Interactive Multiple Neural Adaptive Observer based Sensor and Actuator Fault Detection and Isolation for Quadcopter

Abstract

This paper presents a fault detection and identification (FDI) method that can simultaneously deal with motor and sensor faults in a quadcopter. The method integrates Neural Adaptive Observers (NAOs) that predicts the errors in the dynamic model due to fault into an Interactive Multiple Model (IMM) framework. Two NAOs are constructed to deal with two different categories of faults – sensor faults and actuator faults, which are represented as two different models in the IMM filter. The stability of the proposed FDI scheme is theoretically analyzed, and validity of the method is demonstrated on a virtual physics engine environment.