Marching toward a better future

When it comes to engineering a solution to a problem, Ted Pavlic looks to a specific source for inspiration ­— nature.

Pavlic, who joined ASU’s Ira A. Fulton Schools of Engineering this year as an assistant professorPavlic is appointed jointly with the School of Computing, Informatics, and Decision Systems Engineering and the School of Sustainability., studies decision-making strategies for autonomous agents — like, for instance, a robot vacuuming the carpet.

“An effective vacuum cleaner must make decisions that are sensitive to its remaining charge, remaining space in its sweeper bin and the demographics of the kinds of tasks that are available for it,” PavlicPavlic is also the associate director for research at the Biomimicry Center at ASU, and he is formally affiliated with ASU’s BEYOND Center for Fundamental Concepts in Science as well as the Center for Social Dynamics and Complexity. Outside ASU, he serves as external faculty at the Human Computation Institute, a non-profit innovation center that advances the science of scalable crowd-power to tackle wicked societal problems. He earned a doctorate in electrical and computer engineering, with an emphasis on optimization and intelligent control from Ohio State University in 2010. said.

In other words, it needs to “think” like a human. Or an ant ... more on that in a bit.

“One day, it may make sense to concentrate on cleaning dirt around the baseboards,” he said. “Another day, baseboards might be ignored because of more valuable tasks that are plentiful in the middle of the floor. But the vacuum cleaner must make these decisions on-line in real time as it monitors its own state and discovers the state of the world around it.”

This is just one of many ways the solutions and ingenuity of the natural world can shape the development of engineering designs. Pavlic says that the goal of his research isn’t to simply translate these concepts, but to find inspiration from these biological solutions that can inform design contexts.

“Naturally evolved biological solutions are tailored for their ecological context, and often the human technological context has important mismatches even when analogies are used,” said Pavlic (pictured at left). “Furthermore, scientists often find the ingenious natural solution that was once thought to be operating in a natural scenario is actually entirely different than the actual solution being used.”

A good example of that is the passively cooled Eastgate Centre in Harare, Zimbabwe. The ventilation and cooling system in the shopping tower was inspired by an untested idea about what might be going on in towers in the termite mounds of Africa. Excited by the concept that African termites might have converged upon a general solution for building African towers, designers built the Eastgate Center to mimic this biological narrative.

But when researchers studied African termite mounds with modern technology, it was found that the actual natural story had nothing to do with the elaborate narrative once put forth.

“So a towering example of biomimicry turned into just another example of sustainably minded human ingenuity,” he said. “Nevertheless, that human ingenuity would not have been possible without first thinking about the termites, the problems that they have to solve, and the unique constraints imposed on them by their physiology and their ecology. The Eastgate Centre may not be biomimetic, and it might only be weakly bio-inspired, but it is certainly bio-informed. It is that kind of novel thinking that I try to catalyze by looking at natural problems through the eyes of an engineer.”

Insect inspiration

Termites aren’t the only insects with behavior worth modeling.

Several ant species have colonies that are able to regulate the colony-wide levels of protein and carbohydrate intake. This means, rather than bringing in whatever the nutrient mixture is in the background environment, individual ants are able to make coordinated decisions that reduce the intake of some food sources to compensate for having to increase the intake of other food sources that better meet the colony’s nutrient requirements.

“Each species of ant has a good solution to a particular problem that is modulated by details of its local environment,” Pavlic said.

How is this coordination across foraging individuals accomplished? 

“We certainly have similar multi-objective problems to solve in technological spaces,” Pavlic said. “For example, managers of the power grid must decide how to alter the state of one generator in order to better meet the increased demands of one area of the grid while recognizing that such a change may require reducing the supply from another generator.”

Currently, engineering solutions to similar problems are effectively centralized — either requiring a great deal of explicit coordination among distributed computing agents or requiring a central decision-maker. However, these ant colonies can reallocate their resources in a totally decentralized way with apparently very little communication and explicit coordination.

“Certainly there are many aspects of macronutrient regulation in ant colonies that would be foolish to mimic in the design of building power systems,” Pavlic said. “However, having a mathematical framework to pinpoint exactly what makes the decentralized implicit coordination possible in the ants gives insights on what things can be added to or taken away from conventional technological ways to solve these multi-objective resource allocation problems.”

One way to accomplish this might be to study the ants’ foraging habits.

“As we learn about what allows animals to work in highly effective groups, we can learn about how to make humans work more effectively in groups.”
— Ted Pavlic, assistant professor in ASU’s Ira A. Fulton Schools of Engineering

When given two food choices that differ in quality, some ant species will quickly determine the best of the two choices and allocate all foragers to it, but they will be ignorant of a third food item added that is of even better quality. Other species will allocate all foragers to the best choice, but they will do so more slowly and will be able to discover and reallocate foragers to a third better choice after it is introduced. And other species will not choose one food, but they will instead continuously match the distribution of foragers to the relative quality of the available foods.

By understanding the different recruitment mechanisms in these cases, researchers like Pavlic can inform the design of technologies that help to aggregate information from groups of humans and coordinate the group decisions.

“Consider a large online shopping website. If products sold on that website receive a large number of good reviews and products are sorted by how well they are reviewed, then it is unlikely that the top-selling product will ever be dislodged by a new and better competitor unless viewers sometimes make mistakes and look at products other than the best-rated,” he said.

The idea that “mistakes” allow for discovery of new options is how you differentiate between the first two classes of ants that commit to one choice or, in the second class, are able to find a later better choice. However, the third class of ants uses a very different recruiting scheme that is analogous to having reviews that disappear after a certain number of people have read them. If a similar review mechanism existed on an Internet shopping site, users would likely discover new products much more quickly and all products on the site would receive attention proportional to their relative merit.

“So by studying the differences in recruitment mechanisms across ants, we can propose different mechanisms for sharing information in human groups,” Pavlic said. “This idea could be enormously useful in the case of crowd computing, where anonymous individuals contribute to some group outcome that may not even be possible to observe at the individual level.”

Looking to the future

Pavlic notes that the work he does isn’t all about making automation smarter.

“As we learn about what allows animals to work in highly effective groups, we can learn about how to make humans work more effectively in groups,” he said.

Over the next few years, he plans to illuminate how behavioral analysis can be used in the design of sustainable, resilient automation systems across a wide range of applications. In doing so, Pavlic will help to grow the field of operations research to better address the needs of high-performance multi-agent robotic systems as well as decentralized algorithms for the sustainable built environment.

“I think that decision-making algorithms are everywhere, whether we designed them ourselves or they were discovered by the process of evolution under natural selection,” Pavlic said. “If we approach them with a common engineering perspective, we can express them with a common language that helps to illuminate similar problem structures even if the application space is apparently different. This common language helps to recognize both the similarities and differences in solutions for similar problems.

“We need to view nature as a catalog of diverse solutions. We can participate in adding to that diversity with our own solutions, but we should also learn and borrow whatever we can from existing solutions when possible. Through this process, not only can science enhance engineering, but engineering can find a way to participate in the scientific process and contribute more than just new technology.”