World’s Smallest Autonomous Racing Drone

MAVLab TU Delft reports the world’s first smallest autonomous racing drone created by scientists. The Micro Air Vehicle Laboratory (MAVLab) of TU Delft aims is to make light-weight and cheap autonomous racing drones. Such drones could be used by many drone racing enthusiasts to train with or fly against. If the drone becomes small enough, it could even be used for racing at home.

Drone racing by human pilots has become a rage amongst drone enthusiasts and a huge challenge for artificial intelligence and control. For autonomous racing, drones need to use state-of-the-art solutions for visual perception, like building maps of the environment or tracking drone movement over time accurately. This means high-performance processors, with multiple, high-quality cameras and/ or laser scanners equipped drones, making them relatively heavy and expensive.

Credits: Yingfu Xu, TU Delft
Credits: Yingfu Xu, TU Delft
Credits: Yingfu Xu, TU Delft
Credits: Yingfu Xu, TU Delft
Credits: Yingfu Xu, TU Delft

The drone racing team of the MAVLab has developed a drone 10 cm in diameter weighing just 72 grams. It uses only a single camera and very little onboard processing in order to autonomously fly through a racing track with a speed that rivals that of the fastest, bigger autonomous racing drones.

Asserting that the main challenge in achieving this feat is creation of extremely efficient yet robust algorithms, Christophe De Wagter, founder of the MAVLab says, “The wireless images in human drone racing can be very noisy and sometimes not even arrive at all. So, pilots rely heavily on their predictions of how the drone is going to move when they move the sticks on their remote control.”

The interpretation of the images by small drones can sometimes be defective-they might miss a gate or evaluate their position relative to the gate incorrectly. So a prediction model is central to the approach. Since the drone has very little processing, the model only captures the essentials, such as thrust and drag forces on the drone frame.

Stating that a new robust state estimation filter was used to combine the noisy vision with the model predictions measurements in the best way possible, Shuo Li, PhD student at the MAVLab on the topic of autonomous drone racing adds, “When scaling down the drone and sensors, the sensor measurements deteriorate in quality, from the camera to the accelerometers. 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 drone used the newly developed algorithms to race along a 4-gate race track in TU Delft’s Cyberzoo. It flew multiple laps at an average speed of 2 m/s, which is at par larger with state-of-the-art autonomous racing drones.

Acknowledging that racing drones can be useful in multiple cases other than just racing Guido de Croon, scientific leader of the MAVLab says some applications, such as search and rescue or package delivery becoming 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.”

Citation: Shuo Li, Erik van der Horst, Philipp Duernay, Christophe De Wagter, Guido C.H.E. de Croon,“Visual Model-predictive Localization for Computationally Efficient Autonomous Racing of a 72-gram Drone”, ArXiv Preprint arXiv:1905.10110 (2019) Link: https://arxiv.org/abs/1905.10110

Share