The modular robot can be reconstructed into different topologies and can meet complex and changeable task requirements. How to find the optimal topology of a modular robot with specific working capabilities is the key to fully appropriate its reconfiguration characteristics. To solve this problem， this paper proposes a modular robot topology optimization method based on grey wolf optimizer （GWO）. First， the kinematic chain representation is constructed for the modular robot， and the kinematic/dynamic modeling is completed based on the screw method and the Newton-Euler method. Second， the topology decision variables are designed based on the representation of the motion chain， and the objective function is designed considering the number of modules， the maximum joint driving torque and dexterity of the robot. The concepts of crossover and mutation are introduced to improve the GWO， and the topology optimization model is established and solved. Finally， the optimal topologies corresponding to two experiments are compared to verify that the algorithm can effectively solve the optimal topology of a modular robot.