Robust Fault Detection and Isolation for Proprioceptive Sensors of Robot Manipulators

In this paper a scheme for detecting and isolating proprioceptive sensor faults in industrial robot manipulators is devised. To the purpose, an analytical redundancy approach has been pursued, based on a bank of state observers for residual generation. Namely, an extended H_inf approach is adopted and the compensation of poorly known dynamics in each observer is improved by the use of a Radial Basis Functions (RBFs) neural network. The design of the observer matrix gain is achieved by solving a Linear Matrix Inequality (LMI) feasibility problem, where constraints on the position in the complex plane of the poles of the estimation error dynamics are taken into account. Finally, in order to test the effectiveness of the proposed approach, a case study is developed, based on experiments performed on a six-degree-of-freedom Comau Smart-3 S industrial manipulator.