Model-based fault detection of Modular and Reconfigurable Robots with joint torque sensing

Abstract

In this paper, a Model-based approach to fault detection is proposed that is not reliant on measurement or estimation of joint acceleration. The proposed fault detection algorithm is intended for Modular and Reconfigurable Robots (MRR) with joint torque sensing. It functions by comparing the filtered joint torque command with a filtered torque estimate derived from the corresponding nonlinear dynamic model of MRR incorporating joint torque sensing. The proposed fault detection scheme is ideal for detecting faults in modular robots because of its independence on motion states of other modules for fault detection. Experiments were performed using three common faults associated with joint actuator and the results confirmed the theoretical proposal by successfully detecting faults independently for each joint module.