Sensor fusion for impedance control of robot manipulators

Extensive work has been carried out for improving the robot computed torque control method using different types of parameter identification methods, modelling the unknown dynamics and developing more robust controllers. Analytical computed torque has been considered often prohibitive for real time controller implementation. Operational space impedance control proved to be an efficient way to achieve trajectory generation and obstacle avoidance. The implementation of the impedance controller requires exact linearization in operational space which is achieved by the computed torque approach. One of the limits of computed torque schemes is due to the presence of unknown dynamics. In order to account for unknown and unmodelled dynamics terms, a perturbation observer has been added to the impedance controller to improve its robustness end performance. In the paper a perturbation observer has been used in one case to obtain the complete linearization torque and in another case to obtain only the compensating torque for unmodelled dynamic terms. Operational space control was verified experimentally using outputs from joint resolvers and link accelerometers for a data fusion scheme using In extended Kalman filter. The solutions proposed were tested experimentally on a manipulator and the results for the Cartesian control of the manipulator are presented