Q&A: How does ‘collective intelligence’ emerge among tiny robots?
Microrobots are engineered entities that contain very simple electronic circuits and propulsion machinery that allows them to push or drive forward. Individually, these robots can execute simple behaviors such as forward movements. Collectively, these microrobots can work together with seemingly intelligent behaviors to respond to signals, maintain and restore a larger shape or detect threats. How does intelligent-like behavior emerge in the system of simple interacting units?
That’s the big question posed by Igor S. Aronson, Huck Chair Professor of Biomedical Engineering, Chemistry and Mathematics. Aronson and his co-principal investigator Erwin Frey, professor of physics at Ludwin-Maximilians Universität in Munich, Germany, received a $1,058,000 grant from the John Templeton Foundation to advance this research by considering different populations of agents, how they evolve and the different characteristics of such agents. The research team is focused on understanding how simple physical interactions and signal exchange between microrobots lead to complex, or intelligent-like, behavior such as threat detection, disassembling and reorganizing on signal, and shape retention.
“More is different,” said Patrick Drew, interim director of Huck Institutes of the Life Sciences and professor of engineering science and mechanics, of neurosurgery, of biology and of biomedical engineering. “Emergent phenomenon can arise from bringing together many interactive elements. In a very real sense, this is what life is. Understanding how these emergent dynamics can imbue intelligence into a collection of ‘dumb’ elements is a key problem in interdisciplinary science. I am delighted that Dr. Aronson and his team are getting support to work on this important area, and I am excited to see what they do next.”
The research team is expanding their work by creating computational models aimed at elucidating how self-sufficient, simply designed robots can collectively perform intelligent-like behaviors. Such concepts could be potentially useful to accelerate tissue healing and regeneration, among other applications, Aronson said.
Penn State News spoke with Aronson, who also co-directs the Center for Mathematics of Living and Mimetic Matter, about the concept of collective intelligence and how understanding it might inform self-sufficient, intelligent platforms to synchronize microrobots.
Q: What kinds of methods will be developed in this project and what is the purpose?
Aronson: We will develop computational and theoretical methods to understand collective interactions with the purpose of understanding the agents’ emergent behavior. We will combine these methods with numerical analytical methods, and we will use clusters of graphical processing units, or GPUs, for our large-scale computations. Our computational models are used to describe the behavior of these microrobots — how the mircorobots interact and what kind of emergent behavior they exhibit.
Each of these models is based on a collection of agents — a select group of microrobots — that serves a simple mission, such as “move forward.” If nothing is in the agent’s environment, the microrobots will move straight. However, if a neighbor is present or there are certain signals that present louder and surpass a threshold limit, the agents will respond and form structures or shapes such as snakes; larvae; volvox, which are spherical shapes similar to algae; or oroboros – a fairy tale character showing a snake eating its tail and react accordingly to such signals. These structures self-organize and emerge as part of aligning to react towards strong signals. Collectively, they then exhibit intelligent-like functionalities such as threat detection; directed migration such as moving forward; disassembling and reorganizing as well as shape retention. If there are multiple signals, the microrobots will self-organize and react to the loudest or most immediately pressing. Each robot detects what is in its environment and acts fully independently. The structures can also move at different speeds and release unique sound signatures due to the interactions they have with their environment.
Q: You have researched collective intelligence for two years. How will this grant build on what you already understand from your prior work?
Aronson: We are expanding this research from our preliminary findings from 2022. When we originally explored this model, we didn’t expect these complex behaviors to come from such simple elements. We anticipated something much simpler. This grant will enable a much deeper understanding of these phenomena while understanding the basic mechanisms or simple physical interactions leading to self-organized intelligent-like behavior. In the future, we want to further expand this framework to make the elements, or microrobots, more complex and have slightly different thresholds. We want to make them evolve by changing the parameters depending on the environment, which could manifest more intelligent, functional behaviors.
Q: How can this be applied to other disciplines or have a broader impact on various communities?
Aronson: There are many possible applications. We plan on working with other researchers to implement the work in biological systems, helping accelerate tissue and organ formation or reassembly, by having individual cell components working together on a much faster time scale. It also simplifies developing and producing swarm robotics for use in targeted drug delivery. An individual robot cannot perform sophisticated tasks and solve complex navigation problems. Swarms of simple robots could solve much harder tasks, like passing through a constriction, detecting a threat and various other tasks while increasing efficiency and lowering costs.
We want to expand this framework and make the elements more complex and more intelligent. We aim to eventually gain funding to further develop these robots to better control how they function, manifesting different and more complex intelligent behaviors.