UC San Diego

Robots have shown prowess in demonstrating navigation in many extreme environments, except in granular media (GM), which remains relatively unexplored. GM such as sand, dry snow, and gravel are some of the most common substrates on Earth as well as other terrestrial planets, yet GM is one of the most challenging environments to traverse. To navigate through GM, robots have to overcome large depth-dependent forces and contend with non-zero yield stress that may cause unpredictable fluid/solid resistance forces, all with extremely limited capabilities for sensing. In this thesis, I address the problems of navigating GM by designing bioinspired, underactuated, and passively compliant robot limbs.

Locomotion on GM poses high demands on foot placement and joint control as GM can exist as a solid and can flow like a liquid causing robots to sink and slip. Taking inspiration from passive compliance in camel hooves. I designed a robot foot that uses granular jamming. The foot changed shape passively when in contact with the ground to reduce sinking, and actively changed stiffness for the ability to apply sufficient propulsion forces which led to improved locomotion parameters. For locomotion within GM, I proposed a novel autonomous, untethered robot that swims with underactuated appendages, which enable both large propulsion forces through limb motion and obstacle sensing over a wide range around the robot. To optimize the design of appendages, I experimentally identified the optimum morphological and actuation parameters for generating thrust. I also investigated how the presence of an obstacle buried in GM influenced the granular flow around a moving appendage, enabling the ability to sense obstacles in grains. The results from sensing and propulsion experiments were integrated into an untethered robot capable of subsurface locomotion with a speed of ~1.2 mm/s at a depth of 127 mm. Obstacle detection was demonstrated through experiments with embedded force sensors on the appendages of the robot. Overall, this thesis sheds light on how passively deforming and underactuated structures can enable movement on and within GM with limited limb control while still enabling sensing capabilities.