Research on agile scheduling system based on Multi

2022-09-21
  • Detail

Research on agile scheduling system based on Multi-Agent and rule scheduling

Abstract: the dynamic scheduling problem of manufacturing workshop production process in agile manufacturing environment is studied. According to the special requirements of agile scheduling, this paper puts forward a method to realize the dynamic scheduling of production process in agile manufacturing workshop by comprehensively using multi-agent mechanism and rule scheduling, establishes the framework of production process dynamic scheduling system based on multi-agent production organization and operation mode, studies the method to realize rule scheduling on the basis of multi-agent structure, and creates the model structure of agent that adapts to the actual production environment, The idea and feasibility of the proposed method are illustrated by the simulation of dynamic scheduling of a class of agile machining workshops

Keywords: multi-agent; Rule scheduling; Dynamic scheduling; Agile manufacturing

0 introduction

as an advanced manufacturing mode with a new concept for enterprises in the 21st century, agile manufacturing integrates the mechanisms of many advanced manufacturing modes such as concurrent engineering and sophisticated manufacturing, and produces personalized products with customer satisfaction at the lowest cost [1]. Production scheduling mainly realizes the rational allocation and effective utilization of shared resources such as machines, materials and manpower under certain constraints according to the current situation and production objectives of the production system, so as to achieve the purpose of minimizing production costs [2]. Because the production of agile manufacturing workshop has the characteristics of multiple varieties, small batches, short manufacturing cycle and high quality requirements, the frequency of uncertain events (such as sudden addition or cancellation of processing tasks, randomness of workpiece arrival, operation delay, machine failure, etc.) is significantly higher than that of traditional manufacturing environment. When the production task changes and emergencies occur, the original operation plan should be modified in combination with the feedback information on site, and the system resources should be dynamically reorganized, so that agile manufacturing can operate continuously, orderly and efficiently. Due to the special requirements of agile manufacturing, production scheduling is more of a dynamic scheduling problem

common methods of dynamic scheduling include rule scheduling method, operations research method, cooperative solution method based on artificial intelligence, simulation method, etc. [3]. Rule scheduling method refers to the method of deciding the next operation according to certain rules and policies when the system is running. Its advantage is that it does not need to do a lot of calculations and avoids the "combinatorial explosion problem". As long as the appropriate rules are selected, the corresponding scheduling strategy can be generated, which is convenient and easy. The disadvantage is poor flexibility, which makes it difficult to adapt to uncertain changes. Therefore, it is difficult to effectively solve the dynamic scheduling problem of agile manufacturing workshop by simply applying rule scheduling method. In recent years, agent has been applied more and more in many fields, such as concurrent engineering, distributed solution and so on. Its main characteristics are autonomy, sociality, initiative and responsiveness [4]. The multi-agent system has the powerful advantage that the topology can be changed dynamically, that is, the age nt in the system can be increased or withdrawn dynamically without affecting other components in the system, and the system does not need to restart

in order to meet the requirements of agile manufacturing system for flexibility and rapid reorganization, this paper proposes to take the distributed multi-agent system as a new production organization and operation mode, and comprehensively use the multi-agent mechanism and rule scheduling method to realize the dynamic scheduling of the production process in agile manufacturing workshop

1 framework model of multi-agent scheduling system

multi agent system is a distributed system with coordination and negotiation mechanism, which is composed of management agent, production agent group and workpiece agent group. The main agents of the system are defined as follows

1.1 manage agent - a virtual workshop dispatcher

the functions of the management agent mainly include task planning, management, coordination and monitoring of other agents, as well as the update and maintenance of its own rule base

1.2 manufacturing agent - agent of production equipment in the workshop

all processing, transportation, assembly, warehouse and other equipment and devices mainly used for production in the agile manufacturing workshop can be defined as agents to form a production agent group

(1) machine agent - the agent of the processing device in the workshop, which is a subset of the production agent. Each processing agent is connected with an independent processing unit through the device interface. The machining unit is composed of one or more machining centers or machine tools and has machining functions. It can be defined as a triple:

m =

where: Mid - the unique identifier of the processing agent; σ—— The attribute, state and information characteristics of agent; α—— A set of operations that act on the agent, describing its behavior

(2) transportation agent - the agent of the transportation device in the workshop, which is a subset of the production agent. Each transport agent is connected with an independent transport machine, automatic guided vehicle (AGV) or mobile transport robot. Transportation agent can be defined as a triplet:

C =

where: CID - the unique identifier of transportation agent; τ—— The status and information characteristics of transportation agent; β—— A set of operations that act on the agent, describing its behavior

similarly, major equipment such as assembly devices, component warehouses or finished product warehouses in the manufacturing workshop can also be similarly defined. Workpiece is the processing object of other agents. During the processing, its state changes constantly, which directly reflects and affects the operation of other agents, so it is also defined as an agent component

