Hard Act to Follow: A Practical Leader-Follower System using Curvilinear Formation Motion Plans

This paper presents a comprehensive leader-follower system designed to address the multifaceted challenges of real-world human-robot following scenarios, emphasizing defense applications. Our approach integrates perception, motion planning, and control into a unified framework. By embracing a curvilinear planning domain and treating the leader-follower problem as a teach-and-repeat-style path-following task, we facilitate high-quality formation control for large-scale robotic vehicles. Extensive field trials, conducted across 2 km of driving in an unstructured urban environment, underscore the system's reliability and adaptability across a spectrum of formation types, even in the presence of difficult leader path singularities.