Active fault tolerant control on multi-rotor UAVs using ELM-based neuro-fuzzy system

Abstract

Multi-rotor UAVs have been used prevalently for a multitude of purposes and the reliability of these UAVs has to be ensured to enable safe operations, especially in urban airspace. Hence, an accurate and efficient fault tolerant control strategy is essential to actively detect hardware faults and perform controller reconfiguration. In this paper, an active fault tolerant control strategy is introduced, which uses Online Sequential Fuzzy Extreme Learning Machine (OS-Fuzzy-ELM) to diagnose actuator faults. A gain scheduled PID (GS-PID) controller then uses the fault information to reconfigure among pre-tuned PID controller gains in real-time. The extreme learning neuro-fuzzy based fault tolerant control algorithm has high generalization performance despite nonlinearities and uses low computational load, which makes it suitable for online application. Upon implementing the fault tolerant controller into the system model of a quadrotor UAV, testing is conducted for hovering and trajectory tracking flight scenarios. The results indicate that the fault tolerant controller is successful in quickly identifying faults, suppressing large deviations and subsequently perform quick recovery.