Daniel BuenoControlling a fleet of robots so that they not only act individually, but also as a group, autonomously, is one of the goals of researchers at the Massachusetts Institute of Technology (MIT) and the University of Southern California (USC). It is a new system that connects existing control programs so that multi-agent systems with multiple robots can collaborate in very complex situations. Collaboration must occur in uncertain situations like when a communication link fails or a particular algorithm inadvertently leads a robot to a dead end. The system will be used to automatically correct these situations and ensure the combination of programs can produce the best results possible, given the uncertainty in the environment. Researchers at the MIT Artificial Intelligence and Computer Science Laboratory are testing the new system in a storage application, in which robot teams recover arbitrary objects in unspecified locations, simulating a heavy cargo transport situation. They want the system to govern the robots’ behavior and accumulate statistical data.