Abstract: Considering uncertainty of design parameters,the paper develops a two-stage stochastic programming model for optimal design of a single product,capacitatedreturnedlogistics network,which gives sets of finite and discrete values of stochastic parameters with the help of sampling technique,comp utes and compares construction and run expenses of different networks by mixed genetic algorithm,so effectively avoids network facility numbers and random vector dimensions' negative effect on model calculation efficiency.In order to get a robust returned logistics network,using a larger quantity of samples estimates the feasible networks resulting from time after time necessary optimizations.Taking into account the effect of sample randomness,It presents the design step s to acquire a realistic returned logistics network according to the proposed model.In addition,a useful example illustrates the model's validity,also verifies the fact that determinate model has no advantage to stochastic programming problem.

Key words: returned logistics network, stochastic programming, mixed genetic algorithm, sampling theory