Towards Aerial Embodied Intelligence: Current Status and Challenges

Abstract

Unmanned Aerial Vehicle (UAV) technology is undergoing a paradigm shift from sensing platforms to Aerial Embodied Intelligence (AEI) systems capable of physical manipulation. This paper provides a structured review of this emerging field, clarifying its definition, core tasks, and current technological bottlenecks. Specifically addressing the challenges of contact-based manipulation, we analyze the dynamic coupling between aerial platforms and aerial manipulators and discuss interaction stability in unstructured environments. The implementation path of such tasks is further elucidated through a case study on UAV-based bulb replacement. Subsequently, the paper outlines future research directions, highlighting highly stable aerial platforms, dexterous manipulation, learning algorithms for intelligent adhesion, data-driven skill generalization, and the optimization of Human-Robot Interaction and multi-robot coordination as key priorities. Finally, the necessity of high-fidelity open simulation platforms is emphasized for accelerating the verification and deployment of AEI algorithms.