Team CSIRO’s Data61 representing Australia in ‘robot Olympics’
Team CSIRO’s Data61 will represent Australia – and the Southern Hemisphere – in the ‘robot Olympics’ this month in the United States.
Experts from around the world have spent the past three years pushing the boundaries of autonomous robotic technology to map, navigate and search underground environments as part of the Subterranean Challenge, run by the US Defense Advanced Research Projects Agency (DARPA).
With teams eliminated from the competition after every phase, only eight remain to contest the 2021 grand final. CSIRO group leader Dr Navinda Kottege said he and the team were thrilled to be one of the final eight teams to compete in the systems track of the challenge.
“In the world of robotics, these challenges are like our Olympics,” Dr Kottege said.
“To the best of our knowledge, this is the first time any Australian team has made it to a DARPA Challenge final, and we’re very proud to showcase Australia’s capabilities in this area on the world stage.”
Breakthroughs discovered through the DARPA Challenge have helped push real-world applications forward, including improving safety and enhanced efficiency in local mining sectors, and promising significant potential in agriculture and manufacturing.
The team’s six autonomous robots will need to locate and report back on items and environmental conditions throughout three underground courses built inside the Louisville Mega Cavern in Louisville, Kentucky.
Challenges are designed to simulate real-world scenarios and involve locating objects representing lost or injured humans, backpacks, or phones, as well as variable conditions such as pockets of gas.
Points are awarded for correct identification and location of items, a task that means the robots need to be capable of mapping the terrain, and maintaining autonomy and communications throughout.
A unique feature of each robot on team CSIRO’s Data61 is that they employ homogeneous sensing. The decentralised multi-agent SLAM solution on each robot (both ground and UAVs) enables them to collaborate on mapping and localisation, which in turn helps autonomous navigation.
As a result, rather than relying on a communication channel back to the base station, it was possible to rely on multi-agent autonomy. Team member and software engineer Fletcher Talbot explains that operators have learned a surprising lesson from this.
“We learned that operators can actually hamper the progress of the robots, because we don’t really know what we’re doing sometimes.
My approach during the different runs was to just let the robots be more autonomous, and only give them very high-level commands rather than trying to do any kind of finessing. Trust the autonomy.”
All ground robots are equipped with hardware designed and developed by CSIRO’s Data61, including the CatPack perception system, plus accompanying perception software and Wildcat SLAM technology that allows highly accurate mapping, localisation, object detection, multi-bot navigation and communication.
CSIRO spinout and Challenge partner Emesent’s aerial robots are equipped with their Hovermap systems to provide state-of-the-art mapping and autonomy capabilities. Unlike many industrial drones, they can operate in the absence of GPS, a highly useful feature for a subterranean challenge.
“Having a fleet of driving, walking and flying robots that are achieving complex missions autonomously and collaboratively is a major technological milestone and an important step towards using autonomous systems for saving lives and helping people,” CTO and Emesent co-founder Dr Kendoul said.
The team’s tracked ‘work horse’ all terrain robots were purchased from Brisbane-based robotics company BIA5, while the two quadrupeds are Boston Dynamics iconic Spot bots.
Professor Ronald Arkin is Director of the Mobile Robot Laboratory at the Georgia Institute of Technology, which helped design the team’s multi-robot task allocation system, enabling robots in the field to collaboratively decide who is best placed to perform a particular task.
“This partnership unites the strengths of field robotics and perception already present in Australia with our experience in AI planning, multi-robot teams, and previous work with DARPA to form a solid and highly competitive effort to win the Challenge,” Professor Arkin said.
Though based in Australia for the competition, CSIRO’s Australian team members will appear at the event via telepresence. US-based representatives and partners Emesent and Georgia Tech will be on the ground.
The winner will receive $US2 million to conduct further research and development, with second place awarded $US1 million and third $US500,000. CSIRO’s Data61 placed fourth in the previous challenge to secure their place in this grand final.
The DARPA Subterranean Challenge will take place from 21-23 September GMT4: https://www.subtchallenge.com/