Sequential parameter estimation for fault diagnosis in mobile robots using particle filters

Abstract

The autonomous detection and handling of faults is an important skill for mobile robot systems. Faults in the motion-control system can strongly decrease the robots' performance or compromise its mission completely. In this paper, we demonstrate how a mobile robot system can, in case of a fault, switch to a richer internal system model and estimate the newly introduced parameters to reliably diagnose its state and possibly continue its operation. We discuss three methods for sequential parameter estimation using particle filters and evaluate their performance in physically accurate simulation runs.