Bacterial trick keeps robots in sync

You don’t have to watch Dancing with the Stars to know that keeping in sync is tough — and it’s even tougher for a robot. A new approach keeps several robots in step, and even enables a dancing robot that loses its footing to seamlessly rejoin its synchronized peers.

One way to synchronize a group of robots is for each to communicate with one another about their positions, but distance between the robots can lead to time delays. And when many robots are involved, the complexity of this communication network grows. To skirt such problems, researchers from MIT have taken inspiration from bacteria that synchronize their behavior not by checking in with each other, but by checking in with their environment.

Synchronizing robots this way might work well in rescue operations where robots are damaged and need to be replaced, says Paola Flocchini, a distributed computing expert at the University of Ottawa in Canada.

Many bacteria coordinate via a process called quorum sensing, releasing a steady stream of signaling molecules into the environment and also sensing the signaling molecules. When enough bacteria are around that the local concentration of these molecules soars, it’s time for group action: Genes get turned on, molecular switches are flipped and the bacteria all change their behavior in sync.

Similarly, MIT’s Jean-Jacques Slotine and Patrick Bechon coordinated the behavior of eight dancing humanoid robots by having the bots send information to — and get information from — an external computer server. The work was posted May 14 on arXiv.org.

The robots go through cycles of prescribed actions, such as bobbing their heads, and send the  server information about where they are in these cycles. The server then sends the average of this information back to all the robots. So a robot joining its dancing peers will check in with the server about what the other robots are doing. It can then calculate what the next movement is in the synchronized cycle and rejoin the group. Information about the music — in the test case, Michael Jackson’s “Thriller” — is also embedded in the information sent back to the robots.

Incorporating math that describes the oscillating movements of body parts, such as arm and heads, is quite clever, says Mehran Mesbahi of the University of Washington in Seattle, whose research includes spacecraft navigation and control. It’s much harder to incorporate information on position, angles and music, he says, than to have a simple command such as “March.”