Multi-Agent Systems - An Approach to Design and Implement Applicable Software Architecture, World Modeling, and Decision Making

The need to have well-structured and managed software architecture for control, behavior, and decision making of multi agent environments is inevitable. This is due to the fact that these intelligent agents are expected to perform within a dynamically complex environment where proper and quick response in a cooperative manner is a primary key to successfully complete a task. This thesis follows two main trends. The first trend is to design a model for the agents’ software architecture, and the latter trend is to implement a nondeterministic decision making using electric fields and high-level decision making. Different layers are designed, defined, and implemented for the software architecture with focus on system adaptability, sustainability, and optimization. The approach proposed in this thesis for the agents’ decision making is based on layered artificial intelligence implemented using vector-based fuzzy electric fields and a decision tree. Furthermore, an approach to model the world which is, in this thesis, called Agent Relative Polar Localization is introduced. This world model is based on fuzzy measurements and polar coordinates. This thesis also proposes a special supplementary flexible decision tree which may and can be modified based on the application under which it is intended to be used or implemented. The aim of designing such decision tree is to provide more flexibility for the decision making process. In order to optimize the overall performance of the system learning methods have been introduced to the system. The proposed system in this thesis has been implemented on soccer robots to evaluate the performance of the system and see that theoretical studies also can be applied to the real applications. The results show that the proposed system implemented on the soccer robots is reliable and robust