Scholarlink

Indoor environments often contain a mix of fixed and movable objects, making it challenging for floor-cleaning robots to achieve high area coverage, a key metric for their effectiveness. This paper proposes a novel human-robot co-design framework aimed at improving the performance of floor-cleaning robots by optimizing indoor furniture layouts. A Large Language and Vision Assistant (LLaVA) is integrated to semantically interpret objects in the environment using visual and spatial cues, classifying them as movable or non-movable. Layout optimization is then carried out using a Multi-objective Genetic Algorithm (MGA) to maximize cleaning coverage while minimizing layout disruption. The robot communicates the suggested modifications through structured natural language, allowing users to make informed adjustments. Experimental validation in a real-world environment demonstrates that the framework achieves higher coverage improvement with minimal disruption, confirming its effectiveness in practical applications.

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

Human–Robot Co-design for Cleaning: Leveraging Vision Language Model and Multi-Objective Optimization for Adaptive Layouts

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • S. M. Bhagya P. Samarakoon,
  • M. A. Viraj J. Muthugala,
  • W. K. R. Sachinthana,
  • Mohan Rajesh Elara

摘要

Indoor environments often contain a mix of fixed and movable objects, making it challenging for floor-cleaning robots to achieve high area coverage, a key metric for their effectiveness. This paper proposes a novel human-robot co-design framework aimed at improving the performance of floor-cleaning robots by optimizing indoor furniture layouts. A Large Language and Vision Assistant (LLaVA) is integrated to semantically interpret objects in the environment using visual and spatial cues, classifying them as movable or non-movable. Layout optimization is then carried out using a Multi-objective Genetic Algorithm (MGA) to maximize cleaning coverage while minimizing layout disruption. The robot communicates the suggested modifications through structured natural language, allowing users to make informed adjustments. Experimental validation in a real-world environment demonstrates that the framework achieves higher coverage improvement with minimal disruption, confirming its effectiveness in practical applications.