Fault detection and isolation for robotic systems using a multilayer perceptron and a radial basis function network

Usually, fault detection and isolation schemes for robotic manipulators use the system mathematical model to generate the residual vector. However, modeling errors could obscure the faults and could be a false alarm source. In this paper a multilayer perceptron trained with backpropagation algorithm is employed to reproduce the robot input/output behavior generating the residual vector. Then, a radial basis function network is utilized to classify the residual vector generating the fault isolation. Three different algorithms have been employed to train this network. The first employs subset selection to choose the radial units from the training patterns. The second utilizes regularization to reduce the variance of the model. The third algorithm also uses regularization but, instead of one penalty term, each radial unit has an individual penalty term. Simulations employing a two-link manipulator are showed demonstrating that the system can detect and isolate correctly faults that occur in nontrained trajectories.