Andreas H. Hamel,Andreas Löhne
Abstract:Optimization problems with random data have a wide range of applications. A typical feature of many such problems is that some variables have to be optimized before certain random coefficients have been realized and for other variables it is sufficient to decide on them afterwards. This leads to a multiple stage decision process. To optimize the variables in the first of two subsequent stages the stochastic problem is transformed into a deterministic program, called the (deterministic equivalent of the) recourse problem. In case of stochastic linear programs with finitely distributed random data this non-stochastic substitute is just a linear program. In the same way a multiple objective linear program is obtained if the original problem has multiple objective functions. In the first of the two stages, a decision maker usually would chose a feasible point out of the set of all Pareto-optimal points. This choice however has consequences to later stage decisions. We claim that the decision process in the earlier of the two stages is not fully transparent if a classical multi-objective decision process is applied: in addition to the original objectives of the problem a decision maker may have a preference for largest possible flexibility in later stage decisions. This additional objective is taken into account if the recourse problem in case of multiple objectives is taken to be a polyhedral convex set optimization problem instead of a multi-objective linear program only. We also discuss several surrogate problems to the recourse problem such as the wait-and-see problem and the expected valued problem for the multi-objective case. The new approach based on set optimization is illustrated by an example, the multi-objective newsvendor problem.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to optimize multiple objective functions while maintaining decision - making flexibility in a multi - stage decision - making process in multi - objective stochastic linear programming. Specifically, the paper focuses on how to design a decision - making process in multi - stage stochastic linear programming when there are multiple objective functions, so that decision - makers can not only optimize according to the original objectives, but also retain as much decision - making flexibility as possible in subsequent stages.
### Background and Motivation
Multi - stage stochastic linear programming problems are widely used in various practical scenarios, such as inventory management and resource allocation. A typical feature of such problems is that some variables need to be optimized before the realization of random coefficients, while other variables can be determined after the realization of random coefficients. This leads to a multi - stage decision - making process. Traditional multi - objective decision - making methods have certain limitations when dealing with such problems, especially when decision - makers hope to retain more flexibility in the early stage to deal with different situations that may occur in the future.
### Main Contributions of the Paper
1. **Introduction of Set Optimization Framework**: The paper proposes a set - optimization - based method to deal with multi - objective stochastic linear programming problems. This method not only considers the original objective functions, but also takes into account the decision - maker's preference for decision - making flexibility in subsequent stages.
2. **Transparency of the Multi - stage Decision - making Process**: Traditional multi - objective decision - making methods may not be transparent enough in the decision - making process in the early stage because they do not fully consider the decision - maker's need for flexibility. By introducing the set - optimization framework, the paper improves the transparency of the decision - making process.
3. **Specific Application Examples**: The paper shows the application of the new method through a specific multi - objective newsvendor problem. In this example, the decision - maker not only wants to maximize the profit, but also wants to minimize the working hours, and at the same time hopes to retain more options when purchasing different types of newspapers.
### Mathematical Model
The mathematical models used in the paper mainly include:
- **Multi - stage Stochastic Linear Programming**: Formulated as a two - stage or multi - stage optimization problem, where the decision variables in the first stage need to be determined before the realization of random data, and the decision variables in the second stage are determined after the realization of random data.
- **Set Optimization Problem**: Transforms the multi - objective stochastic linear programming problem into a set optimization problem, where the value of the objective function is a convex set rather than a single value.
### Formula Representation
- **Multi - objective Stochastic Linear Programming Problem**:
\[
\text{minimize} \quad Cx + E[Qy] \quad \text{s.t.} \quad \begin{cases}
Ax = b \\
Tx + Wy = u \\
x, y \geq 0 \\
x \in \mathbb{R}^n, y \in (\mathbb{R}^m)^\Omega
\end{cases}
\]
where \(C \in \mathbb{R}^{d \times n}\), \(A \in \mathbb{R}^{k \times n}\), \(b \in \mathbb{R}^k\) are deterministic, and \(Q \in (\mathbb{R}^{d \times m})^\Omega\), \(T \in (\mathbb{R}^{\ell \times n})^\Omega\), \(W \in (\mathbb{R}^{\ell \times m})^\Omega\), \(u \in (\mathbb{R}^\ell)^\Omega\) are random.
- **Set Optimization Problem**:
\[
\text{minimize} \quad E[Z](x) \quad \text{s.t.} \quad \begin{cases}
Ax = b \\
x \geq 0 \\
x \in \mathbb{R}^n
\end{cases}
\]
where \(Z(x)=\{Cx + Qy\mid Tx\)