In this paper we describe a distributed approach to the block stacking problem. Each block is assumed to be an autonomous mobile agent. Motion conflicts among the agents are resolved by means of ( 1 ) motivation calculation which is based on individual goal attainment request, and ( 2 ) promising destination determination by ANN-based behavior function which is trained using an evolutionary programming technique. The proposed method eliminates the necessity for a central planner and exhibits good scalability. Some experiments are carried out and the mechanism for cooperative goal attainment by multiagents is examined.