Distributed multi-level adaptation for dynamic multi-agent systems

This paper proposes a distributed and multi-level adaptation system (MADAM) for agents to react to the changes of the world. The design of three adaptation levels in MADAM-global, middle, and local - allows agent to have both conscious and unconscious adaptive behaviors. At the local level, we use a rule-based instinctive reaction mechanism to simulate the unconscious adaptation. At the middle level, we use probabilistic reasoning to produce a set of feasible actions based on the current situation, then a fuzzy estimation is used to measure the fitness of these actions in the default cooperative plan and to choose the most suitable one. Finally, at global level, when the cooperative plan does not contain feasible actions, an adaptation is activated by changing the cooperative plan to suit the feasible actions. We apply the proposed adaptation in the soccer game. Our experimental results show that the adaptation system, MADAM, helps the agents (soccer players) achieve goals more frequently