The problem of controlling autonomous mobile robots (AMR's) in real world environments has received great attention in recent years. In order to face the high complexity of this problem, a recent research trend involves the use of distributed architectures for designing and implementing robot controllers. In this paper we present a novel distributed architecture for control of autonomous mobile robots. The proposed architecture includes many autonomous agents able to communicate and to cooperate in order to achieve the global problem solution. Each agent is specialized in a facet of the AMR control problem and no agent has enough information and capabilities to solve the entire problem. Agent are decentralized and loosely coupled, and their cooperation is based on a task-sharing approach. An example of application of the proposed architecture to an instance of the homing problem is given.
PDF file