There are many multi-objective and multi-constrained robust optimization problems in engineering design field, and their inputs or parameters are often uncertain. The two-level robust optimization is widely used in the field of robust optimization to obtain the optimal solutions which are less sensitive to uncertainties. Since most complex engineering design problems rely on time-consuming simulation calculations, the two-level robust optimization may become computationally expensive. In order to solve this problem, an incremental Kriging multi-objective robust optimization algorithm based on the fast Kriging surrogate model(IKA-MORO) is proposed in this paper. Firstly, we introduce an incremental learning based fast Kriging modeling approach. Combining the Kriging models, we propose a model management strategy for the two-level robust optimization framework, and then, a new infill criterion to improve the approximate accuracy of the surrogate models is proposed. The results of the benchmarks show that IKA-MORO can effectively reduce the computational cost for the most of the complex multi-objective robust optimization problems.