MIT and TU Delft's soft robot arm senses shape and contact with 97% accuracy using fluidic channels

Soft continuum robots need embedded sensing for both self-awareness (proprioception) and detecting external interactions, but integrating sensors into highly deformable architected structures is challenging. A team led by Daniela Rus (MIT) and Cosimo Della Santina (TU Delft) introduces a model-based decoupling strategy that reuses the same fluidic pressure sensors for both tasks. Their design, called the Innervated Trimmed Helicoid (ITH), contains six air channels zigzagging along each segment's circumference. With only three principal kinematic degrees of freedom (axial compression, bending in x, bending in y), the six pressure readings create an overdetermined system. A piecewise constant curvature model maps pressures to shape, while Huber regression identifies outlier channels whose residuals signal external contact. On a single ITH segment, the approach achieves shape estimation with a relative bending error of 0.11 ± 0.02 and a contact detection rate of 97% across 178 trials.

The team integrated eight ITH segments into Air-Helix, a tendon-driven soft continuum manipulator, and demonstrated three whole-arm capabilities: tactile teaching by demonstration (the arm learns from human-guided motions), admittance-controlled force regulation (adjusts stiffness in response to applied forces), and tactile object reconstruction (recreates 3D shape by touching objects). The results suggest that localized fluidic innervation combined with model-based redundancy resolution offers a practical, hardware-light path to concurrent proprioception and contact sensing in architected soft robots. Accepted for publication at IROS 2026, this work addresses a key bottleneck in soft robotics: how to make compliant manipulators that can both feel their own shape and interact safely with the environment without dedicating separate sensors to each function.