A bottom-up approach to programming swarms of robots can be achieved by taking inspiration from complexity science, allowing the robots to interact with one another and learn and adapt their behaviors based on the environment.
The combination of leveraging existing biological principles, such as copying searching behaviors observed in ants, and applying mathematical scaling laws enables successful programming and scaling of robot swarms.
Deep dives
Programming swarms of robots using complexity science
In this podcast episode, Melanie Moses, a computer science professor, discusses how a bottom-up approach to programming swarms of robots can be achieved by taking inspiration from complexity science. By observing the behavior of foraging ants, Melanie and her team built a swarm of robots called IANTS that could collectively identify and gather resources. Using evolutionary algorithms, the robots were able to learn and adapt their behaviors based on the environment, such as when to return to a location, when to communicate with other robots, and how to efficiently collect resources. This approach demonstrated the possibility of engineering emergent properties in robot swarms and achieving cooperation to accomplish a common goal.
Fractal branching network for scalable robot swarms
As the team explored scalability in robot swarms, they encountered challenges with congestion and inefficiency as the swarm size increased. Inspired by scaling laws observed in biology, particularly the three-quarter scaling power law, they designed a fractal branching network for the robots. The network allowed for efficient transport of resources by implementing a bucket brigade-like system, where robots would pick up resources and pass them along to the next robot. By organizing the swarm using this network structure, they found that efficiency was maintained regardless of the swarm size, making it possible to scale up the swarm to thousands of robots if needed.
Leveraging existing biological and mathematical principles
To address the challenges of programming robot swarms and achieving scalability, Melanie and her team drew inspiration from both biology and mathematics. They copied the searching behaviors observed in ants and used evolutionary algorithms to evolve the desired behaviors in the robots. Additionally, they applied mathematical scaling laws to overcome limitations in congestion and transport efficiency as the swarm size increased. This combination of leveraging existing biological principles, such as pheromones and observation of local density, along with mathematical principles, resulted in a successful approach to programming and scaling robot swarms.
Imagine you were going to Mars with a swarm of robots, and you needed to send those robots out foraging. How would you program them? A traditional top-down approach to programming would mean programming what every single robot is going to do, and that's going to get complicated fast.
So in this episode, we're joined by Melanie Moses, Professor of Computer Science at the University of New Mexico, and External Faculty at the Santa Fe Institute. Melanie is going to explain how you can take lessons from complexity science, and utilise a bottom-up approach to programming a swarm. In other words, she's going to explain how you can program the robots to interact with one another. And if you thought you'd heard the end of scaling or power laws, then you're in for a surprise, because Melanie is going to share how scaling fits in with her work on getting robots to work as a team.