Intelligent Control And Decision Making Based On The Self-organizing Multi-agent System

The design and implementation of an intelligent control and decision making system is currently a hot issue in the field of artificial intelligence research. This dissertation introduces the self-organizing multi-agent system into the research of the intelligent control and decision making system. In this dissertation, studied with the intelligent control and decision making theories and applications, the Physics-Inspired Multi-agent Interaction model and the Intelligent Control and Decision Making Model based on the Self-organizing Multi-agent System are designed. Meanwhile, a complete platform of the Physics-Inspired Self-organizing Multi-agent Intelligent Control and Decision Making System is also designed and implemented. The main contributions of the dissertation are summarized as follows:1. The basic concepts of self-organization and multi-agent system are illustrated. The presentations of the problem of combining the self-organization theory and multi-agent system theory are also illustrated, i.e. the self-organizing behaviors in the multi-agent system and the self-organizing multi-agent systems.2. A Physics-Inspired Multi-agent Interaction Model is proposed. The basic principle of this model is based on the classical Newtonian physics - Newton's law of universal gravitation and the artificially constructed virtual interaction forces among multiple agents. The model is modeling through the following four steps - the definition of interaction scenarios, the choice of attraction or repulsion interaction mechanisms, the definition of multiple agents'roles and behaviors and the choice of the multi-agent interaction process. Experimental results show that this model can be easily implemented and can overcome the shortcoming of insufficient calculating capacity of the traditional multi-agent interaction model with strong versatility and reliability.3. An Intelligent Control and Decision Making Model based on the Self-organizing Multi-agent System and a paradigm of its implementation algorithms are proposed. This model is consisted of two major components– Environment and Self-organization Multi-agent System. It is modeling by the following four steps - the definition of environment, the identification of environment, the process of multi-agent control and decision making and the output of multi-agent control and decision making. It can be specifically achieved by flexible choosing a variety of algorithms, and the paradigm of its implementation algorithms given in the dissertation is based on the interaction computing among multiple agents in the microscope, while as the core of the overall statistical computing of the multi-agent system by using the statistical physics in the macro scope. The model is full of use the group decision making advantages of the self-organizing multi-agent system. Experimental results show that the model and the paradigm of its implementation algorithms can be easily implemented with great flexibility and versatility.4. A complete platform of the Physics-Inspired Self-organizing Multi-agent Intelligent Control and Decision Making System is proposed and implemented. In the composition of this system platform, the Physics-Inspired Multi-agent Interaction subsystem and the Intelligent Control and Decision Making subsystem based on the Self-organizing Multi-agent System are as the core components. In order to realize the particular intelligent control and decision making functions, the additional subsystems include the environmental identification and input subsystem, the control and decision making output subsystem and the human-computer interaction subsystem. Experimental results show that this system platform has good validity, flexibility, robustness and adaptability. It also has a strong ability to the complex real-time control environment and the feasibility to the actual environment for specific applications.5. The summary of the whole text and the ideas of the further research works are given at the end. The prospect of future research directions is also discussed.