Computationally intelligent strategies for robust fault detection, isolation, and identification of mobile robots

In this paper, new fault detection and isolation/identification (FDI) schemes are proposed by using adaptive threshold bands that are generated with locally linear models (LLM) as well as model error modeling (MEM) techniques. The performance capabilities of our two proposed adaptive threshold bands are compared relative to each other as well as with the performance of a fixed threshold bands. To demonstrate and illustrate the capabilities of our proposed FDI methodology, the developed techniques are applied to a high fidelity model of a two wheeled mobile robot that is subject to the most physically possible faults in these systems. The mobile robot is modeled implicitly by utilizing two computationally intelligent methodologies. Specifically, locally linear models (LLM) as a neuro-fuzzy technique and a radial basis function as a neural network are used to identify and represent the model of the mobile robot. The resulting improvements in the FDI performance by employing our proposed adaptive threshold bands are demonstrated and illustrated through extensive simulation case studies.