Exploring the DARPA SubTChallenge
Dr. Timothy Chung, DARPA program manager for the SubT Challenge, explains the competition’s search-and-rescue goals, technology demands, simulation tools, and ways to participate at a Silicon Valley Robotics event hosted by Andra.
In this talk
- 5:32
Robots for underground rescue
Chung describes the challenge’s tunnel, urban, and cave environments, which test different combinations of robots and sensors. The goal is to give emergency commanders useful information before sending people into dangerous spaces, during the one-to-few-hour window when robotic scouts can map conditions and locate targets.
- 14:09
A systems challenge, not just mapping
The competition brings together autonomy, perception, networking, and mobility. Teams must find and classify artifacts, then report their locations within five meters; reports count only when they reach the base station. Courses can change between runs, with obstacles and artifact positions altered.
- 21:20
Virtual competition and shared tools
In the virtual competition, teams submit Docker containers to explore hidden simulated worlds autonomously. The SubT Tech Repo provides robot and sensor models, environment tiles, and artifact models, while the SubT Simulator uses Ignition Gazebo and ROS. Chung also describes marsupial robots, communication breadcrumbing, and dynamic obstacles such as a rockfall that blocks a robot’s route.
- 32:18
Ways to join and contribute
Chung welcomes individuals, student groups, and companies to the Systems or Virtual competitions, and notes that prior participation is not required for upcoming events. The SubT community forum has a teaming category, though participants arrange their own collaborations. Researchers can also use the open-source simulator and public venue datasets, or contribute robot models, sensors, payloads, worlds, and training data.
- 37:23
Community questions and future research
In the Q&A, Chung confirms that one-person teams can compete and says the simulator and datasets are available for research beyond the competition. He points to potential work on machine learning with Ignition and presenting complex 3D spaces to responders. He also encourages research proposals through DARPA’s Seedlings process, including models that help robots decide how to search unfamiliar buildings.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.