Reinforcement Learning and PD Control Based Trajectory Tracking for a Quadcopter UAV

Abstract

Nowadays, quadcopter unmanned aerial vehicles (UAV) are used in a wide variety of areas, such as reconnaissance and surveillance, firefighting, search and rescue, agricultural spraying, cargo transportation, photography, and mapping. The use of quadcopters in a very wide area makes their trajectory tracking control important. In order for quadcopters to perform their duties successfully, they must be able to follow the given trajectory with the least error. In this study, the quadcopter’s trajectory tracking under random noise is provided by an algorithm based on reinforcement learning and a proportional derivative (PD) controller. Modeling, simulations, and reinforcement learning algorithms were carried out using the MATLAB program. Simulations were made under noise for the x, y, z trajectories and roll, pitch, and yaw angles of the quadcopter. A detailed time response analysis was performed by obtaining rise time, overshoot, and settling time data. It has been observed that the references given were successfully followed thanks to the algorithm based on reinforcement learning.