DARPA Subterranean Challenge

Team CoSTAR preparing for the DARPA Subterranean Challenge Finals in Kentucky. I am pictured in the back row, first from the left.

DARPA Subterranean Challenge - JPL’s Team CoSTAR

The DARPA Subterranean (SubT) Challenge (2018–2021) was a landmark competition aimed at developing autonomous robot teams capable of mapping, navigating, and searching underground environments, ranging from natural caves and mining tunnels to built underground infrastructure, without GPS or human guidance. Team CoSTAR (Collaborative SubTerranean Autonomous Resilient Robots) was led by NASA’s Jet Propulsion Laboratory (JPL) in collaboration with Caltech, MIT, KAIST, and Luleå University of Technology, comprising around 60 engineers. JPL spearheaded the core autonomy software suite, NeBula (Networked Belief‑aware Perceptual Autonomy), which provided robust decision-making, mapping, planning, and networking across diverse robot platforms - wheeled, legged, tracked, and aerial.

Roles

Jan 2021–Sep 2021: Postdoctoral Scholar @ NASA JPL

Team CoSTAR | Artifact Team
  • Developed YOLOv5m6-based perception stack to enable semantic object mapping of e.g. backpacks, cellphones & human survivors using the sensor systems of a team of ground & aerial robots in challenging urban, tunnel & cave-like environments.
  • Designed a multi-modal method for camera/LiDAR robot-relative object localization.
  • Gained hundreds of hours of experience with experimental work, technical challenges, logistics & teamwork during field tests at a variety of locations across the US.

Awards

Videos

JPL’s Team CoSTAR at the DARPA Subterranean (SubT) Robotics Challenge

I make a brief appearance in the above video at around the 1:54 mark.

Video: JPLraw YouTube Channel. Credit: NASA/JPL-Caltech.

Quest For Robotic Autonomy in Extreme Terrains and Conditions: Team CoSTAR at Subterranean Challenge

Video: TeamCoSTAR YouTube Channel. Credit: NASA/JPL-Caltech.

SubT Challenge Team CoSTAR Final Event Full Run

Video: DARPAtv YouTube Channel. Credit: DARPA.

Publications

An Addendum to NeBula: Toward Extending Team CoSTAR's Solution to Larger Scale Environments. IEEE Transactions on Field Robotics, 2024.
Early Recall, Late Precision: Multi-Robot Semantic Object Mapping under Operational Constraints in Perceptually-Degraded Environments. 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022.