SARSA $(\lambda)$ Reinforcement Learning Based Path Planning of Unmanned Aerial Vehicles

This paper studies the path planning problem of space Unmanned aerial vehicle model based on SARSA
$(\lambda)$
Reinforcement learning method in complex environment. First, we utilize three-dimensional mathematical modeling to describe the positions of drone at each moment and motion trajectories, to facilitate subsequent analysis of state and action spaces. Then, by introducing the concept of eligibility traces into the traditional SARSA reinforcement learning algorithm, we optimize the path planning problem of drones in a three-dimensional environment. Finally, through simulation, it is demonstrated that this method not only accelerates the learning process but also smoothly updates states, thereby enhancing the exploration efficiency of drone path planning in space compared to traditional path planning algorithms.