Probabilistic approach to feedback control enhances multi-legged locomotion on rugged landscapes
Published in IEEE Transactions on Robotics, 2025
This paper demonstrates that a lightweight, interpretable feedback controller—driven only by binary foot-contact sensing—can substantially improve multilegged locomotion on rugged, unstructured terrain. By introducing a probabilistic model that links terrain-induced contact disruption to locomotion speed, we derive a cycle-to-cycle control law that modulates vertical body undulation to recover effective ground contact. Extensive laboratory and outdoor experiments show consistent gains in both average speed and speed stability, highlighting a practical approach to robust field locomotion with minimal sensing and computation.
Recommended citation: He, Juntao, Baxi Chong, Jianfeng Lin, Zhaochen Xu, Hosain Bagheri, Esteban Flores, and Daniel I. Goldman. (2025). "Probabilistic approach to feedback control enhances multi-legged locomotion on rugged landscapes." IEEE Transactions on Robotics.
Download Paper