Fault-Tolerance and Error Recovery in an Autonomous Robot with Distributed Controlled Components

Most of the existing autonomous robot systems have a centralized hierarchical control architecture. In such robot systems, all planning, execution control, and monitoring tasks are performed by a single control unit on a defined level. In case of an error that occurs during the execution, this central control unit has the complete knowledge about the past executed actions and is able to reason on the error situation. Besides the centralized control architectures, distributed and decentralized control architectures have been developed to overcome some problems with the centralized systems. Because of the missing overall control, error recovery is more difficult than in centralized systems. This paper presents concepts to obtain fault-tolerance behaviour and error recovery in a distributed controlled robot system. As an example for such a robot system, the Karlsruhe Autonomous Mobile Robot KAMRO is considered that is being developed at IPR. Many experiments were performed with the former centralized control architecture. Our intention is to achieve the same and better results with the distributed control architecture KAMARA.