Two-Loop Online Data-Driven Control for Oral Implantation Robots in Human–Robot Interaction

For oral implantation robots, precise, and safe implantation demands the prediction of human intent and implant position. In this article, a two-loop online data-driven control is proposed for oral implantation robots in human–robot interaction (HRI). First, a feedforward variable-weight outer loop is designed to predict human movement. It relies solely on online moving window of historical movement and tracking errors, thus avoiding the offline human intent training or tuning phase. Second, the inner loop employs a fixed-step model-free adaptive sliding mode controller to predict the unknown implant position, designing the reaching law such that the tracking error converges within a fixed step and is independent of the initial error, thereby ensuring the convergence speed of the tracking error without any model information. In addition, two metrics are designed to evaluate the predictive performance of human movement and implant position estimations, serving as criteria for parameter tuning. This approach is validated through trajectory tracking experiments with two robots: one simulating human movement using publicly movement datasets, and the other tracking target placement position.