This work investigates adaptive switching mechanisms for reinforcement learning (RL) controllers in high-speed autonomous drone racing. While domain randomization (DR) improves generalization, single-policy controllers remain constrained by their training distributions and may fail under unmodeled conditions such as wind, hardware wear, or sensor degradation. To address this, we introduce two complementary online switching strategies: the look-back switch, which retrospectively identifies the dynamics model that best matches the current flight state, and the look-ahead switch, which prospectively selects the most suitable policy by simulating candidate controllers over short horizons. Both mechanisms leverage pre-trained policies without requiring retraining, enabling safer and faster flight. We evaluate the methods in simulation on two platforms (5-inch racing quadcopter and the high-fidelity A2RL competition drone) and across two tracks (Figure-8 and A2RL Grand Challenge). Results show that adaptive switching improves robustness and reduces lap times compared to fixed-policy baselines, demonstrating its potential for real-world deployment.

Reaching fast and autonomous flight requires computationally efficient and robust algorithms. To this end, we train Guidance & Control Networks to approximate optimal control policies ranging from energy-optimal to time-optimal flight. We show that the policies become more difficult to learn the closer we get to the time-optimal ’bang-bang’ control profile. We also assess the importance of knowing the maximum angular rotor velocity of the quadcopter and show that over- or underestimating this limit leads to less robust flight. We propose an algorithm to identify the current maximum angular rotor velocity onboard and a network that adapts its policy based on the identified limit. Finally, we extend previous work on Guidance & Control Networks by learning to take consecutive waypoints into account. We fly a 4×3m track in similar lap times as the differential-flatness-based minimum snap benchmark controller while benefiting from the flexibility that Guidance & Control Networks offer.

In this study, we present a first step towards a cutting-edge software framework that will enable autonomous racing capabilities for nano drones. Through the integration of neural networks tailored for real-time operation on resource-constrained devices. A lightweight Convolutional Neural Network, with the Gatenet architecture, is adjusted for reduced computational demand and is successfully deployed on a GAP8 processor at a rate of 16$Hz$. This network provides gates' size and location data for the subsequent positioning algorithm. A second neural network, trained through reinforcement learning, governs the drone's guidance and control systems, demonstrating a remarkable rate of 167$Hz$ on an STM32F405 processor. The attitude rates and thrust outputted by this network are then fed to an attitude rate PID controller.

The research shows that state-of-the-art neural networks for drone racing can be deployed on nano drones, despite their limited processing power. Nonetheless, the study demonstrated specific limitations, such as the perception network's sensitivity to white pixels in the image reducing its effectiveness when light sources are present in the scene. These findings underscore the importance of dataset composition and the need for diverse training scenarios to enhance the neural network's generalizability and performance in real-world applications.