EAGER: First Order Decision Diagrams for Relational Markov Decision Processes

Tufts University, Medford MA

Investigators

Abstract

Decision making under uncertainty in dynamic, structured, and complex environments is a major area of research in AI. Research under this award aims to make novel contributions in this area by developing and expanding a new compact representation known as First Order Decision Diagrams (FODDs)and new learning and planning algorithms based on abstract Markov Decision Processes that have relational structure. The development of such structures, semantics and algorithms is a crucial step towards optimization and planning in complex real-world domains, such as emergency response, product delivery, and other service domains.