An unifying architectural methodology for robot control design with adaptive fault tolerance

Autonomous mobile robots typically interact with unknown environments and are subject to many kinds of faults. Adaptive fault tolerance is the dynamic adjustment of behavior of a system, as a consequence of variation of external or internal conditions or changes in immediate goals. This work presents a methodology for the design of robot control including adaptive fault tolerance. We use the hybrid architecture model, with control represented by finite state machine and dataflow graph. Redundancy is expressed by multiple paths to the same data element in the dataflow, resulting in different control configurations, allowing adaptation. Our approach intends to simplify the task of the robot control designer, transferring design complexity from program coding to description of components and interconnections. A control platform is responsible for the control flow, including phase transition, adaptations, and fault recovering. Using this platform we designed the control of a simple mission for a real robot