Perception & AI for Defence-Related Drone Applications

Project Description

This project is developing smarter and more reliable artificial intelligence (AI) for defense-related drones, enabling them to operate autonomously in complex and hostile environments where resources like GPS may be unavailable. Current drone systems often handle navigation (“where am I?”) and object detection (“what’s around me?”) as separate, inefficient tasks. Our research explores a unified approach, leveraging powerful, large-scale AI vision models to create a single system where understanding the drone’s location improves its ability to spot objects, and vice versa. A central challenge is adapting these large AI systems to run on small, battery-powered drone hardware, so we are developing techniques to make them more lightweight and efficient without sacrificing performance. The ultimate goal is to advance drone autonomy for challenging defense scenarios, such as navigating without GPS and detecting deliberately concealed threats, with potential dual-use benefits for civilian applications like search and rescue, environmental monitoring, and infrastructure inspection.

Funding

This project is funded by Innovation Fund Denmark (IFD) under grant ID 4354-00002B, and by Danish National Defence Technology Centre (NFC) under grant ID 24032.

Contact

PhD Fellow: Peter Guld Leth
Mail: pgl@create.aau.dk

Supervisor: Andreas Møgelmose
Mail: anmo@create.aau.dk

Co-supervisor: Thomas B. Moeslund
Mail: tbm@create.aau.dk