The present disclosure relates to the field of modular robot control, and more particularly to a method for controlling a modular robot and a system thereof. The method includes the following steps: T1: providing a plurality of module units; T2: assembling the plurality of module units into an initial entity structure; T3: acquiring initial virtual configuration information of the initial entity structure; T4: generating an initial virtual configuration based on the initial virtual configuration information; T5: setting an action frame to generate preset action control information; and T6: transmitting the preset action control information to the modular robot which executes a motion according to the preset action control information.