A Review of the Applications of Vision-Language Models in the Low-Altitude Economy

The Low-Altitude Economy (LAE), a new economic form driven by Unmanned Aerial Vehicles (UAVs) and electric Vertical Takeoff and Landing (eVTOL) aircraft, is a rapidly emerging strategic industry. However, its potential is constrained by the intelligence level of current aerial platforms, whose reliance on predefined routes and rules is inadequate for complex, dynamic environments. Concurrently, Vision-Language Models (VLMs) represent a revolutionary breakthrough in AI, endowing machines with unprecedented capabilities to understand, reason, and interact with the physical world. This paper’s core thesis is that the deep integration of VLMs with the LAE is driving a paradigm shift from simple aerial automation to advanced Aerial Embodied Intelligence. This article first review the evolution of VLMs and analyze the LAE’s core requirements for advanced intelligence. Then elaborate on three key VLM applications in the LAE: natural language-based interactive navigation, dynamic scene understanding in open-world environments, and multi-agent collaboration for swarm intelligence. The paper also addresses critical challenges, including model robustness, real-time edge deployment, and the scarcity of domain-specific data. Finally, it looks toward future research frontiers, highlighting unified Vision-Language-Action (VLA) models and human-in-the-loop learning as key to achieving truly autonomous aerial systems.