Sliding mode fault diagnosis with vision in the loop for robot manipulators

This chapter is devoted to the problem of Fault Diagnosis (FD) for industrial robotic manipulators within the framework of sliding mode control theory. According to this control concept, a set of unknown input higher order sliding mode observers are designed to detect, isolate and identify multiple actuators faults and corruptions. More specifically, the whole FD architecture is based on the inverse dynamics-based feedback linearized robotic MIMO system, which is equivalent to a set of linearized decoupled SISO systems, affected by uncertain terms. The FD process includes a residual generation, followed by a decision making through the evaluation of the achieved residuals. The advantages of the sliding mode approach are the good performance in terms of stability and robustness, as well as satisfactory estimate of the occurring faults. Furthermore, in order to extend the FD strategy to multiple sensor and actuator faults, a low cost vision servoing architecture is used in the scheme, allowing one to design a fault tolerant control strategy in case of sensor faults. The effectiveness of the proposed FD architecture has been carried out in simulation on a realistic simulator as well as experimentally on a COMAU SMART3-S2 anthropomorphic robot manipulator.