This paper describes a novel algorithm for activity coordination in multiagent systems that combines joint planning and joint learning. The basic idea underlying this algorithm is to combine the advantages of planning and learning, while at the same time avoiding their disadvantages.