The robust optimization method has progressively become a research hot spot as a valuable means for dealing with parameter uncertainty in optimization problems. Based on the asymmetric cost consensus model, this paper considers the uncertainties of the experts’ unit adjustment costs under the background of group decision making. At the same time, four uncertain level parameters are introduced. For three types of minimum cost consensus models with direction restrictions, including MCCM-DC,\(\varepsilon \)-MCCM-DC and threshold-based (TB)-MCCM-DC, the robust cost consensus models corresponding to four types of uncertainty sets (Box set, Ellipsoid set, Polyhedron set and Interval-Polyhedron set) are established. Sensitivity analysis is carried out under different parameter conditions to determine the robustness of the solutions obtained from robust optimization models. The robust optimization models are then compared to the minimum cost models for consensus. The example results show that the Interval-Polyhedron set’s robust models have the smallest total costs and strongest robustness. Decision makers can choose the combination of uncertainty sets and uncertain levels according to their risk preferences to minimize the total cost. Finally, in order to reduce the conservatism of the classical robust optimization method, the pricing information of the new product MACUBE 550 is used to build a data-driven robust optimization model. Ellipsoid uncertainty set is proved to better trade-off the average performance and robust performance through different measurement indicators. Therefore, the uncertainty set can be selected according to the needs of the group.

鲁棒优化方法作为处理优化问题中参数不确定性的一种有价值的手段，逐渐成为研究热点。本文基于非对称成本共识模型，考虑了群体决策背景下专家单位调整成本的不确定性。同时引入了四个不确定的水平参数。对于三类有方向限制的最小成本共识模型，包括MCCM-DC，\(\varepsilon\)-MCCM-DC和基于阈值（TB）-MCCM-DC，建立了对应四种不确定性集（Box set、Ellipsoid set、Polyhedron set和Interval-Polyhedron set）的稳健成本共识模型。在不同的参数条件下进行敏感性分析，以确定从稳健优化模型获得的解的稳健性。然后将稳健的优化模型与达成共识的最低成本模型进行比较。示例结果表明，Interval-Polyhedron 集的稳健模型具有最小的总成本和最强的稳健性。决策者可以根据自己的风险偏好选择不确定性集和不确定性水平的组合，以最小化总成本。最后，为了降低经典鲁棒优化方法的保守性，新产品 MACUBE 550 的定价信息用于构建数据驱动的稳健优化模型。椭球不确定性集通过不同的测量指标被证明可以更好地权衡平均性能和鲁棒性能。因此，可以根据组的需要选择不确定性集。