Abstract:This paper presents an interactive approach based on a discrete differential evolution algorithm to solve a class of integer bilevel programming problems, in which integer decision variables are controlled by an upper-level decision maker and real-value or continuous decision variables are controlled by a lower-level decision maker. Using the Karush--Kuhn–Tucker optimality conditions in the lower-level programming, the original discrete bilevel formulation can be converted into a discrete single-level nonlinear programming problem with the complementarity constraints, and then the smoothing technique is applied to deal with the complementarity constraints. Finally, a discrete single-level nonlinear programming problem is obtained, and solved by an interactive approach. In each iteration, for each given upper-level discrete variable, a system of nonlinear equations including the lower-level variables and Lagrange multipliers is solved first, and then a discrete nonlinear programming problem only with inequality constraints is handled by using a discrete differential evolution algorithm. Simulation results show the effectiveness of the proposed approach.
automation & control systems,computer science, theory & methods,operations research & management science
What problem does this paper attempt to address?
This paper attempts to solve a class of Integer Bilevel Programming Problems. The characteristic of this type of problems is that in the bilevel decision - making structure, the upper - level decision - maker controls integer decision variables, while the lower - level decision - maker controls real or continuous decision variables. Specifically, the objective of the paper is to develop an interactive method based on the discrete differential evolution algorithm to solve this type of problems.
### Main Contributions of the Paper
1. **Problem Transformation**:
- Using the Karush - Kuhn - Tucker (KKT) optimality conditions, the original bilevel programming problem is transformed into a single - level nonlinear programming problem with complementary constraints.
- Applying smoothing techniques to deal with complementary constraints, and finally obtaining a single - level mixed - integer programming problem.
2. **Algorithm Design**:
- An interactive method based on the Discrete Differential Evolution (DDE) algorithm is proposed to solve the above single - level programming problem.
- In each iteration, for a given upper - level integer variable, first solve the nonlinear equation system containing the lower - level variables and Lagrange multipliers, and then use the DDE algorithm to solve the integer programming problem with only inequality constraints.
3. **Experimental Verification**:
- The effectiveness and feasibility of the proposed algorithm are verified by testing 15 integer bilevel programming problems. These problems include instances selected from the existing literature and are appropriately modified to fit the framework of integer bilevel programming problems.
### Mathematical Model
The integer bilevel programming problem in the paper can be expressed as:
\[
\min_{x \in \mathbb{Z}^n} F(x, y)
\]
\[
\text{s.t. } G(x, y) \leq 0
\]
where \( y \) satisfies:
\[
\min_{y \in \mathbb{R}^m} f(x, y)
\]
\[
\text{s.t. } g(x, y) \leq 0
\]
Here, \( x \) is the upper - level integer decision variable, \( y \) is the lower - level continuous decision variable, \( F \) and \( f \) are the upper - level and lower - level objective functions respectively, and \( G \) and \( g \) are the upper - level and lower - level constraint functions respectively.
### Solution Methods
1. **Application of KKT Conditions**:
- Using the KKT optimality conditions, the bilevel programming problem is transformed into a single - level programming problem with complementary constraints:
\[
\min_{x, y, \lambda} F(x, y)
\]
\[
\text{s.t. } G(x, y) \leq 0
\]
\[
\nabla_y L(x, y, \lambda) = 0
\]
\[
\lambda^\top g(x, y) = 0
\]
\[
g(x, y) \leq 0
\]
\[
\lambda \geq 0, \quad x \in \mathbb{Z}^n, \quad y \in \mathbb{R}^m
\]
where the Lagrange function \( L(x, y, \lambda) = f(x, y) + \lambda^\top g(x, y) \).
2. **Smoothing Technique**:
- Applying the smoothing function \( \phi_\mu(a, b) = a + b - \sqrt{(a - b)^2 + 4\mu^2} \) to deal with complementary constraints, and further transforming the problem into:
\[
\min_{x, y, \lambda} F(x, y)
\]
\[
\text{s.t. } G(x, y) \leq 0
\]
\[
\nabla_y L(x, y, \lambda) = 0
\]