Search

CN-121998354-A - Multi-agent dynamic scheduling method and system based on efficiency and energy multi-objective optimization

CN121998354ACN 121998354 ACN121998354 ACN 121998354ACN-121998354-A

Abstract

The invention provides a multi-agent dynamic scheduling method and system based on efficiency and energy multi-objective optimization. Belongs to the technical field of intelligent manufacturing, multi-agent system and production scheduling optimization intersection. The method comprises the steps of constructing a global coordination and local execution double-layer agent framework, generating a hierarchical dispatching network topology, carrying out dynamic modeling on robot clusters, generating a hundred-stage robot state data set, initializing a task pool and a real-time environment map to form initial dispatching scene data, and enabling a robot to complete tasks in a more reasonable mode through joint optimization of task allocation and path planning, so that the production throughput rate is remarkably improved.

Inventors

  • YANG YIMING
  • LIU WEI

Assignees

  • 深圳墨影科技有限公司

Dates

Publication Date
20260508
Application Date
20260128

Claims (10)

  1. 1. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization is characterized by comprising the following steps: S1, constructing a global coordination and local execution double-layer agent architecture, generating a hierarchical scheduling network topology, carrying out dynamic modeling on robot clusters, generating a hundred-stage robot state data set, initializing a task pool and a real-time environment map, and forming initial scheduling scene data; S2, inputting initial scheduling scene data into an improved NSGA-II algorithm module, performing multi-objective optimization modeling to generate a pareto front solution set, and performing space-time collision detection and dynamic obstacle avoidance processing on candidate paths in the pareto front solution set through a space-time A algorithm to generate a combined optimization path scheme set; S3, evaluating the matching degree of tasks and robots according to the combined optimization path scheme set, generating a double-target weighted fitness function value, screening an optimal task allocation scheme, synchronously generating an individual path planning instruction set of the robots, and transmitting the individual path planning instruction set to a local execution layer agent to form initial scheduling instruction data; s4, collecting robot motion state data and task execution progress data in real time through a local execution layer agent, generating a dynamic scheduling feedback data stream, detecting a real-time disturbance event, and generating a disturbance type identification and influence range evaluation report; S5, triggering an online rescheduling mechanism according to the disturbance type identification and the influence range evaluation report, solving the updated joint optimization problem through the cooperation of the improved NSGA-II algorithm module and the space-time A algorithm, generating a rescheduling path correction scheme set, and performing fusion processing on the rescheduling path correction scheme set and the initial scheduling instruction data to generate a dynamic scheduling update instruction; S6, performing real-time scheduling operation of the hundred-stage robot system based on the dynamic scheduling update instruction, generating scheduling performance evaluation data, adjusting multi-objective optimization weight parameters of the global coordination layer, forming closed-loop self-adaptive optimization circulation, and finally generating optimized scheduling result data.
  2. 2. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 1, wherein the S1 comprises: S11, integrating a distributed communication protocol and a hierarchical decision mechanism, constructing a global coordination layer and a local execution layer double-layer intelligent body architecture, and generating a hierarchical scheduling network topology; s12, carrying out parameterization modeling on a kinematic model, an energy consumption model and a task execution capacity model of the hundred-stage robot according to node function division of hierarchical dispatching network topology, and generating a hundred-stage robot state data set; S13, initializing a multi-type task pool based on capability boundary parameters of a hundred-stage robot state data set, synchronously acquiring environment obstacle information and a passable path network through a laser radar and vision SLAM technology, and generating initial scheduling scene data fusing task information and environment information.
  3. 3. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 1, wherein the step S2 comprises: S21, inputting task constraint conditions, robot performance parameters and environment constraint information in initial scheduling scene data into an improved NSGA-II algorithm module, constructing a multi-objective optimization model with minimized task completion time and minimized energy consumption accumulation through a self-adaptive weight coding mechanism, and generating a pareto front solution set meeting the pareto optimality; S22, extracting space-time coordinate sequences of candidate paths in the pareto front solution set, constructing a four-dimensional collision detection model through a space-time A-based algorithm, performing millisecond time slice division and space overlapping degree calculation on a robot motion track, and performing real-time planning on a dynamic obstacle avoidance path; S23, carrying out energy consumption secondary calibration on the candidate paths by combining the obstacle avoidance result, and generating a combined optimization path scheme set.
  4. 4. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 3, wherein S22 comprises: extracting space-time coordinate sequences of candidate paths in the pareto front solution set to generate a candidate path space-time coordinate sequence set; Based on a candidate path space-time coordinate sequence set, constructing a four-dimensional collision detection model through a space-time A algorithm, and generating a four-dimensional collision detection model structure; based on a four-dimensional collision detection model structure, millisecond time slice division is carried out on the motion track of the robot, and a track time slice division result is generated; Calculating the spatial overlapping degree of the motion trail of the robot in each time slice by combining the trail time slice dividing result, and generating spatial overlapping degree data; and identifying potential collision risks based on the spatial overlapping data, performing real-time planning of the dynamic obstacle avoidance path, and generating a dynamic obstacle avoidance path planning result.
  5. 5. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 1, wherein the step S3 comprises: S31, key features are extracted from the joint optimization path scheme set, a task and robot double-target matching degree evaluation matrix is constructed, each feature weight is determined through an entropy weight method, and a double-target weighting fitness function value is generated; s32, non-dominant sorting and crowding screening are carried out on all task allocation combinations based on the double-target weighted fitness function value, invalid schemes exceeding the load capacity of the robot and violating task time constraint are removed, and a global optimal task allocation scheme is screened out; S33, generating a robot individual path planning instruction set aiming at each robot individual in the optimal task allocation scheme, and performing instruction format adaptation and integrity check through a communication interface of the global coordination layer and the local execution layer; S34, issuing the verified robot individual path planning instruction set to the local execution layer intelligent agent, and carrying out instruction analysis and execution preparation by combining a robot hardware driving protocol to form initial scheduling instruction data.
