12 September 2021

Swarms May Offer Next Level Artificial Intelligence

John Breeden II

Swarms of drones have gotten a lot of time in the spotlight lately, mostly for their use in potential military operations. The U.S. military is testing out swarm operations in simulations, while the British Army is using live drones operating in swarms during actual training operations. Other militaries are also interested in deploying swarms.

One of the biggest advantages a swarm of drones has when performing military operations is its resiliency. If a swarm enters combat and several individual drones get shot down or otherwise incapacitated, it really doesn’t reduce the combat effectiveness of the swarm, nor the tactics that it uses. A swarm of 550 drones is just about as powerful and flexible as a swarm of 600, even if the former has “lost” almost 10% of its initial strength.

And while that is noteworthy, it’s probably the least interesting aspect of swarms. What makes them really amazing in both military and civilian applications is their so-called swarm intelligence, a term first coined by Gerardo Beni and Jing Wang in 1989 when describing the potential for cellular robotic systems.

Swarm intelligence can enable drones to act very similar to how swarms of insects behave in nature. Take bees, for example. Individual bees each have their own intelligence and jobs within a colony. But when a bee finds something good, like a patch of blooming flowers, it will report that information back to other bees that happen to be nearby when it flies back to the colony. Interestingly enough, bees generally communicate the distance, direction and elevation of the discovered resource, as well as any dangers that might exist along the way. These are exactly the kinds of things that an artificial swarm of flying robots would need to know about a target or a potential point of interest. The bees that receive the message can then either go off to gather juicy pollen from those newly discovered flowers, or share the information with more nearby bees until it becomes well-known within the swarm.

What makes swarms of drones, and the intelligence that drives them, different from a single flying robot is that individual drones interact with both other drones and their environment. Then the swarm collectively makes decisions and pursues its mission based on different points of data that no single device could have provided. In this way, setting relatively simple instructions or overall goals for a swarm and then giving it freedom to operate can result in surprisingly complex behaviors that extend beyond even what a single AI could accomplish.

A Hive Mind For Weather Data

NASA wants to push the concept of swarm intelligence to new heights, adding it to a network of small weather satellites that could coordinate their actions and even change their flight paths and altitudes as needed to study weather events from multiple angles. The program is being worked on at the Goddard Space Flight Center in Greenbelt, Maryland.

The program wants to use swarms of small satellites, called SmallSats or CubeSats, in order to observe related weather phenomena occurring simultaneously around the planet and determine how they influence one another. In that way, multiple satellites working together as a swarm may be able to piece a weather puzzle together in a way that no single satellite ever could.

“We already know that Saharan dust blowing over to the Amazon rainforests affects cloud formation over the Atlantic Ocean during certain times of the year,” said NASA Engineer Sabrina Thompson, who works at the Goddard Space Flight Center. “How do you capture that cloud formation? How do you tell a swarm of satellites what region and time of day is the best to observe that phenomenon?”

According to Thompson, the answer is that climate scientists should create a set of parameters for high profile weather events and general rules that a swarm of satellites would follow when gathering weather data. Then it would be up to the swarm AI to act however it thinks is best to collect that information, coordinating multiple satellites working together to capture the most useful data as quickly and efficiently as possible.

According to NASA, different swarm designs are being considered, but in all cases the swarm AI would have a lot of control over how it deploys its individual members, and how it coordinates coverage of weather events happening around the globe. In one possible configuration, individual satellites could even use the drag force of the earth to drop to lower altitudes, which might give them a different angle on important events like the formation of rain or storm clouds.

NASA already has a prototype swarm satellite in orbit, the Hyper-Angular Rainbow Polarimeter (HARP) CubeSat, which was launched from the International Space Station last year. An updated version with more advanced instrumentation and a wider field of view, the HARP2, is scheduled for launch in 2023.

The HARP and HARP2 satellites were developed in conjunction with The University of Maryland – Baltimore County (UMBC). Professor Jose Vanderlei Martins, who worked on the project, said that operating in a swarm would allow CubeSats to collect better weather data, and do so much more quickly, as long as the swarm AI was given a lot of autonomy over its actions.

“Enabling the swarm to make decisions and share information is crucial,” Martins said. “These sorts of decisions need to be made in minutes. You don’t have time for ground control to be involved.”

In addition to speed, having a system of smart satellites working in a swarm would free up its reliance on ground control operations. That might allow satellite swarms to operate using much smaller budgets.

NASA is proving that better weather prediction and climate change analysis is yet another possible application for emerging swarm AI technology. It’s one area which continues to prove the old adage that two (or maybe more) heads are better than one. Except in the case of swarms, they are using multiple heads to merge back into a single hive mind that can likely accomplish things that no individual person or even AI ever could when operating alone.

No comments: