The Reduced Degree-of-freedom (RDOF) concept is introduced in optimizing the modular reconfigurable robot configurations for specific task requirement. With fewer modules and degrees of freedom (DOFs), a modular robot would have a simple configuration, high loading capacity, and low power consumption rate, so that it can perform the task effectively. In this article, we formulate RDOF problem as a design optimization problem. A weighted sum of the number and types of modules is chosen as the objective function. Several task-related kinematic performance measures are considered as design constraints. Based on the problem specific coding scheme, the evolutionary algorithm (EA) approach is employed to search the optimal solutions. A virtual module concept is also introduced in the coding scheme to guarantee the operation of EA. The effectiveness of this approach is demonstrated by an example.