Title: Enhancing drone performance: using reinforcement learning for active pan and tilt rotor fault tolerance strategy

Authors: Zairil Zaludin

Addresses: Department of Aerospace Engineering, Faculty of Engineering, Universiti Putra Malaysia, Serdang 43400 Selangor, Malaysia

Abstract: Quadcopter drones rely on their four rotors to maintain control of altitude and attitude. However, if one rotor fails, the drones ability to stay airborne is compromised. This article proposes a solution that improves attitude control during a single rotor failure by minimising roll, pitch, and yaw deviations. This was done by actively panning and tilting one of the functioning motors. The controller for the pan and tilt was designed using the reinforcement learning method. The solution was obtained after the agent accumulated as many reward points as possible through 5,000 training episodes. The study demonstrated the feasibility of improving the uncontrollable rolling, pitching, and yawing usually seen in a drone after a single rotor total failure during flight.

Keywords: unmanned aerial vehicles; UAV; fault tolerance; quadrotor; quadcopter; quadcopter malfunction; drone; reinforcement learning; RL; DDPG agent; drone rotor pan-tilt.

DOI: 10.1504/IJAMECHS.2025.145731

International Journal of Advanced Mechatronic Systems, 2025 Vol.12 No.2, pp.132 - 146

Received: 24 Sep 2024
Accepted: 16 Jan 2025
Published online: 17 Apr 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article