Robots aren’t one-size-fits-every-situation. There isn’t one super machine that can complete any task or solve any problem, despite what Siri and Watson might lead you to believe. And sometimes, depending on the task, it’s much better to have a bunch of simple, inexpensive robots rather than one brilliant, expensive robot.
“If I want a robot that’s going to help a doctor with surgery, I want one that’s very precise and careful,” said BYU computer science professor Michael Goodrich. “But if I want a robot to clean up an oil spill, I’d rather have lots that are cheap and expendable so it wouldn’t be a problem if a few got gummed up in the process.”
Goodrich and a team of graduate students researched the logistics of getting these kinds of cheap robots to function well together in groups (known as swarms). Their recently published study in the Journal of Human-Robot Interaction details how humans can control these swarms so the individual robots will work cooperatively to perform tasks. Goodrich and his team looked to nature for inspiration.
“Nature has created some really cool designs,” Goodrich said. “You have flocks of birds and schools of fish that do things no individual could do by themselves. So there’s a real interest to see if we can do something similar with robots.”
Understanding how to control swarms is a critical step in actually using them to solve problems. Eventually, swarm robots could be used to encircle and clean up oil spills, detect cluster bombs in war zones, perform search and rescue missions, find victims in collapsed buildings, and team up with SWAT to surround a building and contain threats. But before this can happen, researchers need to establish the most effective way to lead and control the robots.
Essentially, humans would direct a group of these very simple robots from a remote location by giving the necessary instructions to one or a few of the robots within the group. These robots with more information become the leaders and are the only robots the humans can communicate with and location track.
Goodrich and his colleagues tested three different methods of leadership within this framework, each modeled after a relationship found in nature: predator, shepherd and stakeholder. The predator model was effective at having a leading robot disperse the swarm (as if they were being chased by a predator), while the shepherd model was effective at guiding the entire group in a single direction.
The stakeholder model differed because there were multiple leaders within the group that were then able to manage more sophisticated commands. Humans directing the stakeholders in the swarm could use this model to not only disperse and move the swarm, but arrange it into different shapes and patterns as well.
“Stakeholder is a good way to balance tradeoffs,” Goodrich said. “It’s not necessarily the best at dispersing and leading, but it can do those things as well as getting the swarm to create and maintain a shape or transition between shapes.”
Once Goodrich found the best way to lead a swarm, the next challenge was to figure out how many swarms a human could possibly control simultaneously.
“In any organization, I can’t have 37 people who directly report to me,” Goodrich said. “I don’t have the time, memory or capacity to manage all of those. It’s the same in the robotics world—I need to know how many things I can influence.”
This information is essential, because in application of the swarm technology, the person controlling the swarm won’t be able to see the entire swarm completing the task in a remote or classified location. They’ll likely only be able to observe signals and communicate with the stakeholder robots in that group.
The possibilities for the future of the technology are endless—but Goodrich sees it mainly for its ability to help others.
“Long term, I’d really love to see swarm robotics used for something humanitarian,” Goodrich said. “I’m really excited about applying swarm robotics to problems that have very positive social impact, like performing search and rescue or cleaning up an oil spill.”