Biped type robot is more adaptable than the mobile type in various terrains. Therefore, recently it becomes an interesting topic for control researchers. Yet, the biped could topple easily, and it is difficult to control a biped robot for stable walking. Thus, the generation of the optimal walking trajectory is an important task for a biped robot. In this view, this paper proposes a method for generating stable walking trajectories that results the best performance. The proposed method uses multi-objective evolutionary algorithm, based on the known mechanical information of the biped robot. We formulate a trajectory generation as a multi-objective optimal problem, and obtain all Pareto-optimal solutions within the feasible solution region. The solution trajectories are verified for generating variable patterns by simulations.