Learning-Based Robust Control Policy Design for Vehicular Navigation via Ultra-Wideband Communication

In this paper, we investigate a learning-based framework that addresses the point-to-point vehicular navigation problem through a robust output regulation approach and test it through Ultra-Wide Band (UWB) communication noise data. Our focus is on the learning-based control policy design and ensuring the stability of the closed-loop system for vehicle dynamics. It is seen that our proposed solution, which relies on reinforcement learning, is data-driven and does not require accurate knowledge of the vehicle motion model. Specifically, we explore learning-based optimal tracking control through state feedback and output feedback, which we categorize as policy iteration and value iteration, respectively. To validate our approach, we incorporated noise samples from a real-world UWB positioning platform into our simulation. Our findings demonstrate that the learning-based method may be effective in both vehicular point-to-point tracking and interference tolerance, even in the presence of real-world noise interference.