  6. 6. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 1, wherein the step S4 comprises: S41, acquiring robot data in real time through a multi-sensor fusion module carried by a local execution layer intelligent agent, and generating a dynamic scheduling feedback data stream through Kalman filtering noise reduction and data format standardization processing; s42, based on the dynamic scheduling feedback data flow, a disturbance event feature library is constructed, real-time data matching and anomaly detection are carried out through a deep learning model, disturbance event types are identified, and disturbance type identifiers are generated; s43, performing quantitative analysis on the influence range according to the identified disturbance event and combining the task dependency relationship and the robot cluster topological structure in the initial scheduling scene data to generate a disturbance influence range evaluation report.
  7. 7. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 1, wherein the step S5 comprises: S51, setting a rescheduling triggering threshold according to disturbance type identification and an influence range evaluation report, starting an online rescheduling mechanism, and synchronously updating a task pool state, a robot state and an environment map; S52, inputting updated scheduling scene data into an improved NSGA-II algorithm module, carrying out multi-objective optimization modeling again, and carrying out collision detection and obstacle avoidance optimization on the corrected path by combining a space-time A algorithm to generate a rescheduling path correction scheme set; s53, comparing and analyzing the rescheduling path correction scheme set and the initial scheduling instruction data by adopting a time sequence alignment and conflict resolution algorithm to generate instruction fusion intermediate data; s54, millisecond compression processing and transmission delay compensation are carried out on the instruction fusion intermediate data, a real-time control protocol of the robot is adapted, and a dynamic scheduling update instruction meeting millisecond response requirements is generated.
  8. 8. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 7, wherein S53 comprises: Adopting a time sequence alignment and conflict resolution algorithm to perform time sequence dimension alignment treatment on the rescheduling path correction scheme set and the initial scheduling instruction data to generate a double data set after time sequence alignment; Based on the double data sets aligned in time sequence, carrying out item-by-item comparison analysis of instruction contents, identifying an instruction item affected by disturbance and an effective instruction item unaffected by disturbance, and generating an instruction influence identification result; Based on the instruction influence identification result, extracting effective instruction items which are not influenced by disturbance, and summarizing to form an effective instruction subset; Aiming at the affected instruction items marked in the instruction influence identification result, invoking the corresponding correction instruction content in the rescheduling path correction scheme set to replace, and generating a replaced instruction subset; and fusing the effective instruction subset and the replaced instruction subset, and integrating to form instruction fusion intermediate data.
  9. 9. The multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization according to claim 1, wherein the step S6 comprises: S61, based on a dynamic scheduling update instruction, driving a robot executing mechanism to carry out motion adjustment and task execution through a local executing layer agent, and synchronously collecting data in a scheduling process to generate multidimensional scheduling performance evaluation data; s62, carrying out normalization processing and weight sensitivity analysis on the scheduling performance evaluation data, and adjusting task completion time weight and energy consumption weight parameters of the global coordination layer through a hierarchical analysis method; S63, feeding back the adjusted weight parameters to a multi-objective optimization modeling link, and updating objective function weight configuration of an improved NSGA-II algorithm to form a closed-loop self-adaptive optimization loop; S64, continuously improving the multi-objective optimization performance of the scheduling scheme through multi-round closed loop iterative optimization, and finally generating the optimized scheduling result data.
  10. 10. A system for implementing the multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization of claim 1, the system comprising: The dynamic modeling module is used for constructing a global coordination and local execution double-layer agent framework, generating a hierarchical scheduling network topology, carrying out dynamic modeling on robot clusters, generating a hundred-stage robot state data set, initializing a task pool and a real-time environment map, and forming initial scheduling scene data; The dynamic obstacle avoidance module inputs the initial scheduling scene data into an improved NSGA-II algorithm module, performs multi-objective optimization modeling to generate a pareto front solution set, and performs space-time collision detection and dynamic obstacle avoidance processing on candidate paths in the pareto front solution set through a space-time A algorithm to generate a combined optimization path solution set; The path planning module is used for carrying out task and robot matching degree evaluation according to the combined optimization path scheme set, generating a double-target weighted fitness function value, screening an optimal task allocation scheme, and synchronously generating a robot individual path planning instruction set; The dynamic scheduling module is used for acquiring the motion state data of the robot and the task execution progress data in real time through the local execution layer agent, generating a dynamic scheduling feedback data stream, detecting a real-time disturbance event and generating a disturbance type identification and influence range evaluation report; The joint optimization module triggers an online rescheduling mechanism according to the disturbance type identification and the influence range evaluation report, and solves the updated joint optimization problem in a cooperative way through the improved NSGA-II algorithm module and the space-time A algorithm to generate a rescheduling path correction scheme set, and performs fusion processing on the rescheduling path correction scheme set and initial scheduling instruction data to generate a dynamic scheduling update instruction; And the performance evaluation module is used for executing real-time scheduling operation of the hundred-stage robot system based on the dynamic scheduling update instruction, generating scheduling performance evaluation data, adjusting multi-objective optimization weight parameters of the global coordination layer, forming closed-loop self-adaptive optimization circulation and finally generating optimized scheduling result data.

