Soft-Rigid Coupled Blade Leg Achieves Spatio-temporal Terrain Classification with Minimal Sensor Configuration
H. P. Chapa Sirithunge, Vijay Chandiramani, Yue Xie, Helmut Hauser, Andrew Conn, Fumiya Iida
Abstract
Fast-legged humanoid robots are transforming industries from manufacturing to medical robotics, with the global market projected to grow from $0.67 billion in 2024 to $2.27 billion by 2033 at a 14.3% CAGR. Despite rapid advancements, challenges remain in navigating complex terrains, especially uneven, deformable, and high-friction surfaces. This paper presents the first minimally sensorised blade leg made by coupling soft and rigid materials for robots: an alternative approach for multimodal sensing and advanced control algorithms in terrain navigation. This incorporates a passive leg design embedded with barometric pressure sensors that are proven to retain high dimentional spatio-temporal data. Hence we hypothesized that barometric pressure sensors can capture multidimensional terrain data and subtle surface compliance changes through spatiotemporal pressure patterns. The blade was mounted on an UR5 robotic arm and tested in terrains of varied textures, including aluminium, pebble, coir, and sandpaper; materials spanning a diverse range of stiffness. Spatiotemporal data from the sensors were recorded and analyzed to assess terrain characteristics and leg-terrain interactions under different conditions. The results demonstrated that barometric pressure sensors could accurately recognize different terrains with as few as three sensors in a 2-second time frame. Recognition accuracy improved with more sensors, demonstrating the effectiveness of morphologically adapted composite structures with optimally placed minimal sensors.
BibTeX
@inproceedings{iros2025_softrigidcoupled,
title = {Soft-Rigid Coupled Blade Leg Achieves Spatio-temporal Terrain Classification with Minimal Sensor Configuration},
author = {H. P. Chapa Sirithunge and Vijay Chandiramani and Yue Xie and Helmut Hauser and Andrew Conn and Fumiya Iida},
booktitle = {IROS 2025},
year = {2025}
}