Hungry lions on the savanna or porpoises in the sea work as teams to catch prey. Schooling fish dart in unison to escape a predator. Even “Meerkat Manor” depicts complex groups with clearly defined duties. What do these complex activities all have in common? For Stephen Pratt, assistant professor in Arizona State University’s School of Life Sciences, they represent different aspects of the hunt, but in ways that most people could barely begin to imagine.
ASU and Pratt hosted engineers, computer scientists, biologists and social scientists at a recent workshop – Heterogeneous Unmanned Networked Teams (HUNT) – that focused on developing bio-inspired solutions to engineering problems.
The workshop is part of a five-year project of the same name funded by the Office of Naval Research (ONR). The effort is led by 10 engineers and computer scientists from the University of Pennsylvania, University of California, Berkeley, and the Georgia Institute of Technology, in addition to ASU’s Pratt, who is the sole biologist in the group.
Why would naval research and other engineering research institutions look to nature for bio-inspired solutions? “Robustness, scalability, and the ability to function without complex central control are things that are really desirable in an artificial system,” Pratt points out. “All kinds of natural systems have them; from the movement of fish in schools and birds in flocks to social insects building specific, complex nest structures.”
“One of ONR’s long-term grand challenges is how to deal with the interaction of large numbers of fairly sophisticated autonomous vehicles – flying drones, vehicles underwater or on land,” Pratt explains. “All kinds of increasingly diverse and complex artificial systems will have to interact with each other and with humans.”
Military applications present particular challenges because of the large numbers of people, machines and unpredictable situations.
However, the engineering product is more likely to be a focus towards the end of the project, Pratt says. In the meantime, the workshop was all about thinking outside the box and creating an atmosphere for a rapid exchange of ideas.
Some of the more than 20 consultants attached to the project, spanning the fields of engineering, social sciences and biology, joined the project leaders at the workshop held March 9-10 at ASU. The goal was to pair up particular complimentary problems from biology and engineering.
For example, James Rehg and Tucker Balch of Georgia Tech are using their expertise to design automated computer vision systems to track films of animals in nature – namely, lion teams hunting – and be able to recognize and classify the behavior of each individual. This will enable African savanna ecologist, Craig Parker of the University of Minnesota, to get large quantities of data efficiently and empower him to ask questions about the coordination of group behavior over long time periods. Previously, this could only be done by painstakingly analyzing months of footage.
“Rehg, Balch and their research groups are really interested in doing this kind of thing – partly because it helps them study their own artificial collective systems by testing the ability of their system to analyze complex biological data,” Pratt says with a knowing smile. “Also, they like the idea of a really challenging computer vision problem; they get to do something new and different with our data.”
Another pair of researchers, Lori Marino of Emory University, an expert on porpoise teamwork, and Magnus Egerstedt of Georgia Tech, a robotics engineer, are delving into what porpoises do when they encircle prey. Porpoises are of particular interest because they are very good at switching team strategies quickly to meet changing circumstances. Egerstedt hopes to develop control algorithms for artificial systems that can effect the same sorts of rapid functional changes. Eventually these algorithms could be used to control groups of underwater or fleets of flying vehicles.
Pratt’s first taste of this “HUNTsian” type of collaborative endeavor came when he was asked to be a part of SWARMS (Scalable sWarms of Autonomous Robots and Mobile Sensors; www.swarms.org); an invitation based on his innovative research on collective decision making in acorn ants. Pratt clarifies, “SWARMS was directed at what could be learned from natural systems with many animals; where all the individuals are simple – maybe, even a little bit stupid. Imagine a swarm of army ants.”
The naval research project steps up the complexity in problem-solving, focusing on smaller groups of organisms, each with a unique position and attribute. “The HUNT project leaders originally thought exclusively about vertebrates,” Pratt says. “But, I convinced them that ants carry out tasks which are much more sophisticated than they had initially thought.”
Picture this, Pratt says: “A single ant finds a tasty fig piece (left by a scientist) in an open desert patch. Alone, she cannot rescue the piece from a larger competitor; but wait, five sisters arrive, each picks up a side and together they negotiate this morsel – many times larger than themselves – over a tough terrain and back to the nest. Dinner is served. In order to accomplish this task, the ants must work as a team, each one with a unique job: guiding, spinning or doing the heavy lifting.”
Pratt teams up with Vijay Kumar, who is with the University of Pennsylvania, to address this issue of collective retrieval: a big problem in collective robotics. “I’m getting the biological data; they’re devising tools that I would never be able to make. For example, force sensors that we try to get the ants to carry with them,” says Pratt.
“What I take away is a much more detailed description and measurement of what these insects are actually doing. What they get is inspiration for programming their robots to do the parallel task,” he says.
In some ways, the HUNT teams are analogous to their project, with each member playing a unique and essential role. Alone, each individual might be slow or lack efficiency in carrying the proverbial fig, but by combining their distinctive expertise, together they make one extraordinary, bio-inspired problem-solving team.