Game-Theoretic Interaction Control for Assistive Exoskeletons: a 2-DOF Simulation Study

Enhancing interaction control remains a key challenge for robotic devices designed to assist human movement. Interaction control can be modeled as a dyadic differential game, where both the human and the robot aim to minimize their respective cost functions over a finite time horizon. The difficulty lies in continuously estimating the human cost parameters to design an optimal game-theoretic robot controller. This simulation study evaluates the effectiveness and robustness of a bi-level optimization method in accurately recovering human cost parameters during a trajectory tracking task with a virtual 2-DOF exoskeleton.