A Novel Framework for Fault Identification and Risk Management (F.I.R.M.) of Autonomous Systems
Abstract:
A novel framework based on adaptive neural network is proposed for fault identification and risk management using a three degree-of-freedom nonlinear model of autonomous underwater vehicles (AUVs). The proposed approach can be used for real time fault detection, identification and recovery (FDIR) in AUV actuators and sensors. An adaptive identifier (AID) is designed based on neural network (NN) and extended Kalman filter (EKF), which integrates a nonlinear observer for the AUV’s dynamic model to detect faults in the actuators and sensors. Based on the detected faults, an active fault-tolerant controller, including inner and outer loops of controller, is implemented for online compensating the faults. The simulation results showed that the proposed approach can detect and mitigate different types of faults on the AUV actuators, and the underwater vehicle. i.e. REMUS100 tracks the planned trajectory without any interruption.
