نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
The aim of this research is to design an intelligent controller capable of achieving accurate trajectory tracking of a quadrotor in environments with varying wind conditions. To this end, the dynamic modeling of the quadrotor was first carried out. Wind modeling was also implemented by adding horizontal wind acceleration to the translational dynamic equations of the vehicle.
Next, to ensure precise position and attitude control of the quadrotor in the presence of wind disturbances, a hybrid control framework was designed, consisting of a baseline proportional–integral–derivative (PID) controller, a reinforcement learning–based gain tuner, and a disturbance observer along with its compensator. For adaptive tuning of the PID gains during trajectory tracking, the DDPG and TD3 reinforcement learning algorithms were utilized.
Finally, to evaluate the performance of the control framework, various experiments were conducted under no-wind conditions and under different wind intensities across multiple trajectories. The simulation results indicated that in environments with varying wind, adding the observer and compensator to a fixed-gain PID controller reduced the tracking error by 15% compared to the standalone PID controller. Additionally, PID control with reinforcement learning–based gain tuning, combined with the observer and compensator, reduced tracking error by 25% compared to the fixed-gain case. In the lemniscate trajectory, the DDPG algorithm performed 10% better than TD3; in the circular trajectory, the TD3 algorithm performed 5% better than DDPG; and in spiral trajectories, DDPG outperformed TD3 by 20%.
کلیدواژهها English
