Autonomous Control System Of AGVs Based On Multi-Agent

With the large scale, dynamic and open development of industrial automated assembly linesnowadays, Automatic Guided Vehicle System (AGVS) should become greater scale and morecomplex collaboration to serve the needs. With the increase of AGV number, traditional centralizedcontrol faces great challenges in calculation efficiency and load balancing. Multi Agent System (MAS)technique is based on Distributed Artificial Intelligence and presented to solve large scale andcomplicated problems. This paper researches the application of Agent technology in AGVs for moreefficient and flexible control.According to the researches on the basic principles and structure of agent and MAS technology, afull evaluation of agent technology application in AGVs is given, from individual behavior to thesystem operating characteristics. Specific implementation plans are confirmed that fit for AGVs.Based on the feasibility studies above, an intelligent single vehicle model is set following theindividual characteristics structure and working process of AGV. The model re-defines the originalsystem in four levels: decision-making, interaction, perception and behavior. This makes single AGVhave autonomous control and quickly adaption quality.On the basis of the intelligent model of single AGV, a MAS-based framework of AGVs is built.Coordination mechanism in autonomous control of AGVs is deeply investigated. Improvedblackboard architecture is used to solve the problem of conflict avoiding and traffic management. Thecontract net is used for task allocation in AGVs. A crossing control method based on adaptive traffic isproposed.Distributed control unit of single AGV and cooperative processor of the AGVs is developed.Both of the controller are based on embedded technology such as ARM and Zigbee. An user-friendlyclient management software for AGVs is released where object-oriented idea and database is widelyused. A combined testing is set up on the hardware and software above. Results indicate that theproposed model and running environment have good performance in multi-AGV control system.Finally, the whole work is summarized. Some suggestions are made on what needs in furtherstudy.