8. Intelligent Human–Robot Interaction Systems Using Reinforcement Learning and Neural Networks

In this chapter, an intelligent human–robot system with adjustable robot autonomy is presented to assist the human operator to perform a given task with minimum workload demands and optimal performance. The proposed control methodology consists of two feedback loops: an inner loop that makes the robot with unknown dynamics behave like a prescribed impedance model as perceived by the operator, and an outer loop that finds the optimal parameters of this model to adjust the robot’s dynamics to the operator skills and minimize the tracking error. A nonlinear robust controller using neural networks is used in the inner loop to make the nonlinear unknown robot dynamics behave like a prescribed impedance model. The problem of finding the optimal parameters of the prescribed impedance model is formulated as an optimal control problem in the outer loop. The objective is to minimize the human effort and optimize the closed-loop behavior of the human–machine system for a given task. This design must take into account the unknown human dynamics as well as the desired overall performance of the human–robot system, which depends on the task. To obviate the requirement of the knowledge of the human model, reinforcement learning is used to learn the solution to the given optimal control problem online in real time.