Distributed Structure And Motion Planing For Autonomous Robots

System architecture is an important subject in robotics.These years, many models have been proposed.They are the same in that the directions of information flows are fixed.In mater of fact,Intelligent robots are something very complex.It is impossible for some fixed model to satisfy all the basic requirements.A distributed system model is proposed based on the basic concept of multi-agent system(MAS)of distributed artificial intelligence(DAI).The model is composed by a series of functional units(agents),which is communicated and coordinated with each other by "news".The only requirement is that the agents are able to communicate.The objective of the system is achieved through the concurrent actions of all the agents.Performance could be changed on-line by information shielding,and different models are interchangeable.The supervisor theory of discrete event dynamic system(DEDS)based on finite state automaton is used as a toll in this paper.A constructional theory for DEDS system is established in this paper.The controllability of the proposed basic structure and extend structure are proofed;the conditions witch preserve the properties of consistence, controllability and supervisability for the agent construed by the basic and extend structure are proposed and proofed;several prototypes of supervisor satisfy some controlled behaviors and constrains are proposed and the effectiveness proofed;the condition for supervisor conjointion is proposed and proofed.A complete process for construction of discrete control system is proposed base on all these works.The process keeps the functions of the theory but get ride of all the complexity,so to bring the theory into play.A robot motion control architecture is studied in this paper as an example of distributed system.A guided global planning method including nonholonomic constraint is proposed.The method brings the flexibility and quickness of the local method into the reliable frame of global method.It could be embedded into different global methods.An autonomous vehicle and a manipulator are studied as a unit in this paper.The condition of parameter solvability under the work requirement of the end effector is discussed,and a method for inverse kinematics is proposed.The feasibility of the method is tested.An emulation system for control of a simplified wheeled robot is completed with the proposed distributed structure and the proposed construction method.The system has four agent,the task agent,the vision agent,the planning agent and the drive agent.The correctness and effectiveness of the proposed work are exhibited by the result of the emulation.