Reinforcement Learning-Based Attitude Control of an Unmanned Aerial Vehicles (UAV): a Comparative Analysis with Sliding Mode Control

Abstract

This article presents a detailed study addressing the challenges of maintaining the stability of Unmanned Aerial Vehicles (UAVs). The complexity of this research domain arises from the intricate nature of UAV control and the inherent unpredictability of disturbances encountered during operation. In response, this study introduces a method for UAV control centered on artificial intelligence to regulate the attitude of a quadcopter drone under real-world flight conditions. While traditional control methods (such as predictive and adaptive controls) have yielded satisfactory results under various flight conditions, they are often based on models involving multiple approximations and fail to account for real-world disturbances. Reinforcement learning, by contrast, is a powerful approach that enables an agent to learn from mathematical models and adapt its behavior to real-world environments.