Robotic manipulation in real-life, human-focused settings is a tricky balancing act, requiring the precision for delicate free-space tasks while also safely managing unexpected contacts. Traditional robotic wrists tend to rely on overly complex controls or bulky designs that still don’t quite hit the mark. That’s why we developed BiFlex—a flexible robotic wrist built with a soft, buckling honeycomb structure that naturally toggles between two modes. In high-stiffness mode, it manipulates household objects with pinpoint accuracy, and in low-stiffness mode, it adapts to external forces. Designed to keep fingertip deflection under 1 cm while supporting up to 500 g, BiFlex integrates with various grippers such as Panda, Robotiq, and BaRiFlex. Our experiments—from surface wiping and precise pick-and-place to using environmental constraints for grasping—show that BiFlex not only simplifies control but also enhances both accuracy and safety in real-world scenarios.
We modeled each honeycomb cell as a network of beams and applied Hooke’s law to derive a simple formula for its vertical stiffness (Keq). By applying Euler’s buckling formula to the diagonal beams, we identified the critical load—the point at which the cell suddenly becomes softer when the input force (Fi) surpasses this threshold. This analysis revealed that the tilt angle (γ) largely influences BiFlex’s vertical stiffness, while the beam thickness (b) mainly adjusts the critical buckling load. By tuning these two parameters, BiFlex can be customized to match the specific needs of various robotic end-effectors.
@misc{jeong2025biflexpassivebimodalstiffness,
title={BiFlex: A Passive Bimodal Stiffness Flexible Wrist for Manipulation in Unstructured Environments},
author={Gu-Cheol Jeong and Stefano Dalla Gasperina and Ashish D. Deshpande and Lillian Chin and Roberto Martín-Martín},
year={2025},
eprint={2504.08706},
archivePrefix={arXiv},
primaryClass={cs.RO},
url={https://arxiv.org/abs/2504.08706},
}