Highly Versatile Aerial-Terrestrial Robot - PogoX: System Design and Efficient Data-Driven Control

Abstract

The versatility of hybrid aerial–terrestrial robotic platforms continues to highlight their potential for general-purpose mobility across complex and unstructured environments. However, the diverse dynamics in distinct aerial and terrestrial locomotion modes, have placed critical challenges for existing platforms. The first fundemantal challenge is designing elegant and efficient, most existing work try to add extra actuation for ground locomotion, which imposed challenge to aerial mobility. The other is on the control side, most of them struggle to robustly manage contacts, either continuous or impact-driven interactions. In this thesis, we introduce PogoX, a highly integrated hybrid robot capable of agile flying, efficient wheeled skating, and dynamic hopping. To cope with hybrid dynamics and modeling uncertainty, we develop a unified data-driven predictive controller that augments model-based strategies to compensate the modeling error across modes. Finally, We combined the reference steering framework with our design, and experimentally validate the system with agile motion and stable control in each locomotion mode. The results demonstrate that PogoX performs effectively in all three modes, showcasing its potential as a general-purpose hybrid mobility platform.