Scholarlink

<p>Soft devices critically require design approaches that can fully realise their embodied potential. Soft robots are an especially problematic soft system as behaviour is complex and data is difficult to obtain, leading to narrow exploration of potential embodiments and inaccurate behavioural assessment. We create the first high-throughput design approach for soft robotics to address these challenges. Scaleable and automated, our approach adaptively combines simulated and experimental assessment to efficiently explore a design space of soft grippers in an automated closed loop. In a prototype study using this high-throughput regime, we demonstrate discovery of higher-performing grippers than comparative methods, and automatically identify and close the simulation-to-reality gap, as well as recording an order of magnitude more experimental grasps than comparative approaches in the literature. Our experimental regime is significantly differentiated from the current literature, offering a realistic route to turn soft robotics into an increasingly data-rich domain and opening up previously unattainable opportunities for the design of soft systems. Data for this paper is publicly available through CSIRO’s Data Access Portal: <a href="https://data.csiro.au/collection/csiro:65672">https://data.csiro.au/collection/csiro:65672</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

High-throughput soft robot design via an adaptive experimental platform

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Rafael Oliveira,
  • Josh Pinskier,
  • Xing Wang,
  • Lois Liow,
  • Sarah Baldwin,
  • James Brett,
  • Vinoth Viswanathan,
  • Richard Scalzo,
  • David Howard

摘要

Soft devices critically require design approaches that can fully realise their embodied potential. Soft robots are an especially problematic soft system as behaviour is complex and data is difficult to obtain, leading to narrow exploration of potential embodiments and inaccurate behavioural assessment. We create the first high-throughput design approach for soft robotics to address these challenges. Scaleable and automated, our approach adaptively combines simulated and experimental assessment to efficiently explore a design space of soft grippers in an automated closed loop. In a prototype study using this high-throughput regime, we demonstrate discovery of higher-performing grippers than comparative methods, and automatically identify and close the simulation-to-reality gap, as well as recording an order of magnitude more experimental grasps than comparative approaches in the literature. Our experimental regime is significantly differentiated from the current literature, offering a realistic route to turn soft robotics into an increasingly data-rich domain and opening up previously unattainable opportunities for the design of soft systems. Data for this paper is publicly available through CSIRO’s Data Access Portal: https://data.csiro.au/collection/csiro:65672.