Evan LernerPenn Engineers to Send Robots Underground in DARPA Subterranean Challenge
By Lauren Salig
While the future of robotics is looking up, many innovative minds in the field are looking down — below the Earth’s surface. Robots owe their increasing autonomy to continuous improvements in satellite-based navigation and on-board sensors, but those advantages are often taken away when a robot goes underground.
In unexplored subterranean environments, even the most advanced robots or devices are often unable to function properly, due to unstable, dangerous environments or an inability to access data. This technological gap may seem insignificant to many, but to construction workers, miners, first responders, and military personnel, it can seriously affect their ability to work safely and effectively.
DARPA, the Defense Advanced Research Projects Agency, is hosting the Subterranean (SubT) Challenge in an attempt to address this issue. DARPA hosts challenges like the current SubT Challenge and their 2015 Robotics Challenge to encourage labs from around the world to develop envelope-pushing robotics technology for addressing real-world problems.
Teams competing in the SubT Challenge are tasked with creating a robotics system that can successfully navigate underground tunnels, urban transportation systems and natural caves. Such a feat will require the robots to quickly navigate unstructured terrain while mapping their location and sending useful data back to the user.
The teams will be evaluated on their performance in a series of tests between August 2019 and August 2021 but were given the opportunity to test their systems in a cave environment at the SubT Integration Exercise (STIX) in early April 2019.
Camillo “CJ” Taylor, professor in the Department of Computer and Information Science, is leading a team in the DARPA SubT Challenge, along with Alejandro Ribeiro, professor in the Department of Electrical and Systems Engineering, and Vijay Kumar, Nemirovsky Family Dean of Penn Engineering and professor in the Department of Mechanical Engineering and Applied Mechanics.
These GRASP Lab members and their students have joined forces with Exyn Technologies and Ghost Robotics, two GRASP Lab spin-offs, to secure their place as one of seven DARPA-funded teams in their track of the SubT Challenge.
Their PLUTO team, short for Pennsylvania Laboratory for Underground Tunnel Operations, will be competing in the “Systems” track of the competition, which requires them to physically create a robotics system and test it in the real world. Other competitors in the “Virtual” track will create computer-based models that will explore virtual underground environments.
When they began approaching the SubT Challenge, Taylor and his team quickly realized that they would need to create a system that incorporated multiple robots with different capabilities to meet DARPA’s demands for the system’s functionality.
“In our analysis of the problem, the feeling was that it involves a number of different elements. With typical aerial robot flight times at 15 minutes and DARPA’s challenge requiring explorations of about an hour, it immediately seems like an all-flying system might not get you there. Having something that’s more terrestrial, like a legged system, offers you a lot more duration,” says Taylor. “On the flip side, a legged vehicle is wonderful, but there are times when you want to explore a 3D environment. You might reach a point in a cave where you want to explore up or down, and you want to use a flying vehicle.”
Based on this initial analysis, Taylor reached out to Exyn Technologies, which specializes in autonomous aerial vehicles, and Ghost Robotics, which specializes in legged robots, and both companies joined the PLUTO team.
However, Taylor knows that the advantages of creating a multi-robot system also come with challenges when it comes to getting the parts of that system to function cohesively underground, where signal availability is inconsistent.
“How do you get a bunch of agents that can only communicate sporadically to collaborate effectively? It can involve making some very complex decisions. Each robot needs to ask, ‘Should I explore?’, ‘Should I act as a communication relay?’, ‘Should I go back to the base station?’. You have to make these decisions in the face of uncertainty and deal with the question of how to do robust communication underground,” says Taylor.
Beyond the issue of coordinating robots’ actions in the face of communication constraints, Taylor says there are at least two other major challenges the PLUTO team will have to address. The first is engineering a robot that can move around safely. Navigation may seem like a basic skill for a robot, but navigating underground terrain is a complicated task that roboticists cannot take for granted.
The second challenge will be meeting DARPA’s perception and mapping standards. The SubT Challenge asks teams to design robots that can create scalable 3D maps of underground environments that are many kilometers in size. Being able to map these large environments accurately will require teams to push the state of the art in robotics.
At the STIX practice event in early April, the PLUTO team, along with the other DARPA-funded teams, had the chance to gauge their progress in stacking up to DARPA’s standards by sending their robots into a Colorado mine. According to Taylor, the event was a glimpse into how PLUTO’s robotics system will perform moving forward to the evaluated stages of the competition.
“We certainly learned a lot. You can see it on paper and hear people talk, but it’s nice to be on site and get a sense of what the challenge is,” says Taylor. “I was pleased with the way the entire team prepared for it. It was joint effort between us, Exyn Technologies, and Ghost Robotics, and we ended up working very well together on site which was good to see, considering it’s a very demanding experience with lots of tight deadlines and not a lot of time.”
For the STIX event, the team tested a robotics system with one flying and one legged component to get an initial feel for how both would function individually in the subterranean environment. For the judged events, PLUTO plans to scale up to a much larger number of robots to perform at the level demanded by DARPA’s parameters.
Now, Taylor and his team are preparing for the first evaluated event this August and are excited to be engineering an innovative robotics system that could be a precursor to future technology with important applications.
“Underground areas are notoriously dangerous to operate in. The value that a robot brings is the ability to send it to places you really don’t want to send a person,” says Taylor. “The opportunity to be able to work on developing systems to operate in these environments is an interesting challenge and something that we hope will be very useful moving forward.”