Linear-quadratic-Gaussian torque control: Application to a flexible joint of a rehabilitation exoskeleton

A linear-quadratic-Gaussian regulator is proposed for the torque control of flexible robotic joints with built-in torque sensor. The regulator requires the joint-torque sensor information only and features: (1) a Kalman filter that, beside reducing the noise and evaluating the derivative of the torque sensor measure, is able to estimate all the external and internal torques acting on the joint; 2) a controller that optimizes system stability, responsiveness, accuracy and effort. The regulator is implemented on a flexible joint of a rehabilitation exoskeleton. Simulation and experimental results are provided which demonstrate regulator performance and efficacy.