A Method for behavior planning is presented for a modular robot that applies a randomized planner to coherent structure of the robot To cope with difficulty in planning of many degrees of freedom (DOFs) of a modular robot, coherent structure is introduced in terms of control system and robot configuration. As the control system, a simple phase synchronization mechanism is introduced that can control die robot with many DOFs with reduced number of parameters. Together with symmetrical configuration, this control system generates various dynamic motions. In order to plan the behaviors of the modular robot determined by the parameters of the control system, we adopt a randomized planner called rapidly exploring random trees (RRTs). This can benefit from a number of advantages of RRT planner, including simple implementation, uniform search, and applicability to a dynamic system with differential constraints. By exploring parameter space of the coherent control system, behaviors including dynamic motions can be planned. We have applied the proposed planner to M-TRAN modular robot to demonstrate the effectiveness of the proposed method through preliminary simulation results.