Study On Conflicts Resolution For Multi-Agent Systems And Dynamic Path Planning For Intelligence Robots

This Dissertation theoretically analyses the Multi-Agent Systems (MAS), coordinate principles and strategies of conflicts resolution (CR),and, based on the BDI mental state model, studies the Extension inference principle and Extension methods for conflicts resolution. The dissertation also proposes the dynamic path planning for mobile robots and decision making Expert system for soccer game. The major results of the thesis are outlined as follows:1. Based on the coordinate principle and the conflicts resolution, the conflicts are classified into three categories: resource conflict, goal conflict and result conflict. The concept of Objective conflict and Subjective conflict are presented. According to Extension theory, a matter-element model of Agent which is based on Belief (B) and Intention (I) of opponents, is defined. Concerning the? shortcomings of Arbitration method and Negotiation method, a new way for conflicts resolution is presented, which results in a good resolution of Objective conflicts in MAS and Subjective conflicts in Multi-MAS, and remedies defects in Arbitration method and Negotiation method, by using Extension inference principle based on adapting matter-element properties, matter-element operation and Transform bridge.2. Dynamic eluding for mobile robots is a kind of conflicts. A soccer robot is an Agent and the block-eluding is a Subjective conflict, when the soccer game is a implement platform of MAS. Based on the transposition principle in DAI(Distributed Artificial Intelligent) , Fuzzy inference and control are adapted when opponent's intention between the block and eluding for soccer robots is analyzed. Agent's intention, and intelligent dynamic path planningfor soccer robots are also studied.3. After an environment model described by quad-tree had been analyzed, A algorithm BA* based on binary tree representation for environment and algorithm A*of AI(Artificial Intelligent), is proposed to study the path planning for global and local path planning. Another algorithm DBA*, which represents the environment by dynamic binary trees and reduces the search range, is also proposed to optimize algorithm BA*. The simulation of soccer games shows time saving in real-time search and an optimal path planning.4. Based on an analysis about the construction of Soccer game, a decision making expert system with forward inference is proposed, Which theinference is in layers and the knowledge is in classes. The system supports intelligent inference for decision making, roles assigning and path planning in soccer game.