1.3 part agent - the agent of the workpiece to be processed in the workshop. It can be defined as a triple

p= ID, ω,δ

where: PID - the unique identifier of the workpiece agent; ω—— The status and information characteristics of the artifact agent; δ—— The task information of the workpiece connected to the agent is dynamically assigned or cancelled by the management agent

the framework of multi-agent production scheduling system is shown in Figure 1. Multi agent works under the support of Internet or enterprise internal environment. The system communication and integration adopt standard protocols such as tcp/ip to ensure the information interaction between agents. Because the number of agents working in the system under different conditions is uncertain, the blackboard method can be selected for the communication between agents

2 agent model structure

agent can be divided into intelligent cognitive agent and non intelligent reactive agent. The main function of the workpiece agent in the above system is to dynamically calibrate its own state and stimulate the process of other production agents. It does not have the ability of judgment and reasoning. Therefore, it is designed as a non intelligent reactive agent. The management agent and production agent have certain reasoning ability, and can reason and make decisions based on their own state, task objectives, environmental characteristics and other information, with the support of distributed rule base. Therefore, it can be designed as an intelligent cognitive agent. Each agent formulates and executes specific scheduling strategies according to the combination of independent decision-making and mutual negotiation, so as to drive the operation of the production system in a clear and orderly manner

the structure of the management agent is shown in Figure 2. After receiving the production task from the workshop manager through the man-machine interface, it negotiates with the production agent, determines the operation plan, completes the optimal management of the system production resources and equipment, and receives the feedback information of the production agent to realize the monitoring of the production process. It also releases the workpiece list according to the production task, determines the workpiece processing priority, and dynamically assigns the workpiece to the workpiece agent combined with the working condition information. Agents communicate through communication interface

the structure of each production agent is similar, but the specific functions of their planning modules are slightly different, as shown in Figure 3. The production agent is connected with the controller of the specific equipment through the equipment interface. The planner can be further divided into several functional modules according to the complexity of the controlled equipment, and can work in parallel. All scheduling rule bases of each production agent are designed and maintained by themselves and encapsulated in the agent, which lays the foundation for the realization of distributed rule scheduling

The structure of the

artifact agent is shown in Figure 4. Due to the different types and quantities of workpieces in different processing tasks and production stages, the number of workpiece agents in the system is also changing dynamically. The maximum quantity is set by the management agent according to experience, and the empty workpiece agent is allowed to exist in the specific production scheduling. The task data of the workpiece agent is dynamically assigned through communication with the management agent and stored in the database, including the type code, serial number code, processing procedure and processing priority of the workpiece. The status attribute is dynamically assigned by the processing agent or the transportation agent through the management agent according to the actual situation of the workpiece corresponding to the agent, and becomes a condition to stimulate other agents to call and execute rules

in this way, in addition to the management agent playing a macro role in global scheduling and coordination, each production agent plays a micro role in local planning and scheduling

3 rule scheduling mechanism under multi-agent structure

rule scheduling strategy is applied to agile manufacturing workshop system. First, we need to construct a sufficient rule base that can be dynamically modified and updated. Secondly, it can master the real-time information of the workshop and select the appropriate scheduling rules. Because ag ent has the ability of autonomous decision-making and relative independence, it can respond to environmental changes quickly. Therefore, the rule scheduling mechanism under the multi-agent structure proposed in this paper can meet the above requirements

each agent defines its own rule base according to its own characteristics and tasks, and can dynamically construct, modify and update its own scheduling rules. The rule base is encapsulated in the agent. The management agent has the function of global common rule base, which is mainly used for the setting of the initial state of the agent in the system, the overall planning and coordination of multi-agent systems, etc. The rule base of the production agent is constructed according to its own functions and characteristics, and is called according to specific conditions during operation. When there is no suitable rule matching, the agent automatically starts the real-time scheduling module and updates the rule base with the feasible scheduling strategy as a new rule. The establishment and maintenance of specific rule base can adopt conventional methods (such as if... Then... Form) to describe and establish scheduling rules

when the system is running, each agent works in parallel according to its own rules. When it is necessary to add or cancel some production agents, the interactive objects between agents may change. However, because the rule base is encapsulated in the agent, it will be dynamically reorganized with the reorganization of the agent. Each rule base does not need to be modified. The system uses the intelligence and interaction of the agent to coordinate under the overall planning of the management of ag ent. The production agent shall notify the management agent and other agents every time it completes a work; The management agent understands the operation of the system according to the status of each agent, so that the scheduling system can quickly grasp the real-time information of the workshop, and create good conditions for the selection and implementation of specific rules

agent mode is combined with rule scheduling strategy to meet the concurrency and synchronization requirements of dynamic scheduling. Rules select appropriate operation steps for agents to realize orderly scheduling; The agent structure supports the distribution of rules, which makes it easy for the rule base to modify and supplement independently, and makes up for the shortcomings that the rule scheduling does not adapt to dynamic changes, and the rule base is huge and difficult to build and maintain. Therefore, the complexity of dynamic scheduling is greatly reduced and the effectiveness of rule scheduling is improved

4 agile scheduling example

4.1 problem description

Figure 5 is the schematic diagram of the agile manufacturing workshop with five machining units. Each machining unit in the system can be a CNC machine tool or machining center, which automatically

Copyright © 2011 JIN SHI