A Belief-Space Approach to Integrated Intelligence - Research Area 10.3: Intelligent Networks
Technical Report,01 Aug 2014,31 Jul 2017
Massachusetts Institute of Technology (MIT) Cambridge United States
Pagination or Media Count:
The overall goal of this work is to develop the perception, estimation, planning, and control techniques necessary to enable autonomous agents to perform robustly and intelligently in complex uncertain domains. This includes the ability to intelligently interact and coordinate with humans or other agents so as to achieve goals effectively and efficiently.Specifically our goal has been to develop algorithms for effective and efficient planning in domains that are characterized by a variety of action types, each with continuous parameters, for example the variety of manipulation actions available to a robot picking, placing, pushing, tilting, etc.. Furthermore, the actions take place in the presence of uncertainty both as to the current state of the world and as to the actual result of the action.