Topic

題目

Reinforcement Learning Based Agile Flight Control of Quadrotor Drones 

基於強化學習的四旋翼無人機敏捷飛行控

Abstract

撮要

Quadrotors, with their advantages such as rapid vertical takeoff and landing, wide operational range, low operating costs, high flight accuracy, and strong agility and flexibility, have been widely applied in industries, agriculture, and other fields. Nevertheless, as the application fields expand and the complexity of application scenarios increases, higher requirements and new challenges are required for the autonomous flight control capabilities of quadrotors. These include the agility and accuracy required for autonomous traversal in narrow areas, the adaptive ability to handle model uncertainties and external environmental disturbances, and the real-time response capability for autonomous planning and decision-making. This report will focus on small quadrotors, centering around the underlying control of drones, stable hovering, trajectory tracking, traversal flight in narrow areas, and safe flight under rotorcraft malfunctions. It will particularly introduce the self-learning motion control technology based on reinforcement learning for rotorcraft drones, aiming to develop a relatively comprehensive set of self-learning control methods for rotorcraft drones.