Adaptive Task-Space Force Control for Humanoid-to-Human Assistance

We envision a humanoid robot to serve as source of additional motion-support forces in assistance for frail persons. In this context, we present control strategy adaptively regulate its assistive force contribution. First, identify human model torque optimal execution priori known motion task from sample recordings performed by healthy individual. utilize the identified proposed position discrepancy based observer contribution, unknown and unmeasurable variable. propose an experience-based contribution learning that allows improve estimate trial-to-trial. The target is then calculated difference between required estimated integrated into multi-robot quadratic programming task-space controller compute desired interaction supply necessary model. use non-optimal emulate frailty apply our adaptive demonstrate results successfully assisting simulated restore performance.