In this paper, we propose an automatic motion generation system for a humanoid-like robot based on Interactive Genetic Algorithm(IGA). The proposed system can generate various movements for ten kinds of emotions, i.e., joy, anger, sadness, fear, shame, like, dislike, irritate, easy and surprise. Because of employment of IGA, users' burden is low. Laban features are also used for effective learning. We performed two kinds of evaluation experiments. One is evaluation of generated movements, second one concerns effectiveness of usage of Laban features. In addition, clustering analysis is carried out to find various findings for movements.