Adaptive Neural Control for Safe Human-Robot Interaction

Abstract

This thesis studies safe human-robot interaction utilizing the neural adaptive control design. First, novel tangent and secant barrier Lyapunov functions are constructed to provide stable position and velocity constrained controls, respectively. Then, neural backpropagation and the concept of the inverse differential Riccati equation are utilized to achieve the impedance adaption control for assistive human-robot interaction, and the optimal robot-environment interaction control, respectively. Finally, adaptive neural assist-as-needed control is developed for assistive robotic rehabilitation.