Autonomous drones for safe exploration one step closer

Over the past four years, TU Delft and the National Police Lab AI have collaborated on developing autonomous flying drones that can safely navigate unknown environments with limited human feedback. This technology is ideal for exploring hard-to-reach and potentially dangerous locations, such as drug labs.

Drones offer unique advantages in locating missing persons or exploring high-risk environments. “Currently, operating a drone typically requires three operators. We are developing technology that simplifies control, reducing the number of people needed,” explains associate professor Javier Alonso-Mora. “Additionally, autonomous drones lower the risk for police officers by enabling remote site inspections,” adds assistant professor Laura Ferranti.

Making decisions and avoiding obstacles

Alonso-Mora, Ferranti, Prof. Dr. Robert Babuska, and their team have developed advanced navigation and control algorithms allowing drones to explore an unknown space safely. With their techniques, the robot will start autonomously building a map when entering a building, which will be visible to the human operator in real time. Using an onboard camera, the drone looks for cues in the space to locate, for example, a missing person or potentially dangerous materials. It can also leverage human feedback to speed-up exploration based on the operator’s experience.

To make the exploration safe, the researchers developed a dedicated navigation system that accounts for obstacles, such as walls and trees, and hardware limitations. In addition to software development, the team has also worked on the drones’ hardware to ensure they meet the technical requirements and preferences of the police.

Testing is done in the Mobile Robotics Lab.

A productive collaboration

Although the results are promising, the technology is not yet ready for operational use. Currently, for localization purposes, the drones still depend on external cameras, which are not available in natural or unknown indoor environments. “Hopefully, we can start developing a new platform soon to make the drones independent of additional cameras,” says Ferranti.

This project is a joint effort between TU Delft and the National Police Lab AI (NPAI). TU Delft contributed its expertise in motion planning, control, and AI strategies, while the NPAI provided technical knowledge and challenging test scenarios.