Scholarlink

Soft pneumatic actuators offer numerous advantages, including dexterity in confined spaces and compliance when handling irregularly shaped objects. However, their inherent lack of proprioception remains a major challenge. In this study, we propose a vision-based approach using an internal camera to capture visual features such as stripe patterns and brightness variations within the actuator cavity. For calibration, an external tracking system provides ground truth pose data for each configuration. A neural network is trained to establish the mapping between the internal visual features and the corresponding actuator poses. Finally, we validate the effectiveness of our method through peg-in-hole experiments.

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

Intrinsic Vision-Based Learning for Proprioceptive Sensing of Soft Pneumatic Actuators

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Taoyuan Huang,
  • Jieji Ren,
  • Xinyu Yang,
  • Longyan Wu,
  • Jinnuo Zhang,
  • Ningbin Zhang,
  • Guoying Gu

摘要

Soft pneumatic actuators offer numerous advantages, including dexterity in confined spaces and compliance when handling irregularly shaped objects. However, their inherent lack of proprioception remains a major challenge. In this study, we propose a vision-based approach using an internal camera to capture visual features such as stripe patterns and brightness variations within the actuator cavity. For calibration, an external tracking system provides ground truth pose data for each configuration. A neural network is trained to establish the mapping between the internal visual features and the corresponding actuator poses. Finally, we validate the effectiveness of our method through peg-in-hole experiments.