PITTSBURGH (AP) – After practice in a former limestone mine and an abandoned hospital outside of Pittsburgh, a fleet of robots from Carnegie Mellon University heads to Kentucky for a final test of their ability to autonomously navigate an undefined underground route.
Objective: Collect as many items as possible in one hour while traveling through a cave in Louisville, Kentucky.
Objective: To create autonomous rescue robots to assist in search and rescue operations for the military and first responders.
Kicker: There is only one operator and a fleet that can include a dozen rescue robots, so the machines must be able to move around and make decisions on their own. It’s hard enough to make autonomous robots useful on earth. The CMU team is trying to go deeper into this.
“At the heart of this challenge is a research problem,” said Sebastian Scherer, co-leader of the CMU team and associate professor at the Institute of Robotics. “You find yourself in an environment that you have never been in, so there is a lot of uncertainty, and the robot must be able to decide what to do when I face uncertainty.”
CMU researchers and students have teamed up with Oregon State University to create their own fleet, dubbed Team Explorer.
Next week, the group is competing for the chance to win the $ 2 million grand prize in the final round of the DARPA Subterintage Challenge, a competition sponsored by the Defense Advanced Research Projects Agency of the U.S. Department of Defense to develop autonomous technology that allows control of underground scenarios.
DARPA has run similar races in the past, and CMU has proven itself in several competitions, ranging from those focused on self-driving technologies to developing software that helps with language translation and document understanding, said Martial Hebert, dean of the School of Computer Science.
In 2007, a team from Pittsburgh won the DARPA Urban Challenge, a race to develop an unmanned vehicle that could travel in traffic conditions, navigate intersections, merge and park. This, in turn, led to progress in self-driving car research and spurred some of the team members to start their own companies, including Aurora and Argo AI.
These types of problems help researchers “find real problems,” said Mr. Hebert. “With this scenario of problems, it pushes us to define what the next frontier will be, what the next problem we have to solve.”
The DAPRA Underground Challenge, or SubT, is geared towards autonomous rescue robots that can navigate an underground route and deal with environmental uncertainties they’ve never seen before.
Researchers will also not see the course before sending their cars there. For them, it’s like putting their robots in a black hole, ”said Matt Travers, another co-director of Team Explorer and a systems scientist at the Institute of Robotics. He said there was no GPS or communication.
Each team will have 60 minutes to walk the track in search of approximately 40 objects, including fire extinguishers, ropes and a helmet with a flashlight. The objects and course will mimic what the robot might find in an underground disaster scenario, as if it were sent on a search and rescue mission in a real application, the researchers said.
“In many rescue situations, it takes a long time for people to get inside,” said Mr Scherer. “First you need to make sure that it is safe, that there are no gases. … The idea is that these (robots) can come in very quickly, give you a lot of awareness, and then you can send humans or your other robots. “
Now helping to lead the team, Mr. Scherer has also participated as a student on the CMU team in both the DARPA Urban Challenge and the Grand Challenge, which focused on self-driving technologies for navigating the desert landscape.
The CMU fleet for the SubT mission includes four small drones, three large drones, three ground robots, and two walking robots.
Part of the job of building their fleet was to bring in all the different types of technology, Mr. Scherer said. “Think of it as a human team,” he
suggested. Some coworkers will get along better than others, and when all of the employees come together, there is bound to be “friction”.
Just as a team of human employees has strengths and weaknesses among the group, a fleet of robots is also created to solve various problems. Some robots are better at climbing stairs, while others are built to navigate tight spaces, says Scherer.
For now, the human operator will help the rescue robots determine who is best suited for the various tasks, but ultimately he hopes that decision making will become part of the machine’s autonomous capabilities.
The development of Team Explorer technology also had to consider a “hierarchy of autonomy,” said Stephen Willits, lead test engineer and research scientist at CMU.
It starts with a “local scheduler” or software that helps the machine decide what to do based on its environment. According to Mr. Willits, this is what keeps him from bumping into objects or falling off a cliff.
The team then used additional technologies that create a “global planner” that helps machines paint a picture of the entire environment and communicate with each other. This helps to make sure the robots do not cover the same area as other members of the fleet.
Team Explorer is also preparing for the failure of their rescue robots.
Someone will get stuck, someone will not survive, and someone will turn in the wrong direction. The goal is to make sure the robots and the entire fleet as a whole are resilient enough to recover from these disruptions, says Scherer.
“We expect these systems to run into problems and have to get out of them on their own,” he said.
The teams have been working on their fleets for over three years since the start of the competition in September 2018.
In the trial, CMU and Oregon have already taken home the first place prize in the first round, in which robots moved through a research mine in the town of South Park, near Pittsburgh. The mine is operated by the National Institute for Occupational Safety and Health.
Team Explorer also finished second in the next round of the competition, passing through an abandoned nuclear power plant in Olympia, Washington.
In the final round next Thursday, Team Explorer will face seven other groups for a prize pool that includes $ 2 million for the first place winner, $ 1 million for the runner-up, and $ 500,000 for the third place team.
According to Mr. Scherer, Team Explorer has about 20 people who are the “core” members, and about 100 people who make up the extended team.