Designing Deep Neural Net Controller for Quadrotor Attitude Stabilization

Xudong He

Abstract: Quadrotors are popular small unmanned aerial vehicles (UAVs) deployed in many real-world applications such as aerial surveying. Despite their simple mechanical construction and propulsion principle, quadrotors have nonlinear dynamics and require advanced stabilizing control. This paper presents an approach combining deep learning, control theory, and convex optimization to design neural net controllers to stabilize quadrotors, which provides the basis for developing neural net controllers to stabilize more sophisticated drones.

External IDs:dblp:conf/qrs/He24