RSOME案例：两阶段分布鲁棒优化用于批量生产问题

RSOME: Robust Stochastic Optimization Made Easy

---

Adaptive Distributionally Robust Lot-Sizing

In this section, we are using RSOME to replicate numerical case studies presented in Bertsimas et al. (2021). A capacitated network with $n$ locations is considered, where each location $i$ has an unknown demand $d_i$, and the demand can be satisfied by the existing local stock $x_i$ or by transporting an amount $y_{ij}$ of units from another location $j$, which is determined after the demand is realized. Such a lot-sizing problem can be written as the following two-stage formulation,

$$\begin{array}{rlr}\min & \sum _{i=1}^N{c}_i{x}_i+\mathbb{E}\left[Q(x\text{x}x,\widetilde{d\text{d}d})\right],& \\ \text{s.t.}& 0\le x_i\le K,& i=1,2,...,n\end{array}$$

where the recourse problem $Q(x\text{x}x,d\text{d}d)$ is written as

Q ( x , z ) = min ⁡   ∑ i = 1 n ∑ j = 1 n c i j y i j s.t.  x i − ∑ j = 1 N y j i + ∑ j = 1 N y i j ≥ d i i = 1 , 2 , . . . , n 0 ≤ y ≤ b . \begin{aligned} Q(\pmb{x}, \pmb{z}) = \min~& \sum\limits_{i=1}^n\sum\limits_{j=1}^nc_{ij}y_{ij} &\\ \text{s.t.}~&x_i - \sum\limits_{j=1}^Ny_{ji} + \sum\limits_{j=1}^Ny_{ij} \geq d_i & i = 1, 2, ..., n \\ &\pmb{0} \leq \pmb{y} \leq \pmb{b}. \end{aligned} Q(xxx,zzz)=min s.t. ​i=1∑n​j=1∑n​c