Efficient planning under uncertainty with incremental refinement
In Proc., (UAI-2019), 22.7.-25.7.2019, Tel Aviv, AAAI Press, 2019.
Online planning under uncertainty on robots and similar agents has very strict performance requirements in order to achieve reasonable behavior in complex domains with limited re- sources. The underlying process of decision- making and information gathering is correctly modeled by POMDP’s, but their complexity makes many interesting and challenging prob- lems virtually intractable. We address this is- sue by introducing a method to estimate rel- evance values for elements of a planning do- main, that allow an agent to focus on promis- ing features. This approach reduces the ef- fective dimensionality of problems, allowing an agent to plan faster and collect higher re- wards. Experimental validation was performed on two challenging POMDP’s that resemble real-world robotic task planning, where it is crucial to interleave planning and acting in an efficient manner.