Description

Multi-agent dynamic scheduling method and system based on efficiency and energy multi-objective optimization Technical Field The invention provides a multi-agent dynamic scheduling method and system based on efficiency and energy multi-objective optimization, and belongs to the technical field of intelligent manufacturing, multi-agent system and production scheduling optimization intersection. Background In the field of intelligent manufacturing, as the production scale is continuously expanded, the scene of hundred-stage robot collaborative operation is more common. However, current multi-robot scheduling methods suffer from a number of deficiencies in coping with such complex scenarios. Most existing systems focus on single-objective optimization only, or simply pursue the shortest task completion time, or only pay attention to reducing task delay, but neglect the energy consumption accumulation effect caused by the moving distance of the robot. Although the single-target optimization mode can improve local efficiency, high efficiency is often caused along with high energy consumption, and high-efficiency utilization of global resources cannot be realized. Meanwhile, the traditional scheduling methods mostly adopt static allocation strategies, such as Hungary algorithm, auction mechanism and the like, the methods allocate tasks once at the initial time of the tasks, and real-time disturbance such as dynamic task insertion, robot faults, path blocking and the like is difficult to deal with. Once an emergency occurs, the whole dispatching system may be confused, cannot be adjusted in time, and seriously affects the production efficiency. In addition, the decision level splitting problem of task allocation and path planning is also very prominent. The high-level task allocation and the bottom-layer path planning lack coordination, and the allocated robots are always the closest to the task, but the actual path needs to bypass, so that the energy consumption and the task completion time are increased. Moreover, the centralized solver solves overtime when facing hundred-stage robots, and the pure reinforcement learning method is difficult to converge due to state space explosion and has poor expandability. Therefore, a new scheduling method that can achieve multi-objective optimization of efficiency and energy, has dynamic rescheduling capability, and is suitable for hundred-stage robots is needed. Disclosure of Invention The invention provides a multi-agent dynamic scheduling method and system based on efficiency and energy multi-objective optimization, which are used for solving the problems mentioned in the background art: The invention provides a multi-agent dynamic scheduling method based on efficiency and energy multi-objective optimization, which comprises the following steps: S1, constructing a global coordination and local execution double-layer agent architecture, generating a hierarchical scheduling network topology, carrying out dynamic modeling on robot clusters, generating a hundred-stage robot state data set, initializing a task pool and a real-time environment map, and forming initial scheduling scene data; S2, inputting initial scheduling scene data into an improved NSGA-II algorithm module, performing multi-objective optimization modeling to generate a pareto front solution set, and performing space-time collision detection and dynamic obstacle avoidance processing on candidate paths in the pareto front solution set through a space-time A algorithm to generate a combined optimization path scheme set; S3, evaluating the matching degree of tasks and robots according to the combined optimization path scheme set, generating a double-target weighted fitness function value, screening an optimal task allocation scheme, synchronously generating an individual path planning instruction set of the robots, and transmitting the individual path planning instruction set to a local execution layer agent to form initial scheduling instruction data; s4, collecting robot motion state data and task execution progress data in real time through a local execution layer agent, generating a dynamic scheduling feedback data stream, detecting a real-time disturbance event, and generating a disturbance type identification and influence range evaluation report; S5, triggering an online rescheduling mechanism according to the disturbance type identification and the influence range evaluation report, solving the updated joint optimization problem through the cooperation of the improved NSGA-II algorithm module and the space-time A algorithm, generating a rescheduling path correction scheme set, and performing fusion processing on the rescheduling path correction scheme set and the initial scheduling instruction data to generate a dynamic scheduling update instruction; S6, performing real-time scheduling operation of the hundred-stage robot system based on the dynamic scheduling update instruction, generating scheduling per