A Confidence-based Allocation Approach for USV Trajectory Tracking via Human-Robot Co-Driving

Abstract

Unmanned surface vehicles (USVs) equipped with autonomous controllers have emerged as a viable alternative to human drivers in complex and unstructured task environments. However, challenges remain in terms of social trust and flexibility when handling complex multitasking scenarios, which hinder the widespread adoption of unmanned operations. To address this, a Human-Driving Co-driving framework to achieve the human-in-the loop control is proposed in our work. A nonlinear Model Predictive Control (MPC) framework with nominal dynamic model will be solved for optimal tracking accuracy. And a confidence-based dynamic authority allocation method considering temporal stochastic driver model enable the flexible and efficient shared control. The effectiveness of our proposed approach is validated through comprehensive simulation experiments.