The MIST Lab team stand with their robotic swarm. | SkyNews
The MIST — Making Innovative Space Technology — lab team stand with their robotic swarm. (Caroline Perron)

Montreal team tests robot ‘swarms’ for Moon exploration

A robot swarm could be a small step for a few machines — and a giant leap for artificial intelligence on the Moon.

Sending humans into the unknown sounds daring — but why not have a robot do it for us? Better yet: what if the robot could communicate with other ones to explore a complex environment?

Giovanni Beltrame hopes to make exploration of the Moon happen using these kind of robot “swarms” for environments such as lunar lava tubes — areas that are too treacherous for a rover or human, but could provide valuable information about the Moon’s interior and long history as well as potential radiation shelters for crewed lunar bases.

Beltrame, a professor of computer engineering at Polytechnique Montréal, is leading a team investigating how robot swarms work together and how they resolve arguments. While it sounds like a futuristic concept, Beltrame said we need look no further than the Perseverance rover’s exploration on Mars to see an early-stage effort at collaboration.

“If you send many robots they can explore in parallel, and we’ve seen that already happening with Perseverance and Ingenuity on Mars,” Beltrame said in an interview. Ingenuity is a drone helicopter that rode to the Red Planet along with Perseverance, landing in February 2021. 

The MIST Lab team stand with their robotic swarm. | SkyNews
The MIST — Making Innovative Space Technology — lab team stand with their robotic swarm. (Caroline Perron)

The little helicopter made the first flights on Mars this year and is still going strong, having finished 11 flights by the time of this writing. It is currently making daring hops between “airfields” (pre-identified landing zones) to test out scouting techniques for future missions.

Beltrame wants to step up the role of such robots to give them a measure of decision-making capability, powered by machine learning — meaning the ability for a computer to learn from a dataset and to apply that education to new problems. 

For example, a swarm of robots could act as backups for each other to figure out complex terrain or other problems they may encounter, Beltrame explained.

“The swarm provides multi-robot exploration in an almost fully autonomous way,” he said. “As humans, we want to receive the results and tell the robots what we’re interested in. What we want to do — we want to break this paradigm of one robot, one team, one mission by having multiple systems explore the Moon and Mars.”

In a swarm, the robots would work together, similarly to how the neurons in your brain operate, Beltrame explained. While neurons are simple units of the nervous system, working together they allow the human brain to think over complex problems. The same principle may be applied to swarm engineering, Beltrame continued.

“Simple elements that we design can act together and form something that is more complex,” Beltrame said. “This property is called emergence, and the implications are many. For example, we can use this concept for nanomedicine. Having tiny robots in your bloodstream that do [cell-cleaning] work — despite the fact they are very simple, [they] can do a complex mission.”

Beltrame has numerous space research projects underway. For example, the Canadian Space Agency awarded his research centre — the Making Innovative Space Technology (MIST) lab — a grant in 2019 called Flights and Fieldwork for the Advancement of Science and Technology. FAST supports student research in space-like missions, and CSA said the robots’ “discussions” about where to go next could be key to future space exploration initiatives.

A student operates a robot swarm. | SkyNews
A student operates a robot swarm. (Martin Primeau)

“While underground, the robots work in a democracy,” the CSA stated of their decision-making process. “They will vote on who does what by bidding on tasks based on their skill set, power level, ability to navigate, distance they can travel, etc. The highest bid will win. This system is somewhat similar to that used by air traffic controllers and companies such as Uber who coordinate independent drivers and vehicles.”

Beltrame has also just received support from the European Space Agency to do simulated “prospecting missions” in the Netherlands. Robots will explore a simulated Moon environment to try to identify interesting geological features. 

“We’re going to do a simulated lunar polar mission, where we have very difficult lighting conditions,” Beltrame said, referring to the fact that the region can be plunged into full daylight or full darkness depending on the Moon’s position in its orbit. Moreover, there are deep craters that have permanently shadowed regions — perhaps a good spot for ice, but not so easy to view. The robots will have to navigate a simulated south pole area, pick out the most interesting rocks, and then place a spectrometer on the rocks to learn more about the region’s composition. 

Beltrame’s laboratory will also participate in the Defense Advanced Research Projects Agency’s (DARPA) Subterranean Challenge in Kentucky from September 21 to 23; one of his students is already working with NASA’s Jet Propulsion Laboratory on an interface a human may use to communicate with a robot swarm, which may prove useful next month, Beltrame said.

“There are a bunch of robots that go inside a cave, and they have to explore the cave and find objects of interest,” Beltrame explained of the challenge. “DARPA will communicate which objects have to be found — that could be a fire extinguisher, a jacket, stuff that is not normally in a cave. We have to find it and report the location of these objects with very high accuracy.”

Beltrame’s most optimistic scenario for future research is to send robots to explore the Moon itself. He acknowledges it sounds far-fetched for the moment, but he is keeping an eye on NASA’s Commercial Lunar Payload Services program (CLPS) and CSA’s Lunar Exploration Accelerator Program (LEAP), both of which are opening opportunities to academia and companies for Moon exploration later in the decade, if all goes to plan.

“What I’m mostly interested in is what we call perception … how robots look at the world, how they perceive the world and how they look at the environment with their own sensors,” he said. As the robots make decisions, ideally they will work with each other and also try to use the environment to their advantage, he said. 

Beltrame also is interested in finding a solution to having a single human work with perhaps 100 robots simultaneously, which has implications for industries on Earth such as the growing self-driving market. “How many can you monitor? What kind of interface? What kind of communication can these robots provide to the human, and how can the human send information back?” he asked.

“This is very interesting and complicated, because we don’t want to exceed the cognitive capabilities of humans,” he said. “This is actually one of the most interesting interdisciplinary aspects of this work — how we’ll have the humans play and work with these robots.”

This biweekly column by Canadian science and space journalist Elizabeth Howell focuses on a trending news topic in Canadian astronomy and space.