Talk title: Iterative learning of robot impedance and reference position: from repetitive to non-repetitive tasks.
Abstract: Adaptation of a robot’s impedance and reference is an effective approach to achieve safe, optimal, and desired interaction between the robot and an environment, or a human. For a repetitive interaction, estimation of the environment’s parameters can be improved based on previous iterations, and the robot’s control parameters, i.e., impedance and reference position, can be updated in accordance. In this talk, I will introduce how we applied the idea of iterative learning control in control theory to interaction control, starting with repetitive tasks involving periodic movements. I will then explain the challenges faced by the traditional method, leading us to development of a new learning method that can handle non-repetitive tasks involving random movements. Case studies of this approach in applications of robotic tooling, exoskeleton, and collaborative robotics will be presented.