Scholarlink

Traditional coordination control methods for unmanned aerial vehicles (UAVs) exhibit limited adaptability, particularly in complex and dynamic environments. In this study, we propose a novel formation control framework based on the Proximal Policy Optimization (PPO) algorithm augmented with a dual-reward mechanism, targeting leader-follower structured UAV formations in dynamic scenarios. The state space is constructed by integrating position and velocity vectors, while the reward function is designed to incorporate both relative distance and velocity alignment. Experimental results show that, compared with the Deep Deterministic Policy Gradient (DDPG) algorithm, the proposed approach reduces the mean tracking error by \(64\%\) and significantly mitigates trajectory oscillations.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Unmanned Aerial Vehicle Following Flight Based on Reinforcement Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mingyu Yuan,
  • Yongnan Jia

摘要

Traditional coordination control methods for unmanned aerial vehicles (UAVs) exhibit limited adaptability, particularly in complex and dynamic environments. In this study, we propose a novel formation control framework based on the Proximal Policy Optimization (PPO) algorithm augmented with a dual-reward mechanism, targeting leader-follower structured UAV formations in dynamic scenarios. The state space is constructed by integrating position and velocity vectors, while the reward function is designed to incorporate both relative distance and velocity alignment. Experimental results show that, compared with the Deep Deterministic Policy Gradient (DDPG) algorithm, the proposed approach reduces the mean tracking error by \(64\%\) and significantly mitigates trajectory oscillations.