Design and Control Co-Optimization for Automated Design Iteration of Dexterous Anthropomorphic Soft Robotic Hands

TL;DR

This paper presents an automated co-optimization framework for designing and controlling dexterous soft robotic hands, using simulation-based iterative design and validation through real-world teleoperation tasks, outperforming prior designs.

Contribution

It introduces a novel co-optimization pipeline combining genetic algorithms and policy transfer for automated design iteration of soft robotic hands in simulation.

Findings

Optimized designs outperform existing soft robotic hands in real-world tasks.

Simulation trends in design performance correlate with real-world results.

Automated iteration effectively guides design choices despite the sim-to-real gap.

Abstract

We automate soft robotic hand design iteration by co-optimizing design and control policy for dexterous manipulation skills in simulation. Our design iteration pipeline combines genetic algorithms and policy transfer to learn control policies for nearly 400 hand designs, testing grasp quality under external force disturbances. We validate the optimized designs in the real world through teleoperation of pickup and reorient manipulation tasks. Our real world evaluation, from over 900 teleoperated tasks, shows that the trend in design performance in simulation resembles that of the real world. Furthermore, we show that optimized hand designs from our approach outperform existing soft robot hands from prior work in the real world. The results highlight the usefulness of simulation in guiding parameter choices for anthropomorphic soft robotic hand systems, and the effectiveness of our automated design iteration approach, despite the sim-to-real gap.