Wisdom of Swarm

MAHTA EMRANI | August 29, 2016

What ants, pigeons, and bacteria can teach us about collective intelligence

The Swarm. Source: iStock

The Swarm. Source: iStock

Researchers at the USC Viterbi School of Engineering believe that some of the most complex technologies of our times, such as robotics or cognitive computing, can learn a thing or two from swarms of ants, pigeons and bacteria.

Let’s look at ants for example. Drop a piece of bread on the floor and wait. Sooner or later, one or two straggling ants scoping the area will stumble upon it. What happens next is perhaps one of the most elaborate orchestrations of organized teamwork in nature.

Once the food source is detected, more and more ants begin to march towards it. At first, the patterns of their movements may appear random and erratic. However, as each ant gathers and exchanges more information with its neighbors, their collective intelligence grows. They begin to orient and self-organize. They become more stable. They form a swarm.

A swarm represents a group of agents that interact with one another and exchange information between themselves. Their goal is to get better organized and become collectively smarter and more effective at performing their group tasks.

This is what Hana Koorehdavoudi, a USC Viterbi Ph.D. candidate, has studied for the past three years. Under the mentorship of her advisor, Assistant Professor Paul Bogdan of USC Viterbi’s Ming Hsieh Department of Electrical Engineering, Koorehdavoudi has researched ants, pigeons and bacteria in order to develop a series of algorithms that can actually quantify the degree of complexity within swarms.

Koorehdavoudi’s calculations measure the interactions within complex systems. These calculations are unprecedented because they enable scientists to identify and evaluate the types of exchanges that result in certain forms of collective behavior. This could help scientists engineer specific outcomes by simply tweaking the interactions that exist within a network.

The inspiration behind Bogdan and Koorhedavoudi’s unique approach was the “energy landscape.” The energy landscape represents all the possible dynamic formations of agents within a swarm.

“The energy landscape helps provide an understanding of how the dynamics of the swarm evolves through time,” Koorehdavoudi said. “This lets us identify and extract how the agents relocate themselves with respect to others in the system.”

In addition to the energy landscape, the team also used the concepts of emergence, self-organization and complexity to describe the evolution of the swarm:

  • Emergence occurs when a more complex system is created as a result of the growing number of interactions or interdependencies between individual agents. Imagine the ant swarm described earlier.
  • Self-organization goes hand-in-hand with collective intelligence. The more self-organized a system, the higher its collective intelligence. This is because self-organization emerges through increased interdependencies and interactions, which is fundamentally how information is shared across networks.
  • The last construct, complexity, is the product of emergence and self-organization together. Higher emergence and self-organization mean a more complex system.

Bogdan and Koorehdavoudi believe that these algorithms can have a significant impact on scientists’ efforts to understand, optimize, and control new complex networks. Some of the more obvious applications are advancing research and development in the areas of robotics, urban-planning, or even cancer treatment, among many others.

For instance, the algorithms could contribute to the design of autonomous machines that collaborate. This could improve their performance by helping them adapt, operate and evolve as part of a group. Imagine an army of drones at a disaster area collaborating together in order to maximize their rescue efforts.

Beyond robotics, Bogdan and Koorehdavoudi’s research could also help solve everyday problems such as traffic. Their work might even inspire new approaches for controlling the spread of cancerous cells. Both of these scenarios involve individual agents that interact and move together (i.e., automobiles and cancer cells).

In case of traffic, optimizing the interactions between the automobiles and raising their collective intelligence could potentially reduce congestions and shorten travel times. Similarly, doctoring the interdependencies of cancer cells could hypothetically slow or even diffuse their chance of growth or metastasizing.

Ultimately, the whole phenomenon as explored by Bogdan and Koorehdavoudi is very universal. “You can also take these formulas and apply them to the brain and see how brain organizes a thought,” said Bogdan. “You can model the thinking process.”

Indeed, the possibilities are endless, and all due to the wisdom of crowds – in this case, the humble ants, pigeons and bacteria.

For more information on Koorehdavoudi and Bogdan’s research, find their article in nature.com.