Swarm Robotics

swarm beesGrowing up, Sean Brigandi ’10 loved playing with Legos. So when the mechanical engineering student came home from college hauling a container stuffed with the plastic blocks and a popular programmable robotics kit, his parents were bemused.

“It was kind of a joke for my family because I was doing research and building with Legos and they asked, ‘Didn’t you do that as a kid?’” Brigandi said.

But Brigandi was using the blocks in bigger and more forward-thinking ways—building and programming small-scale robots and contributing to the emerging field of swarm robotics.

The study of swarm intelligence—a key concept in artificial intelligence research—has long been inspired by watching social animal behaviors such as foraging, fish schooling, bird flocking, and animal herding. After all, a solitary ant or bee can’t accomplish much, but a group or a swarm of them can.

Examining “this interplay helps us to better understand how simple individuals can work together to create and even manage seemingly complex and intelligent tasks,” said Yunfeng (Jennifer) Wang, PhD, associate professor of mechanical engineering.

In recent years, swarm research has expanded to include swarm robotics, examining how a team of similar robots can collaborate to complete a task.

Wang said swarm intelligence–based multi-robot systems have “great potential.” In the future, the military might deploy thousands of robots for guarding, patrolling, stalking, and attacking. In industry, they might be used for hazardous inspections, space exploration, assembly, and transportation.

Wang, who joined TCNJ’s faculty in 2002, has focused her research on robotics—specifically multi-robot systems, biological-inspired robots, interactive robots, and autonomous vehicles. Two years ago, she became increasingly interested in swarm robotics.

“The swarm intelligence–based multi-robots project is like fresh blood in my teaching and scholarship,” Wang said. It is especially new, she added, as a course of study at the undergraduate level.

Multi-robot systems and swarm theory

Multi-robot systems are typically made up of a bunch of simple robots. Building and using several simple robots can be easier, cheaper, more flexible, and more fault-tolerant than having a single powerful robot. Key technical challenges are communication, localization, and coordination between each robot; however, the study of swarms could provide the solution.

Swarm intelligence–based multi-robot systems could someday be used for tasks too complex—or impossible—for a single robot to accomplish, or to gain more effective performance. That might include environmental monitoring of large areas, searching a mineshaft or collapsed  building for survivors, or inspecting aircraft wings.

Major research area in the future

Wang, a native of Heilongjiang, China, has a doctorate in mechanical engineering from Johns Hopkins University and did her postdoctoral research in the Field and Space Robotics Laboratory at the Massachusetts Institute of Technology. Over the years, she has developed robotics-related activities through course work, undergraduate research projects, and scholarly work. Her work has been published in several top-tiered journals; she has presented at numerous international robotics conferences; and she served on the editorial board of Robotica, a prestigious scientific journal.

Brigandi, now a graduate student at Drexel University studying mechanical engineering, describes his former professor as “very knowledgeable” and a “great resource.”

“She has a hands-off approach, which allowed us to formulate our own research ideas while still providing guidance and keeping us moving in the right direction,” Brigandi said. “Her methods really helped us to grow as students and as independent thinkers.”

In a recent visit to the robotics laboratory, Wang admitted that when it comes to swarm robotics, she is learning alongside her students.

“The more I learn about this subject, the more fascinating I find it,” she said.

Setting the stage

swarm tuna
Wang's students built a system of robots that used a “flocking” algorithm inspired by the swimming behavior of a school of tuna to determine the robots’ movement.

Under Wang’s direction, Brigandi and Jennifer Field ’10 researched swarm robotics during the Mentored Undergraduate Summer Experience (MUSE) program in 2009. The students’ goal was to build four robots to move successfully as a team in an open environment. After reviewing the latest literature on swarm intelligence and swarm robotics, they built a system of four robots that used a “flocking” algorithm inspired by the swimming behavior of a school of tuna to determine the robots’ movement.

Much of the robotics research at TCNJ takes place in a windowless lab in Armstrong Hall. In addition to a whiteboard scrawled with math equations, electrical circuit diagrams, and drawings of different swarm formations, the lab is packed with various paraphernalia—Mylar balloons, pliers, sensors, soldering irons, LEDs, and wires of assorted sizes.

The students gave the robots names: “Wall-E” and “Ern-E,” in homage to the 2008 science fiction film in which humans have destroyed Earth, leaving the cleanup to an army of robots they’ve programmed. The other two robots are named “Cole” and “Chase,” after Philadelphia Phillies baseball players Cole Hamels and Chase Utley.

Nuts and bolts

The Lego Mindstorms NXT kits have many standard Lego pieces as well as more “robotic” parts, including a programmable “intelligent brick.” The brick is a microcomputer that has its own software and serves as the brain for the kits. The challenge is to manipulate the systems to complete tasks the original sets were not specifically designed to do.

