Dynamic Locomotion on Yielding Terrain

Robotic performance on natural substrates such as sand, snow, and grass is currently quite limited when compared to the performance of most animals.  Much of this difference can be attributed to the complexity of intrusion into yielding ground and to the high degree of spatial variation in substrates compared to other natural environments such as air and water. These factors impose significant challenges on robotic design and control. This project aims to integrate physics models and control methodologies for enhanced robotic legged locomotion on yielding terrain.


Locomotion on soft ground is explored using a one-legged robot jumping on a tunable model substrate generated with a fluidized bed filled with granular material. Ground reaction forces are dependent on intrusion velocity, step spacing, foot geometry and the packing state of the ground material.  Understanding these parameters and how to take advantage of them for effective motion is a key factor for the design and control of new robots whose agility will one day match that of animals.