Drone racing pilots are capable of some impressive control and trickery, but autonomous racing drones are also looking to claim their share of the spotlight. Scientists at TU Delft have now developed what is claimed to be the smallest autonomous racing drone on the planet, a feat that involved some serious innovations in the algorithms that control its flight path.

The US$2.25 million Artificial Intelligence Robotic Racing contest by the Drone Racing League is one example of the spoils on offer for technologists developing racing drones that fly on their own. That particular competition offers $1 million for the winning team of a head-to-head AI racing circuit, and $250,000 for the first team to beat a human-piloted drone across the finish line. Point being, autonomous racing drones are serious business, and as the technology moves forward it may well have flow-on effects for other areas of the drone industry.

"For typical drones with four rotors, flying faster also simply means that they are able to cover more area," says Guido de Croon, scientific leader at TU Delft's Micro Air Vehicle Laboratory. "For some applications, such as search and rescue or package delivery, being quicker will be hugely beneficial. Our focus on light weight and cheap solutions means that such fast flight capabilities will be available to a large variety of drones."

De Croon and his team set out to see how exactly small a package they could fit a competitive racing drone into. The researchers say that the fastest autonomous racing drones can travel at around two meters per second (6.5 ft), but these require high-performance processors and cameras to model their flightpath, which in turn makes them heavy and expensive.

The aircraft they wound up with weighs just 72 g (2.5 oz) and is 10 cm (3.9 in) in diameter, constituting the smallest autonomous racing drone in the world. With only a single camera and minimal sensors, the drone instead relies heavily on efficient algorithms that help it predict a safe path through a course with minimal information.

"When scaling down the drone and sensors, the sensor measurements deteriorate in quality, from the camera to the accelerometers", explains Shuo Li, PhD student at the MAVLab on the topic of autonomous drone racing. "Hence, the typical approach of integrating the accelerations measured by the accelerometers is hopeless. Instead, we have only used the estimated drone attitude in our predictive model. We correct the drift of this model over time by relying on the vision measurements."

The team put its drone to work on an indoor racing course at TU Delft's Cyberzoo, consisting of four gates. The drone was able to navigate the course and travel at an average speed of two meters per second, on par with the performance of larger autonomous racing drones, though the team hopes this is just the beginning, and that its advances can then be applied to other applications for unmanned aircraft.

"We are currently still far from the speeds obtained by expert human drone racers," says Christophe De Wagter, founder of the MAVLab. "The next step will require even better predictive control, state estimation and computer vision. Efficient algorithms to achieve these capabilities will be essential, as they will allow the drone to sense and react quickly. Moreover, small drones can choose their trajectory more freely, as the racing gates are relatively larger for them."

You can see the drone in action, in the video below.

Source: TU Delft

View gallery - 4 images