Scholarlink

<p>Unlike humans, who can perceive, interpret and adapt to physical interactions through their own bodies, enabling adaptive interaction with dynamic and unstructured environments, most robotic systems still lack such embodied sensorimotor intelligence. Conventional robotic arms rely on vision or sparse force sensing that limit their ability to perceive distributed contact, interpret interactions and respond safely in dynamic, human-centred settings. Here we present EmArm, an embodied robotic arm that integrates rigid kinematics with large-area soft tactile skins, proprioceptive sensing and a closed-loop perception–action framework. The system achieves submillimetre tactile localization, high-fidelity whole-arm sensing under environmental disturbances and real-time feature extraction for interaction-aware control, enabling touch-based human intention recognition, contact-rich manipulation and tactile-driven trajectory replanning in visually occluded, human-involved environments. This work establishes a scalable route towards physically intelligent robots capable of safe, adaptive and intuitive operation in unstructured environments.</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

Embodied sensorimotor integration for whole-arm tactile sensing and adaptive robotic manipulation

  • 期刊论文

  • 发表日期：

    2026年7月20日

查看全文
  • Yifeng Tang,
  • Tianci Yin,
  • Tieshan Zhang,
  • Liu Yang,
  • Gen Li,
  • Hao Ren,
  • Zhengrong Ling,
  • Haoxiang Zhao,
  • Ruijia Zhang,
  • Yajing Shen

摘要

Unlike humans, who can perceive, interpret and adapt to physical interactions through their own bodies, enabling adaptive interaction with dynamic and unstructured environments, most robotic systems still lack such embodied sensorimotor intelligence. Conventional robotic arms rely on vision or sparse force sensing that limit their ability to perceive distributed contact, interpret interactions and respond safely in dynamic, human-centred settings. Here we present EmArm, an embodied robotic arm that integrates rigid kinematics with large-area soft tactile skins, proprioceptive sensing and a closed-loop perception–action framework. The system achieves submillimetre tactile localization, high-fidelity whole-arm sensing under environmental disturbances and real-time feature extraction for interaction-aware control, enabling touch-based human intention recognition, contact-rich manipulation and tactile-driven trajectory replanning in visually occluded, human-involved environments. This work establishes a scalable route towards physically intelligent robots capable of safe, adaptive and intuitive operation in unstructured environments.