Scholarlink

<p>Soft continuum robots enable dexterous manipulation but present challenges for shape sensing due to their continuous deformability. We report a soft continuum robot that integrates a conductive polymer composite (CPC) based on graphite and PDMS directly into a node-based lattice structure for intrinsic sensing, coupled with a neural network for near-real-time shape reconstruction. The CPC serves both as the structural material and a distributed strain sensor, with multiple mesh-like sensing segments embedded along the robot to preserve compliance without external sensors. A tendon-driven system with three motors enables actuation, while a compact data acquisition unit monitors resistance changes in the CPC network. To address the nonlinear and hysteretic response of the CPC, we employ a Conformer-based neural network that fuses resistance time-series with tendon inputs for state estimation. Experiments show reliable strain sensing and accurate shape reconstruction even under external loads, while the trained reconstruction model achieved an end-effector RMSE of 6.3 mm and a mean position error of 3.8 mm. In addition, we demonstrate the capability to accurately predict the geometry of grasped objects using the self-sensing soft continuum structure. This approach enables near-real-time proprioception without compromising flexibility, with potential applications in biomedical manipulation and remote inspection.</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

A 3D printing-enabled soft continuum robot with integrated sensing for multi-purpose predictions with machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月14日

查看全文
  • Guo Liang Goh,
  • Chunpeng Yu,
  • Kyohei Watanabe,
  • Guo Dong Goh,
  • Xinchao Gao,
  • Ruichen Zhang,
  • Han Wei Chua,
  • Max Austin,
  • Kohei Nakajima,
  • Wai Yee Yeong

摘要

Soft continuum robots enable dexterous manipulation but present challenges for shape sensing due to their continuous deformability. We report a soft continuum robot that integrates a conductive polymer composite (CPC) based on graphite and PDMS directly into a node-based lattice structure for intrinsic sensing, coupled with a neural network for near-real-time shape reconstruction. The CPC serves both as the structural material and a distributed strain sensor, with multiple mesh-like sensing segments embedded along the robot to preserve compliance without external sensors. A tendon-driven system with three motors enables actuation, while a compact data acquisition unit monitors resistance changes in the CPC network. To address the nonlinear and hysteretic response of the CPC, we employ a Conformer-based neural network that fuses resistance time-series with tendon inputs for state estimation. Experiments show reliable strain sensing and accurate shape reconstruction even under external loads, while the trained reconstruction model achieved an end-effector RMSE of 6.3 mm and a mean position error of 3.8 mm. In addition, we demonstrate the capability to accurately predict the geometry of grasped objects using the self-sensing soft continuum structure. This approach enables near-real-time proprioception without compromising flexibility, with potential applications in biomedical manipulation and remote inspection.