MAPJA: Multi-agent planning with joint actions

Abstract

In this paper, we suggest a domain independent approach, MAPJA, to solve multi-agent planning problems with cooperative goals involving joint actions. We consider the capability of agents where capability is represented using a numeric value. The state-of-the-art multi-agent planners cannot handle joint actions. We have implemented and evaluated MAPJA on some benchmark planning domains and the experimental results are quite promising. We have also compared the performance of MAPJA with an existing approach, that transforms multi-agent planning problems into single-agent (classical) planning problems. The implementation results show that MAPJA outperforms the existing approach.