Brigandi and Field began by connecting the intelligent bricks to computers to develop programs for the robot. The brick was then connected to the robot’s motors so the robot would move based on its programming, Field said. Later, they set up Bluetooth communication between the bricks and computer so everything was wireless.

Each robot had three sensors. A compass sensor told the robot’s orientation, a distance sensor how far a robot was from an object, and an infrared sensor allowed the robot to detect if an object was another robot or not, Brigandi said. The work wasn’t without problems. At one point, two bricks stopped working and had to be sent out for repair.

“I learned that you need to embrace frustration … because that feeling of  disappointment drove us to be more innovative and that much more determined to find a solution to our problems,” said Field, who is now pursuing a master’s degree in mechanical engineering at Stevens Institute of Technology. In the end, she added, the project was “a great success,” because the robots performed a flocking algorithm, “found” each other, marched forward together a predetermined distance and stopped—just as planned.

Making “recognizable contributions”

Wang, Brigandi, and Field presented a published paper on their work last July at the IEEE/ASME International Conference on Advanced Intelligent Mechatronics in Montreal, Canada. According to Wang, three to six experts review papers submitted to the conference, and the acceptance rate is only about 60 percent.

“This shows that our work has recognizable contributions that are agreed on by scholars in the field of robotics,” Wang said, noting that it is “rare for undergraduate students to have scholarly publications.” Bijan Sepahpour ’83, chair of the mechanical engineering department, said successful faculty-student partnerships such as this are proof that the “teacher-scholar” model is flourishing at the College.

“The fact that we do not have graduate programs (and…no graduate students) in engineering…has led to the enhancement of the research and design capabilities of our undergraduates working with the faculty,” Sepahpour said. “We are receiving positive feedback from both our alumni who have engaged in such activities and … [the employers] who are now enjoying the fruits of the experiences that these young men and women gained at TCNJ.”

Fine tuning the project

Brian Geuther ’11, a senior majoring in mechanical engineering with a minor in computer science, continued the research during the MUSE 2010 program. The goal of his work was to have the robots “scout out” an unknown area and continue moving as far north as possible, but the unknown area could contain obstacles that the robots had to avoid.

Since using Bluetooth meant the robots could only communicate with the three other robots, Geuther designed a new communication system that didn’t limit the number of robots that could
communicate. His system used infrared LEDs to passively communicate with the other robots. As the robot made “good choices,” its LED stayed on. If it made “bad choices,” it would turn it off.

“The infrared [was] the only method of identifying other robots, so if the LED was off, it [was] viewed as an obstacle [to] be avoided,” Geuther explained. “Good choices” in this system were based upon the robots’ bearing (how close to north it was moving), the proximity to obstacles, and whether or not there were other robots in formation with it.

Geuther and Wang will detail this work in a paper, tentatively titled “A Decentralized Flocking Algorithm,” which they plan to submit to the 2011 International Conference on Robotics and Automation.

Research continues

jennifer wang
Yunfeng (Jennifer) Wang, PhD, is leading the swarm robotics research at TCNJ. “Her methods really helped us to grow…as independent thinkers,” said one of her former students.

Geuther was so intrigued by swarm robotics that he has made it the focus of his senior research project. He and Eric Johnsrud ’11, also a mechanical engineering major, are designing a multi-robot system to scout and harvest.

Geuther said that when it comes to wind and solar energy, “there could be more efficient positions to gather it, depending upon things like weather  conditions. This could also apply to non-renewable resource gathering, such as rare earth metal mining, or oil spill clean up. Either of these problems could be solved with swarm intelligence.”

Scouting robots would be sensor orientated to find the most efficient point to collect resources while harvesting drones would have minimal sensors and be able to collect energy from the energy source, Geuther explained. He added that the “scouts” could send information to a single computer, and then that computer could send data to the harvesters, which could “make their own decision based upon the scouted areas.”

Back to the future

Brigandi said friends think it’s “pretty cool” when he tells them he has researched swarm robotics.

“One of my friends who is a political science major said ‘that is like 10 times cooler than what I’m doing now,’”  Brigandi said.

Swarm robotics may still be relatively young and growing, but Brigandi said researchers “are exploring so many possibilities” that the field is expanding rapidly. Further in the future, Brigandi said, really small robots might work in swarms inside a person to perform non-invasive surgery, fight off bacteria or a virus, or even deliver medication to a specific part of your body.

“In the not too distant future, you’re going to start reading about and seeing swarm intelligence in your everyday life,” Brigandi said.

Leave a reply

© The College of New Jersey. All Rights Reserved.