Underactuated Robotics in Granular & Underwater Environments

One‑line summary

Designed and experimentally validated an untethered, underactuated digging robot that can both swim through sand and sense buried obstacles, demonstrating the fastest reported subsurface locomotion at depth for an untethered robot and the first such demonstration in natural beach sand. 

Context & scope

Granular environments such as sand, soil, and regolith present extreme challenges for robots due to high resistive forces, unpredictable solid–fluid transitions, and the lack of reliable sensing modalities. This project explored whether underactuated appendages—inspired by biological diggers like sea turtles—could be used to both generate propulsion and enable obstacle sensing in granular media using minimal actuation and compact hardware.

The work spanned mechanical design, terradynamics‑informed experimentation, sensing, and full‑system integration, culminating in an untethered robot capable of subsurface locomotion in both controlled lab conditions and natural beach sand. 

Outcomes (fast facts)

• Fastest reported untethered subsurface locomotion at a depth of 5 inches(127 mm) in granular media (≈ 1.2 mm/s).  Fastest  untethered burrowing robot at the time.

• First demonstration of an untethered robot swimming beneath the surface of natural beach sand. 

• Validated obstacle sensing in sand using force feedback from underactuated appendages. 

• Published in Advanced Intelligent Systems (2023) under a Creative Commons license.

What I owned

• Robot architecture & mechanical design: Designed the full untethered robot, including a sealed terradynamic body, single‑motor actuation, worm‑gear transmission, and anisotropic underactuated appendages.

• Underactuated appendage design & optimization: Designed multi‑link appendages with tunable joint constraints and experimentally identified an optimal locking angle that maximized thrust asymmetry in granular media. 

• Terradynamic experimentation: Designed and ran controlled drag, lift, and torque experiments in fluidized granular beds to inform body shape, appendage geometry, and control surface design. 

• Granular sensing strategy: Developed and validated a sensing approach where changes in granular flow and resistance during appendage motion reveal the presence of nearby obstacles. 

• System integration & field testing: Integrated actuation, power, sensing, and wireless control into a fully untethered system and led lab and beach demonstrations. 

Key technical contributions

• Asymmetric thrust via underactuation: Demonstrated how passive joint constraints create low‑drag return strokes and high‑force power strokes, enabling net forward motion without fully actuated limbs. 

• Granular lift control using “terrafoils”: Designed and experimentally validated granular control surfaces that counter upward lift forces, allowing sustained horizontal subsurface swimming. 

• Haptic obstacle sensing in sand: Showed that force measurements from moving appendages can detect buried obstacles above the robot by sensing disruptions in granular flow. 

How I approached it

• Started with benchtop terradynamics experiments to understand force asymmetries and failure modes

• Used iterative mechanical prototyping to tune appendage geometry and joint constraints

• Combined physics‑based reasoning (RFT) with experimental validation rather than relying on simulation alone

• Progressed from component‑level tests → tethered robot → fully untethered system → field deployment

Publication

Toward Robotic Sensing and Swimming in Granular Environments using Underactuated Appendages - Advanced Intelligent Systems, 2023

Tags

Underactuated Robotics • Bio‑Inspired Design • Granular Media • Experimental Mechanics • Robophysical Modeling • Field Robotics