The paper discusses the use of imitation learning for controlling quadrotors with multiple waypoint constraints. The authors propose a neural network called WN&CNets (waypoint-constrained navigation and control network) to generate control commands online. This approach enables quadrotors to replan in real-time, handle dynamic waypoints, and manage deviations caused by disturbances or model inaccuracies. The method is tested in simulations and real-world experiments, achieving a maximum speed of 7m/s while navigating through 7 waypoints in a confined space.
Publication date: 20 Feb 2024
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2402.11570
