Power Indices for Team Reformation Planning Under Uncertainty

This work is an attempt at solving the problem of decentralized team formation and reformation under uncertainty with partial observability. We describe a model coined Team-POMDP, derived from the standard Dec-POMDP model, and we propose an approach based on the computation of team power indices using the Elo rating system to determine the most fitting team of agents in every situation. We couple this to a Monte-Carlo Tree Search algorithm to efficiently compute joint policies.