Original scientific paper

https://doi.org/10.17559/TV-20250211002362

Implementation of Unmanned Aerial Vehicle Flight Path Planning Software Based on Integral Compensation Reinforcement Learning Algorithm

Xianlong Ma ; School of Aerospace, Northwestern Polytechnical University, Xi'an, Shaanxi province 710072, China *

* Corresponding author.

Abstract

Due to the complex nature of UAVs, flight path planning is a very difficult task. This paper proposes a UAV trajectory planning algorithm that combines model-free reinforcement learning with an improved depth deterministic strategy gradient algorithm with integral compensation. The trajectory planning problem is modeled as a Markov decision process with different degrees of environmental information missing, and the actions given by the trained agent from the observed state are taken as prior knowledge through the integral compensation method. PPO algorithm is used to train agents off-line in the established flight environment simulator, and the curvature of the trajectory is guaranteed by improving the correlation of the agent's action in time. The experimental results show that this method can generate curvature smooth flight path, and has a high success rate in complex flight environments. It can be extended to different flight environments.

Keywords

adaptive search; depth deterministic strategy gradient; flight path planning; integral compensation; track cost; unmanned aerial vehicles

Hrčak ID:

335085

URI

https://hrcak.srce.hr/335085

Publication date:

30.8.2025.

Visits: 308 *