ,in, For task delay rate, if ; This refers to the actual completion time. The deadline for the task; To achieve the accuracy target, the accuracy of a single task is... ( When the accuracy requirement is met, it is judged as meeting the standard. For energy efficiency optimization; The estimated energy consumption for the initial replacement scheme; For actual energy consumption, when This indicates reduced energy consumption and effective optimization. The anomaly recognition knowledge base has been updated with new anomaly types and their corresponding feature parameters, while the robot capability assessment model has been updated with the weights and scoring criteria for each assessment indicator. The specific strategy is as follows: based on the evaluation results of the supplementary scheduling effect and the data in the anomaly handling process, update the anomaly identification knowledge base and the robot capability assessment model: add the newly emerging anomaly types and feature parameters to the anomaly identification knowledge base, adjust the weights and scoring standards of each indicator in the robot capability assessment model, and improve the accuracy and optimization of subsequent scheduling.

[0036] The specific embodiments described are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Identical components are represented by the same reference numerals. Therefore, all equivalent changes made to the structure, shape, and principle of this application should be included within the scope of protection of this application.

Claims

1. A scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios, characterized in that, The scheduling method includes the following steps: S1: Configure a multi-source sensing unit for each industrial robot and build a multi-dimensional anomaly perception model to achieve real-time monitoring and accurate identification of robot anomalies. The multi-dimensional anomaly perception model includes a multi-source sensing unit, anomaly feature fusion rules and anomaly identification knowledge base. The multi-source sensing unit collects robot operating status, working environment and task progress data, and after feature fusion, compares them with the anomaly identification knowledge base to complete anomaly identification. S2: Design a hierarchical replacement scheduling framework to generate an initial replacement scheme. First, use the robot capability assessment model to quantitatively evaluate and rank normal robots. Then, combine the replacement task parameters to screen candidate replacement robots and allocate tasks according to the load margin to form an initial scheme. S3: An improved adaptive scheduling algorithm is used to optimize the replacement scheme. The algorithm is a particle swarm-genetic fusion algorithm. The initial replacement scheme is used as the initial population. The global optimal scheme search is completed by combining the multi-objective optimization function and constraints. S4: Establish a dynamic adjustment mechanism to achieve real-time optimization of the fill-in process, collect fill-in execution data in real time, and update the evaluation model and task parameters and re-optimize the scheme when the preset trigger conditions are met. S5: After the replacement is completed, status feedback and knowledge base update are performed. The task completion status is analyzed through scheduling effect evaluation indicators. Based on this, the anomaly identification knowledge base content is supplemented and the robot capability evaluation model parameters are adjusted.

2. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 1, characterized in that: The multi-source sensing unit in S1 includes an operating status sensor, an environmental sensing sensor, and a task progress acquisition module. The parameters collected by the operating status sensors include joint temperature, motor speed, power supply voltage, load torque, and motion error. The parameters collected by the environmental perception sensor include obstacle distance, ambient temperature, and light intensity; The parameters acquired by the task progress acquisition module include task completion progress, remaining task quantity, and task deadline.

3. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 1, characterized in that: The abnormal feature fusion rule in S1 adopts a weighted fusion algorithm, in which the weight of the running status parameter accounts for 60%-70%, the weight of the task progress data accounts for 20%-30%, and the weight of the environmental data accounts for 10%-20%. The anomaly recognition knowledge base stores feature thresholds and judgment rules for common anomaly types such as mechanical failure, insufficient energy, sensor failure, and task overload.

4. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 3, characterized in that: The anomaly identification process in S1 is as follows: the data collected by the multi-source sensing unit is cleaned, denoised and standardized preprocessed, a uniform scale feature value is generated by a weighted fusion algorithm, the feature value is compared with the feature threshold in the anomaly identification knowledge base, and the abnormal robot number, anomaly type, occurrence time and impact range are output in combination with the judgment rules.

5. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 4, characterized in that: The evaluation indicators of the robot capability assessment model in S2 include current load rate, operation efficiency, motion accuracy, remaining energy, and distance to the abnormal robot operation area. The weights of each indicator are determined by the analytic hierarchy process, with the current load rate accounting for 30%-40%, the distance accounting for 20%-30%, the operation efficiency accounting for 15%-20%, and the remaining energy and motion accuracy each accounting for 5%-10%. The weight calculation results pass the consistency test.

6. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 5, characterized in that: The selection criteria for candidate replacement robots in S2 are: comprehensive capability score not lower than a preset threshold, current load rate not exceeding 50%, and distance from abnormal work area not greater than a preset distance. The initial replacement scheme generation rule is as follows: when the candidate set contains only 1 robot, all tasks are assigned; when it contains multiple robots, tasks are split according to the load margin to ensure that the load rate of each robot does not exceed 70% after replacement.

7. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 5, characterized in that: The optimization objectives of the multi-objective optimization function in S3 include the shortest completion time of the replacement task, the lowest total energy consumption of the robot, and the highest task completion accuracy. The constraints include that the robot's movement speed does not exceed the maximum limit, the load rate does not exceed the upper limit, the energy consumption does not exceed the remaining energy capacity, and the working position is within the preset area boundary.

8. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 5, characterized in that: The optimization process of the improved adaptive scheduling algorithm in S3 is as follows: using the initial replacement scheme as the initial population, firstly, the particle position and velocity are updated through the particle swarm optimization algorithm to achieve a global search and screen for globally better candidate schemes; then, genetic operations are performed on the candidate schemes to perform local optimization, and the above process is repeated until the change in the optimal fitness of the population is less than 5 times for 5 consecutive generations. Output the optimal solution.

9. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 4, characterized in that: The dynamic adjustment triggering conditions in S4 include: a new anomaly occurs in the replacement robot, the deviation between the task completion progress and the plan is not less than 10%, the energy consumption rate exceeds the preset threshold, and a new obstacle appears in the work area; The adjustment process is as follows: update the robot capability assessment model and the replacement task parameters, call the improved adaptive scheduling algorithm in step S3 to regenerate the replacement scheme and issue it.

10. The scheduling method for multiple robots to fill in for abnormal situations in industrial scenarios according to claim 1, characterized in that: The evaluation indicators for the supplementary scheduling effect in S5 include task delay rate, accuracy achievement rate and energy consumption optimization rate. The anomaly recognition knowledge base has been updated with new anomaly types and their corresponding feature parameters, while the robot capability assessment model has been updated with the weights and scoring criteria for each assessment indicator.

Images