Convex Coverage Set Methods for Multi-Objective Collaborative Decision Making

My research is aimed at finding efficient coordination methods for multi-objective collaborative multi-agent decision theoretic planning. Key to coordinating efficiently in these settings is exploiting loose couplings between agents. We proposed two algorithms for the case in which the agents need to make a single collective decision: convex multi-objective variable elimination (CMOVE) and variable elimination linear support (VELS). While CMOVE deals with the multiple objectives on a single agent level, VELS deals with the multiple objectives on a joint decision level, leading to different trade-offs. We also proposed a naive way (approximate optimistic linear support (AOLS)) to apply the scheme of VELS to sequential settings, which does not yet fully exploit loose couplings. The next step in this line of research is to extend the distributed single-shot methods to distributed sequential settings, while better exploiting loose couplings.