Carnegie Mellon Robotics Institute
I have a vision of some of society’s biggest, most complex, dangerous and boring tasks being performed by large heterogeneous teams consisting of intelligent software agents, autonomous robots and highly trained humans. Heterogeneous teams with thousands of intelligent agents processing data, planning, allocating resources and monitoring sensors, combined with thousands of specialized autonomous robots searching, transporting, rescuing and changing their environment, all under the control of a small group of highly trained human operators can fundamentally change how and what can be achieved. From military operations to agriculture to surveillance to disaster response, such teams have the potential to transform current practice in the relatively short term.
Much of the technology required for individual agents or robots has already been developed or is likely to be available soon. Advances in intelligent agents and artificial intelligence have made agents with an incredible breadth of capabilities possible. An important feature of autonomous robots is that once a single robot can be made effective and reliable, manufacturing processes will typically allow large numbers of identical robots to be made at low cost. Thus, we can think about applications where 100s or 1000s of such robots are used to achieve much bigger tasks than a single robot ever could. The heterogeneous teams of interest to my work will have the following characteristics:
Unfortunately, in most cases, it is not sufficient to simply make many copies of robot and spawn many agents and let them loose in a domain. Two general problems need to be overcome. First, the agents and robots must coordinate their actions, at minimum to avoid interfering with each other’s individual activities and at best to exploit synergies in their capabilities. Second, highly trained human operators are a scare resource, hence a relatively small number of human operators need to be able to provide oversight to the team to achieve their potentially ill-defined and dynamic objective. Notice that solutions to these two problems are not completely orthogonal, the approach the team uses to coordinate will impact the control operators have over the team. My work focuses on solving these two problems.
|Research Interest Keywords|
|artificial intelligence, constraint-directed reasoning, human-computer interaction, multi-agent systems|
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions