In order to minimize the mass of primary structure of a Long Slit Spectrometer (LSS), and to improve the robustness of structural response, the primary structure of the LSS was optimized based on robust optimization methods. The effects of the structure geometry tolerance and material property tolerance on the structure robustness were analyzed. Accordingly the structure geometry tolerance which has significant sensitivity to the structure robustness was adjusted. The robust optimization model of primary structure of the LSS with the constrained conditions of the fundamental frequency and the objective function of total mass was established. A Monte Carlo simulation and an ant colony optimization algorithm were applied to solve the optimization model. Experimental results show that the optimized structure has reduced the mass by about 17.5% compared to the original design. Although the mass of the optimal robust design is larger than the mass obtained by deterministic based optimization, it is insensitive to variations on design variables within a feasible region. The optimal robust structure design satisfies design requirements.