Robust Fault Diagnosis and Accommodation for a Robotic Manipulator under Human Interaction

This paper addresses the problem of fault diagnosis in a robotic manipulator designed for human interaction. Faults in the external force sensor are addressed with a constant or slowly time-varying profile. A model-based strategy is considered that takes into account uncertainty in the mathematical representation, where a proportional-integral (PI) observer is in charge of estimating the fault profile. This estimation stage incorporates tools from adaptive estimation theory in order to employ in the design stage the uncertainty information, and a compact set of variation for the fault profile. The robotic manipulator includes a nominal control algorithm based on impedance philosophy to address the human-robot interaction. The nominal controller embeds a fault accommodation structure to add fault-tolerance to the closed-loop response. The effectiveness of the proposed approach was verified in simulation under model uncertainty and human-robot interaction, where the results clearly illustrate an accurate fault diagnosis stage and an improved closed-loop response after fault accommodation.