MIT Herding System Furthers Robotic Collaboration



By Danielle Lucey



Researchers from the Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Laboratory will present a new system at a conference in May that stitches together existing control programs to allow multiple robots to collaborate in new and complex ways, according to a university press release. 



The programming will apply to all “multiagent systems,” meaning either swarms of robots or networks of devices could benefit from the technology. 



The system works by factoring in uncertainty, like the odds that communications will drop out, and then plans around those scenarios. Until now, multiagent systems have typically been restricted to scenarios where information about the environment is reliable or a fairly simple collaborative task can be planned in advance, says the release. 



Jon How, the Richard Cockburn Maclaurin Professor of Aeronautics and Astronautics, and his student Chris Maynor are currently testing the system in simulations that involve small swarms of iRobot Create platforms, based on the body of the popular Roomba vacuum cleaners. 



The pair will present their findings at the International Conference on Autonomous Agents and Multiagent Systems.