Detecting and locating faults in the control software of autonomous mobile robots

摘要

Research on diagnosis has a long history in arti-cial intelligence which includes work dealing with diagnosis and repair of autonomous systems like space probes and software debugging. In this pa-per we extend previous work by introducing model-ing principles for software architectures that allow to detect and locate failures of the robot's control software. Failure detection is based on observers, i.e., software that monitors the behavior of the con-trol software. Localization is based on a model of the architecture and uses model-based diagnosis for computing the failure causes. Moreover, we intro-duce an algorithm for repairing the software and discuss rst experimental results of our implemen-tation. 1