| After reviewing the studies on the theory and application of the Agent and Multi-Agent System, the novel concept of Compound-Agent is proposed, Compound-Agent is a special kind of Multi-Agent System, which consists of Monitor-Agent (MA) and Execute-Agent (EA). Members of Compound-Agent can be Compound-Agents, i.e. it can be compounded more than once. Compound-Agent can conveniently cooperate closely, learn and evolve together, has the capacity of diagnosing fault and recover, and the properties of real time, tolerance and robusticity. Compound-Agent can be used to construct the system with self-adaptability, rapider response and more intelligence, which adapts to the complex dynamic giant system better. The study on the Compound-Agent theory is divided into four parts: structure, function, cooperation model and learning algorithm. The structure of a Compound-Agent and the architecture and function of MA and EA are given. The principle and reason of cooperation for Compound-Agent is given. The cooperation models, including the cooperation model based on commitments and conventions, forecasting and voting, are given. The advantage of the cooperation model based on commitments and conventions is shown by the simulation of pursuit problem. The concept and meaning of Compound-Agent learning is described, and the machine learning algorithms, including reinforcement learning, scoring and revolution of rule base, are detailed, and simulation results are given.This application study is striving to the application of Compound-Agent to enterprise management. In order to improve enterprise management and intelligentize the management information system, we apply the Compound-Agent theory to enterprise management information system and construct Enterprise-Management-Compound-Agent (EMCA), which makes the enterprise own higher 'intelligence quotient', sooner activity and stronger competence. The architecture of EMCA and the function of the members of EMCA are provided. The realization of Production planning and scheduling Compound-agent (including JSCA and FSCA) and supply chain management Compound-Agent are given in great detail.JSCA is mainly based on DNA computing, A DNA-coding genetic algorithm based on the structure and the genetic mechanism of biological DNA is described, which is applied into the agent-based scheduling system for the first time. The studies on the production planning and scheduling are reviewed, and a mathematics model of this problem is constructed. The JSCA structure and the architecture, function and realization of the members of JSCA, workshop manage agent, resource agent and task agent are given. The detail of solving the production scheduling problem using the DNA genetic algorithm and its result is shown through simulation of the classic 6×6 Job-shop problem. A case study of the application of JSCA to the schedule problem of a finishing workshop of a wool textile mill shows its usability. Rule set for reasoning is obtained by mining from the result of DNA genetic algorithm. The rule base can be adjusted on time.The study of FSCA is designed for two types of scheduling objection-the Flowshop scheduling with the objective of minimizing the makespan and the Flowshop scheduling with the objective of delivery on time. The structure and realization of these two types of FSCA are given. The case studies of the production scheduling with the objective of minimizing the makespan in wool strip workshops and the production scheduling with the objective of delivery on time in spinning workshops show its effectiveness.A global economy and increase in customer expectations in terms of cost and services have put a premium on effective supply chain. The supply chain model can be composed from some members of EMCA, the gather of which is the Supply-Chain-Compound-Agent (SCCA). The early research of supply chain is summarized, and the concept of the supply chain and the purport of the supply chain management system are presented. The architecture of SCCA and t...
|
PDF Full Text Request