Parity Relation Based Fault Detection, Isolation and Reconfiguration for Autonomous Ground Vehicle Localization Sensors

This paper considers fault detection, isolation and reconfiguration (FDIR) for the localization sensors, including the dead reckoning and external sensors, of an autonomous ground vehicle (AGV) designed for use in highly unstructured outdoor environments. Ten sensors are considered in this research. None of these sensors are identical, but subsets of them do have the ability to measure or calculate (based on simple algebra) the same kinematical parameters. To improve the localization accuracy, selected sensor outputs are fused using Kalman filters. The fused data and selected sensor measurements are then combined into a set of linearly independent parity equations, which leads to the generation of a bank of residuals. A fault in any one of the ten sensors causes a unique subset of these residuals to grow, which allows the fault to be detected and isolated. This allows a control scheme based on these sensors to reconfigure itself so that only the non-faulty sensors are used for localization. The effectiveness of this FDIR scheme is demonstrated in the context of a recently developed algorithm for maneuvering an AGV in cluttered environments. See also ADM001736, Proceedings for the Army Science Conference (24th) held in Orlando, Florida on 29 November-2 December 2004. The original document contains color images.