Model-Based Fault Diagnosis Algorithms for Robotic Systems

Abstract

In this paper, three model-based fault diagnosis algorithms for robotic systems are designed, compared, simulated, and implemented. The first algorithm is based on a nonlinear adaptive observer (NLAO), where a sufficient condition for the convergence of the estimator is derived in terms of linear matrix inequality (LMI) under persistence of excitation condition. The second algorithm is based on an adaptive extended Kalman filter (AEKF). Unlike traditional approaches, where the fault parameters are considered as augmented state variables, the AEKF directly estimates the fault parameters from measurement data. The third algorithm is based on a cascade of a nonlinear observer (NLO) and a linearized adaptive Kalman filter (LAKF), called the adaptive exogenous Kalman filter (AXKF). The pros and cons for each algorithm are discussed. The performance of the algorithms is compared in a single-link joint robot system. Furthermore, the algorithms are implemented in a ball-balancing robot to detect and estimate the magnitude of the actuator faults.