\n<\/p>\n
\t\t\t\t\t\t08\/07\/2025<\/span> An autonomous drone has beaten human drone racing champions in an international competition for the first time. A team of scientists from Delft University of Technology has taken first place at the A2RL Drone Championship in Abu Dhabi \u2013 and the AI taking their drone to victory got off to a flying start in the lab of the European Space Agency\u2019s Advanced Concepts Team just a few years prior.<\/p>\n<\/div>\n Earlier this year, a team from the Micro Air Vehicle Laboratory, MAVLab, of Delft University of Technology, TU Delft, competed against 13 autonomous drones and three drone racing champions at the A2RL Drone Championship in Abu Dhabi \u2013 beating them all in a historic racing first.<\/p>\n The victorious team\u2019s drone was able to dominate the competition thanks to neural network-based AI control systems, originally developed by ESA\u2019s Advanced Concepts Team (ACT) \u2013 the agency\u2019s multidisciplinary think tank focusing on emerging technologies.<\/p>\n ACT\u2019s scientific coordinator Dario Izzo explains how the collaboration started: \u201cA few years ago, the ACT started working on neural networks for spacecraft guidance.<\/p>\n \u201cTypically, in space settings, Guidance and Control are two separate systems. The Guidance part, involving different manoeuvres, is traditionally planned in detail by engineers on the ground, while the Control part is undertaken by the spacecraft itself.<\/p>\n \u201cTo make the system more efficient, we explored a way to combine the two in one, creating an end-to-end controller. This alternative method, called Guidance & Control Networks \u2013 G&CNets \u2013 involves all the work taking place on the spacecraft.<\/p>\n \u201cBefore using this method to fly real spacecraft, we needed a more accessible testing platform \u2013 this is when our collaboration with TU Delft\u2019s MAVLab started. Our concept tested on MAVLab\u2019s drones immediately delivered incredible results.\u201d<\/p>\n<\/p><\/div>\n The G&CNets approach employs neural networks \u2013 AI models composed of interlinked artificial neurons, mimicking the architecture of animal brains.<\/p>\n Robin Ferede, PhD candidate at TU Delft and member of the winning race team, describes the journey from the first flight tests to the race victory: \u201cThe G&CNets were initially trained using the \u2018behavioural cloning\u2019 method, which is based on giving the AI expert examples to learn from.<\/p>\n \u201cThe method uses a mathematically optimal solution to control a system. When replicated in a simulation, these solutions work perfectly. But flying a real drone using this system is quickly hindered by the \u2018reality gap\u2019 \u2013 what worked in simulation causes the drone to crash.<\/p>\n \u201cLooking at data from these unsuccessful test flights, we realised this system does not replicate the true dynamics of the drone \u2013 and no matter how much we improved the model, this reality gap always remained.\u201d<\/p>\n To close this gap, the team turned to reinforcement learning, a method of training AI by means of trial and error.<\/p>\n \u201cEven though dynamics are somewhat easier to predict for spacecraft then for drones, the reality gap is an issue in both cases,\u201d says Sebastien Origer, ACT\u2019s former Graduate Trainee. \u201cReinforcement learning addresses this by training the network to deal with unforeseen circumstances as they arise during flight.<\/p>\n<\/p><\/div>\n \u201cIn space there is no air, so aerodynamic effects are not an issue. Even then, other factors, such as gravitational variations, can influence the control systems of a spacecraft.<\/p>\n \u201cTo gain confidence that the G&CNets will work in the space environment, it was important to demonstrate that they can be applied to any real platform, whether that be a drone, robot, or a spacecraft.\u201d<\/p>\n Guido de Croon, TU Delft professor and the race team\u2019s supervisor, comments: \u201cThe switch from behavioural cloning to reinforcement learning allowed us to significantly improve the efficiency and robustness of the G&CNets.<\/p>\n \u201cTaking part in the drone race championship made us push the limits of the technology even further \u2013 everything, from the drones to the track and racing conditions, was controlled by the organisers and completely unknown to us. All these aspects made for an amazing challenge for our software as well as the team.\u201d<\/p>\n \u201cWe had to use the time we were given wisely,\u201d says Stavrow Bahnam, PhD candidate at TU Delft and member of the team. \u201cWhen our drone crashed, we couldn\u2019t spend as much time as we would have liked on fixing and recalibrating it, we simply had to reuse it \u2013 this was another challenge that our AI dealt with perfectly.\u201d<\/p>\n<\/p><\/div>\n Robin adds: \u201cIn the final tournament, competing against the world champion, we knew we had to step up our game. In between the individual runs, we finetuned the network quite literally on the fly. These last minute adjustments paid off, and our drone crossed the finish line first.\u201d<\/p>\n Guido concludes: \u201cMuch as the race win was a symbolic culmination of our efforts, we certainly do not stop there. As the next step, we are exploring the use of technologies that would allow our network to fit on even smaller, more energy-efficient chips \u2013 a topic on which we again collaborate with the Advanced Concepts Team.\u201d<\/p>\n \u201cThis achievement demonstrates how blue-sky research from ESA\u2019s Advanced Concepts Team can lead to transformative applications \u2013 both in space and here on Earth,\u201d Dario concludes. \u201cFrom autonomous spacecraft to racing drones, it\u2019s a powerful example of how fundamental research can cross domains and push the boundaries of what we can achieve.\u201d<\/p>\n<\/p><\/div>\n
\n\t\t\t\t8<\/span> views<\/small><\/span>
\n\t\t\t\t\t\t\t\t\t\t0<\/span> likes<\/small><\/span><\/p>\n<\/header>\n
\n\t\t\t\t\t\t\t\t<\/figcaption><\/figure>\n
\n\t\t\t\t\t\t\t\t<\/figcaption><\/figure>\n
\n\t\t\t\t\t\t\t\t<\/figcaption><\/figure>\n
\n\t\t\t\t\t\t\t\t<\/figcaption><\/figure>\n