Scholarlink

Coverage Path Planning (CPP) is a critical challenge in autonomous systems, with applications spanning lawn mowing to search-and-rescue. Existing methods often struggle in dynamic, uncertain environments. This paper presents Think4CPP, a method that integrates a Transformer-based world model with the Soft Actor-Critic (SAC) algorithm to address these challenges. The Transformer’s self-attention mechanism captures long-term dependencies and complex state dynamics in robot-environment interactions. By modeling aleatoric uncertainty through Gaussian variance and incorporating it into SAC training, Think4CPP enhances policy robustness and decision-making efficiency in uncertain conditions. Empirical results show superior coverage efficiency and safety compared to existing methods. This work advances CPP by introducing a Transformer-based model, improving risk assessment via uncertainty modeling, and offering a framework that sets new standards for safe, efficient CPP.

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

Think4CPP: Reinforcement Learning by Thinking with Latent World Model for Safe Coverage Path Planning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhentang Liao,
  • Meng Li,
  • Zhongxue Gan,
  • Lihua Zhang,
  • Zhiyan Dong

摘要

Coverage Path Planning (CPP) is a critical challenge in autonomous systems, with applications spanning lawn mowing to search-and-rescue. Existing methods often struggle in dynamic, uncertain environments. This paper presents Think4CPP, a method that integrates a Transformer-based world model with the Soft Actor-Critic (SAC) algorithm to address these challenges. The Transformer’s self-attention mechanism captures long-term dependencies and complex state dynamics in robot-environment interactions. By modeling aleatoric uncertainty through Gaussian variance and incorporating it into SAC training, Think4CPP enhances policy robustness and decision-making efficiency in uncertain conditions. Empirical results show superior coverage efficiency and safety compared to existing methods. This work advances CPP by introducing a Transformer-based model, improving risk assessment via uncertainty modeling, and offering a framework that sets new standards for safe, efficient CPP.