A Human-Aided Learning Approach to Optimal Control of Human-Robot Systems

Human-Robot Systems (HRS) play a significant role in many applications, including manufacturing, critical infrastructure, and civil applications. However, conventional HRSs require predetermined reference before the start of the mission, which restricts the flexibility of the HRSs. In this paper, we use a recursive least-square approach to design Human-Aided Learning (HAL) algorithm, which allows the robot to learn the reference via the human inputs and system outputs. With multiple learning iterations, the robot can determine the reference. After estimating the reference, the robot can execute the tasks independently. We also design a low-pass filter to remove the impact of a high-frequency noise introduced by the human inputs. Finally, we use simulation to evaluate the performance of the proposed HAL algorithm.