Topology optimization is proven significant for improving the performance of mechanical systems. However, when it is utilized for the optimal design of industrial robots, there are still many challenges such as complicated variable configurations of robots and computational expensiveness. Therefore, this paper presents an integrated optimal design method for lightweight serial robots by the use of part-level topology optimization and parametric system optimization. Firstly, the finite element analysis (FEA) and virtual joint method (VJM) are combined to construct the stiffness model of the robot, based on which deformations of the end-effector (EE) are analyzed. Then, typical robot configurations and load conditions are determined for the optimizations. Secondly, topology optimizations of main structure components i.e. the part-level optimizations are implemented with TOSCA software, and the stiffness-mass metamodels i.e. relationships between the stiffness and mass of these components are constructed. Finally, the system optimization of the robot is obtained by determining the mass division into different components with the robot stiffness model and metamodels. The optimal design of a serial painting robot is implemented to demonstrate the effectiveness of the proposed method.