A method for robust design was presented based on sensitivity analysis with hybrid discrete variables, and two multi-objective robust optimization design models were established for robust optimization design based on adjunctive function of sensibility analysis. The models were solved with absolute degree of grey incidences. The models and method could reasonably deal with value adopting problems of hybrid discrete variables in optimization design. A chaos emigration operator was introduced to carry out improvement on the fundamental genetic algorithm, and the grey compound genetic algorithmic program GSCHGA for the multi-objective optimization of hybrid discrete variables was developed. Finally, the engineering design example shows that this algorithm has no special requirements on the characteristics of optimal designing problems, good universal adaptability, reliable operation of program, and strong ability of overall convergence. 