Scholarlink

Global e-commerce has spurred demand for efficient delivery services, particularly in the last-mile segment. Implementing unmanned aerial vehicles with a slung payload system offers an effective method for addressing this requirement. Controlling this multi-body system poses unique challenges such as real-time adaptability and unpredictable dynamics. This work describes the development of an on-board model free Reinforcement Learning (RL) controller capable of achieving precise waypoint tracking while ensuring stability and accuracy in the presence of payload-induced perturbations. The developed controller was trained and validated in a simulated environment then implemented on a prototype quadcopter in an indoor testing environment using an integrated motion capture system.

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

Reinforcement Learning Application for Controlling a Quadcopter Carrying a Slung Payload

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Nourah Al Saud,
  • Eric Lanteigne

摘要

Global e-commerce has spurred demand for efficient delivery services, particularly in the last-mile segment. Implementing unmanned aerial vehicles with a slung payload system offers an effective method for addressing this requirement. Controlling this multi-body system poses unique challenges such as real-time adaptability and unpredictable dynamics. This work describes the development of an on-board model free Reinforcement Learning (RL) controller capable of achieving precise waypoint tracking while ensuring stability and accuracy in the presence of payload-induced perturbations. The developed controller was trained and validated in a simulated environment then implemented on a prototype quadcopter in an indoor testing environment using an integrated motion capture system.