Drones and ground robots are learning to think, not just obey.

Air-ground collaborative systems pair flying drones with ground vehicles. Drones scan the area from the sky while ground vehicles travel across the terrain below. Sharing what they see lets them achieve far more than any single machine can alone. In the past, these teams followed strict rules or basic reinforcement learning programs. Next, systems added language interfaces through vision-language navigation. Now, researchers show these systems are shifting toward multimodal large language models, or advanced AI that handles both images and text. These models let aerial and ground machines perceive surroundings, reason, and coordinate actions on their own. Instead of just replacing old software tools, this shift fundamentally changes how machines understand tasks, coordinate different types of robots, and make autonomous decisions in unknown environments.

Why it matters

This shift means air-ground teams can handle unfamiliar situations without being pre-programmed for every scenario. Instead of needing engineers to script every behavior, the system interprets instructions and adapts on the fly. For real-world applications like disaster response or environmental monitoring, that makes collaboration faster and more flexible.

Questions answered