Scholarlink

<p>For robots in human–robot shared work environments, it has always been expected that the robot can perform as a human colleague as much as possible. However, in the recent human–robot interaction (HRI) community, research on human-like motion generation is scarce, as related work has predominantly explored improving robot appearance, movement patterns, or human understanding. We propose a method, topology-based motion generalization for HRI (ITMG), to generate human-like motion from demonstrations. This method transforms the motion generalization problem into a mesh/graph deformation optimization and captures the spatial relationship between different parts of the human and robot with a topology-based representation. During live interaction, the interaction relationship is maintained with topology and spatial prior. The task instruction from the human colleague is abstracted as a constraint to limit the deformation of the mesh. We validate the effectiveness of the adaptation in an assembly task involving human and robot collaboration. Compared with a benchmark, our method achieved a 17.5–54.3% lead in configuration similarity under the same or even less training data conditions, which indicates that the proposed method can generate a more human-like interaction. We also demonstrate the excellent re-adaptation capacity of our method. The video of the experiments is available at <a href="https://youtu.be/KJnNzaKHelg">https://youtu.be/KJnNzaKHelg</a>.</p>

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-Like Interaction with Topology-Based Optimization and Node Updating

  • 期刊论文

  • 发表日期：

    2025年3月5日

查看全文
  • Yihui Li,
  • Jiajun Wu,
  • Wangcheng Chen,
  • Yisheng Guan,
  • Haifei Zhu

摘要

For robots in human–robot shared work environments, it has always been expected that the robot can perform as a human colleague as much as possible. However, in the recent human–robot interaction (HRI) community, research on human-like motion generation is scarce, as related work has predominantly explored improving robot appearance, movement patterns, or human understanding. We propose a method, topology-based motion generalization for HRI (ITMG), to generate human-like motion from demonstrations. This method transforms the motion generalization problem into a mesh/graph deformation optimization and captures the spatial relationship between different parts of the human and robot with a topology-based representation. During live interaction, the interaction relationship is maintained with topology and spatial prior. The task instruction from the human colleague is abstracted as a constraint to limit the deformation of the mesh. We validate the effectiveness of the adaptation in an assembly task involving human and robot collaboration. Compared with a benchmark, our method achieved a 17.5–54.3% lead in configuration similarity under the same or even less training data conditions, which indicates that the proposed method can generate a more human-like interaction. We also demonstrate the excellent re-adaptation capacity of our method. The video of the experiments is available at https://youtu.be/KJnNzaKHelg.