Researchers at the Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory say they have developed a new way of modeling robot collaboration.
The researchers say the approach reduces the need for communication by collaborating robots by 60 percent.
MIT professor Julie Shah and graduate student Vaibhav Unhelkar have developed a system that ignores uncertainty about the effectiveness of the actions of agents in a multiagent system. Instead, the system assumes whatever an agent attempts to do, it will do.
The team tested the system on more than 300 computer simulations of rescue tasks in unfamiliar environments. "What I’d be willing to bet, although we have to wait until we do the human-subject experiments, is that the human-robot team will fail miserably if the system is just telling the person all sorts of spurious information all the time," Shah says.
The researchers also believe their model could make it easier to design systems that enable humans and robots to work together.
"This work has applications outside of multi-agent systems, reaching into the critical area of human-agent collaboration, where communication can be costly, but more importantly, human team members are quickly overloaded if presented with too much information," says University of Melbourne professor Tim Miller.
From MIT News
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found