Autonomous AI Drone Tested in Race in Abu Dhabi

Photo Credits: TU Delft

In a drone race conducted in the city of Abu Dhabi, a drone piloted by an artificial intelligence achieved its first victory in an international competition. The AI controlled drone that won the race was developed by Delft University of Technology’s MAVLab. 

The concept of AI controlling drones can be expected to contribute significantly to many sectors. For example, faster drones could be used to deliver blood and defibrillators more urgently or to search disaster zones for survivors more quickly. 

Over the course of two days MAVLab’s AI piloted drone competed against 13 other international teams. The championship was organised by the Abu Dhabi Autonomous Racing League (A2RL) and Drone Champions League (DCL). 

The race was conducted with each of the teams having a  standardised drone with a NVIDIA Jetson Orin NX computing module, an inertial measurement unit for onboard sensing and control, and a single forward-facing camera. The drones raced at speeds as high as 95 miles per hour.  

The championship aimed to boost the research in AI autonomously controlling drones and performing under the pressure of time with very limited computing and sensing resources.  The use of just one camera was a very intentional choice, as most human pilots have a first person view when flying a drone. A single camera attempted to simulate the same for the AI and present visual and perception challenges while flying. 

The 11-gate racetrack was also designed to push the boundaries of testing the ability of AI to fly the drones, featuring wide gate spacing, irregular lighting and minimal visual markers. 

Team lead Christophe De Wagter explained that at the core of TU Delft’s drone is a deep neural network that sends control commands directly to the drone motors. A similar concept of network was originally created by the European Space Agency. He also added that AI-powered drone racing helps develop robust and dynamic AI systems. 

As the capabilities of AI to pilot drones autonomously advance, they can be used in situations where emergency delivery of medical equipment and supplies is needed in disaster-affected zones where rescue personnel cannot go as well as in numerous other situations where vital supplies can be delivered in very hazardous conditions.