Multi-agent planning by distributed constraint satisfaction
Pedro Meseguer · DIGITAL.CSIC (Spanish National Research Council (CSIC)) · 2015
Multi-agent planning has received substantial attention in the last years. A few works have considered the translation into SAT terms, to be solved distributedly by an specialized SAT solver. Here, in the context of interleaving planning and coordination, we propose an approach that uses the ABT (asynchronous backtracking) algorithm to distributedly solve the propositional formulation of a multi-agent planning instance. This algorithm, presented for distributed constraint satisfaction, is generic and provides interesting privacy properties. We show some experimental results on some real-world instances coming from standard planning competitions