A multiagent planning algorithm with joint actions

Abstract

In this paper, we consider multiagent planning with joint actions–that refer to the same action being performed concurrently by a group of agents. There are few works that study specification of joint actions by extending PDDL. Since there are no multiagent planners that can handle joint actions, we propose a multiagent planning algorithm, which is capable of handling joint actions. In a multiagent setting, each agent has a different capability. The proposed algorithm obtains the number of agents involved in a joint action based on the capability of the individual agents. We have implemented our algorithm and compared its efficiency with some state-of-the-art classical planners. The results show that when the problem size increases, our algorithm can solve such problems whereas it cannot be solved by the classical planners.