At our laboratory, we have developed a method to produce dynamic and stable walking trajectories for a humanoid robot, which includes the control of both the upper body and the legs, as well as the implementation of object transport. It is difficult to analytically determine such trajectories with a multiply linked robot. To deal with this, we use straight line approximation and incoherence with zero moment point (ZMP) derivation from the linked movements. Our proposed strategy can find a trajectory to satisfy a target ZMP with high speed calculations, and fully implement an online control system that can achieve both target upper-body and leg movements.