Motion control of robot manipulators under sensor failure

In a robot control system there are dynamics and sensor faults, modeling errors, and system and measurement noises. This paper addresses several techniques to control the robot manipulator under the uncertainty model and sensor failure. First, a neural network is proposed to control the robot to follow the generated path. This network model is able to compensate the structured and unstructured dynamic uncertainties of the robot by using both online and off-line training. The PD controller, computed torque controller and adaptive controller with and without exact model are used to control the robot and compared with the proposed method. Next, a residual generator for detection and isolation of the sensor failure is designed. Assuming one fault at a time, the proposed method can detect and isolate the fault sensor. The simulation study based on a 2-DOF planar robot is included