Title: Managing a Twin Rotor System via Twin Delayed Deep Deterministic Policy Gradient (TD3)

Abstract: This study introduces a reinforcement learning (RL) approach designed to stabilize the Twin Rotor Aerodynamic System (TRAS) at designated pitch and azimuth positions, as well as to facilitate trajectory tracking. Controlling the TRAS is notoriously difficult due to its complex, non-linear dynamics, which often render traditional control algorithms ineffective. However, the potential of RL in managing multirotor systems has garnered significant attention following recent advancements. In this work, the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm serves as the foundation for training the RL agent. TD3 is particularly suited for environments featuring continuous state and action spaces, such as the TRAS, because it operates without requiring an explicit model of the system. Simulation data demonstrates the efficacy of this RL-based control strategy. Furthermore, the controller’s performance was evaluated against standard PID controllers under wind-induced external disturbances. Finally, laboratory experiments were conducted to validate the controller’s practical applicability in real-world scenarios.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC