Learning a Controller Fusion Network by Online Trajectory Filtering for Vision-based UAV Racing
This work addresses the challenge of robust control in fast-paced UAV racing, offering an incremental improvement over existing methods by combining and filtering multiple controllers for better adaptability.
The paper tackles the problem of autonomous UAV racing by proposing a DNN that fuses multiple controllers with online trajectory filtering to overcome limitations of end-to-end policies and single controllers, resulting in performance that beats all baselines and approaches that of a professional human pilot in simulation.
Autonomous UAV racing has recently emerged as an interesting research problem. The dream is to beat humans in this new fast-paced sport. A common approach is to learn an end-to-end policy that directly predicts controls from raw images by imitating an expert. However, such a policy is limited by the expert it imitates and scaling to other environments and vehicle dynamics is difficult. One approach to overcome the drawbacks of an end-to-end policy is to train a network only on the perception task and handle control with a PID or MPC controller. However, a single controller must be extensively tuned and cannot usually cover the whole state space. In this paper, we propose learning an optimized controller using a DNN that fuses multiple controllers. The network learns a robust controller with online trajectory filtering, which suppresses noisy trajectories and imperfections of individual controllers. The result is a network that is able to learn a good fusion of filtered trajectories from different controllers leading to significant improvements in overall performance. We compare our trained network to controllers it has learned from, end-to-end baselines and human pilots in a realistic simulation; our network beats all baselines in extensive experiments and approaches the performance of a professional human pilot. A video summarizing this work is available at https://youtu.be/hGKlE5X9Z5U