Active Fault Management in Autonomous Systems Using Sensitivity Analysis

摘要

The absence of human senses and experience in autonomous systems pose a variety of unforeseen challenges. One of these challenges is the effective health monitoring of autonomous systems. This paper proposes a comprehensive active fault management framework. The proposed framework works on the measured signal and control inputs of the system. The set of residuals and isolation tests, which is part of the passive fault diagnosis system, have the capability of adapting to new and unforeseen scenarios. If the fault is not isolable or detectable in magnitude, it will be excited by manipulating the inputs of the system in a controlled fashion. Once the fault is confirmed, it will be mitigated to minimize the performance degradation and damage to the system. The later part of the framework on active fault diagnosis (sensitivity analysis based fault excitation and mitigation) has been demonstrated for a powertrain of an autonomous electric vehicle. The simulation results confirm the effectiveness of the proposed active fault management framework.