Date: Monday, April 13rd, 2026

Time: 1PM to 3PM ET

Location: Howey building N 201 or Zoom Link

Juntao He

Robotics Ph.D. Student

Woodruff School of Mechanical Engineering

Georgia Institute of Technology

Committee:

Dr. Daniel I. Goldman (advisor) – Schoold of Physics, Georgia Institute of Technology

Dr. Sehoon Ha – School of Interactive Computing, Georgia Institute of Technology

Dr.  Zeb Rocklin –  Schoold of Physics, Georgia Institute of Technology

Dr. Tony Chen – Woodruff School of Mechanical Engineering, Georgia Institute of Technology

Dr. Baxi Chong  – Department of Mechanical Engineering, The Pennsylvania State University

Abstract:

Locomotion in confined, unstructured environments remains a foundational challenge in robotics. Reliable mobility in these settings would unlock high-impact applications ranging from search-and-rescue in disaster debris to autonomous infrastructure inspection and residential pest control. However, robust operation is hindered by (i) nonlinear, contact-rich dynamics, (ii) degraded exteroceptive sensing (LiDAR/vision) under occlusion, dust, and clutter, and (iii) planning and control complexity under severe geometric constraints.

This thesis develops an elongated many-legged robotic system that combines leg redundancy with a flexible backbone, enabling resilient locomotion with minimal sensing. First, I show how leg redundancy can simplify control on rough terrain: beyond a threshold number of leg pairs, open-loop wave-template gaits achieve robust traversal by passively tolerating localized contact disturbances. Building on this foundation, I integrate low-profile tactile sensing to close the loop in environments where vision is unreliable. Foot-contact feedback is used to modulate body undulation and limb stepping online, yielding up to a twofold speed increase in both laboratory and outdoor field tests. A tactile antenna further enables vision-free behaviors critical for confined operation, including vertical obstacle negotiation and rapid turning in cluttered tunnels.

Together, tactile-centric control and a low-profile sensing suite eliminate the need for line-of-sight, wide-baseline exteroceptive payloads (e.g., multiple cameras, LiDAR, and illumination), enabling a smaller cross-section and greater adaptability in geometrically constrained environments. As a result, the robot achieves robust omnidirectional locomotion in passages as narrow as 30 cm and under 30 cm overhang clearance. Finally, with sparse human guidance, the robot autonomously navigates drainage conduits and performs pipe-inspection tasks such as erosion detection and clog-state monitoring.