Teaching robots to be team players with nature


Newswise – Algal bloom, flock of birds, and swarm of insects. This collective behavior by individual organisms can provide a separate and collective benefit, such as improving reproductive opportunities for successful mating or providing security. Now, researchers have harnessed the self-organizing skills required to reap the benefits of natural swarms for robotic applications in artificial intelligence, computing, search and rescue, and more.

They published their method on August 3 at smart computing.

“Designing a set of rules, once implemented by a swarm of bots, that leads to a specific desirable behavior is particularly challenging,” said corresponding author Marco Dorigo, a professor in the Artificial Intelligence Laboratory, called IRIDIA, from the Université Libre de Bruxelles. Belgium. “The swarm behavior is not a single map with simple rules implemented by individual bots, but results from the complex interactions of many bots implementing the same set of rules.”

In other words, the bots must work together to achieve the overall goal of separate contributions. The problem, according to Dorigo and his co-authors Dr. Valentini and Professor Hamann, are that the traditional design of individual units to achieve a collective goal is bottom-up, requiring trial-and-error improvements that can be costly.

“To meet this challenge, we are proposing a new design approach that is global to local,” Dorigo said. “Our basic idea is to create a heterogeneous swarm using combinations of behaviorally different factors such that the resulting swarm behavior approximates user input that represents the behavior of the entire swarm.”

This configuration involves selecting individual agents with predefined behaviors that researchers know will work together to achieve the target group’s behavior. They lose out on the ability to program individual units locally, but according to Valentini, Hamann and Dorigo, the trade-off is well worth it. They cited the example of a monitoring mission, where a squadron might need to monitor a facility that requires more internal monitoring during the day and more external monitoring at night.

“The user provides a description of the desired swarm assignments as a probability distribution over the space of all possible swarm assignments—more factors indoors during the day, more factors outside at night or vice versa,” Valentini said.

The user will define the target behavior by changing the number and position of the distribution modes, with each mode corresponding to a specific allocation, such as 80% agents indoors, 20% outside during the day, 30% indoors, and 70% outside at night. This allows the squadron to change behavior periodically and independently, predetermined by group modes, as conditions change.

“While it is difficult to find the exact control rules for robots so that the swarm behaves as we desire, the desired swarm behavior can be obtained by combining different sets of control rules that we already understand,” Dorigo said. “Swarm behaviors can be engineered microscopically by mixing bots from different pre-defined rule sets.”

This isn’t the first time Dorigo has turned to nature to improve computer science methods. He previously developed an ant colony optimization algorithm, based on how ants move between their colonies and food sources, to solve difficult computing problems involving finding a good approximation of the optimal path on a graph. While Dorigo first proposed this approach to a relatively simple problem, it has since evolved as a way to address a variety of problems. Dorigo said he plans to take the swarm’s methodology in a similar direction.

“Our immediate next step is to validate our methodology across a larger set of swarm behaviors and beyond task assignment,” Dorigo said. “Our ultimate goal is to understand what makes this possible, and to formalize a general theory to allow researchers and engineers to design swarm behaviors without going through a painstaking trial and error process.”

This paper was co-authored by Gabriel Valentini, scientific collaborator at IRIDIA, and Heiko Hamann, professor at the Institute of Computer Engineering at the University of Lübeck, Germany.

This work was partially supported by the European Research Council and the Belgian Foundation for Scientific Research.


Leave a Reply

Your email address will not be published.