Solving a PDE-constrained antenna design problem by mixed-integer programming
Abstract
This paper shows that a mixed integer PDE-constrained design optimization problem can be made amenable to be solved to optimality using standard mixed-integer programming solvers. The concrete problem that we consider is the optimal design of the topology of an antenna, a problem that keeps haunting the engineering community. Indeed, up to this point no exact method was known. Existing methods rely on iterative use of PDE-solvers, embedded in either gradient methods or metaheuristics. However, none of these methods come with performance guarantees. We break this barrier, showing that important antenna design problems can be solved optimally using a mixed integer linear program, plus one singleton second order cone constraint in two variables. The main contribution of the paper is to develop such a formulation. There are two major methodological steps to get there. The first is to show how a standard boundary element method (BEM), based on a discretized design space, translates into a mixed integer program at all. Approached intuitively, this yields a maximization problem with non-concave objective, which is hence not tractable. The second is to make the objective function linear, using a new Charnes-Cooper transform for mathematical programming over complex numbers. We evaluate our approach on the classic dipole antenna, reconfirming well-known experimental results, and apply it to a widely-used test case, the Yagi-Uda antenna. In a successor paper, we report on further computational enhancements using physics-informed cuts.
Related
- Non-convex integer programming: a deterministic solution method of the large block sale problem
- Integer linear programming based methods for computing the maximum chordal subgraph of an arbitrary graph
- On solving integer bilevel optimization problems with a non-convex quadratic follower objective function using disjunctive cuts
- Hidden Convexity via Symmetric Displacement Covers: Mixed-Integer Quadratic Programming and Integer Cubic Programming in Fixed Dimension
- Learning to Solve Mixed-Integer Nonlinear Programming with Neural Networks and Sequential Quadratic Programming Refinement
BibTeX
@misc{hoeksma2026solving,
title = {{Solving a PDE-constrained antenna design problem by mixed-integer programming}},
author = {Ruben Hoeksma and Jesse van Rhijn and Marc Uetz and Wick Wijnholds},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.28132},
eprint = {2610.28132},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.28132},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Solving a PDE-constrained antenna design problem by mixed-integer programming
AU - Ruben Hoeksma
AU - Jesse van Rhijn
AU - Marc Uetz
AU - Wick Wijnholds
PY - 2026
DA - 2026/10/20
PB - arXiv
DO - 10.48550/arxiv.2610.28132
UR - https://arxiv.org/abs/2610.28132
AB - This paper shows that a mixed integer PDE-constrained design optimization problem can be made amenable to be solved to optimality using standard mixed-integer programming solvers. The concrete problem that we consider is the optimal design of the topology of an antenna, a problem that keeps haunting the engineering community. Indeed, up to this point no exact method was known. Existing methods rely on iterative use of PDE-solvers, embedded in either gradient methods or metaheuristics. However, none of these methods come with performance guarantees. We break this barrier, showing that important antenna design problems can be solved optimally using a mixed integer linear program, plus one singleton second order cone constraint in two variables. The main contribution of the paper is to develop such a formulation. There are two major methodological steps to get there. The first is to show how a standard boundary element method (BEM), based on a discretized design space, translates into a mixed integer program at all. Approached intuitively, this yields a maximization problem with non-concave objective, which is hence not tractable. The second is to make the objective function linear, using a new Charnes-Cooper transform for mathematical programming over complex numbers. We evaluate our approach on the classic dipole antenna, reconfirming well-known experimental results, and apply it to a widely-used test case, the Yagi-Uda antenna. In a successor paper, we report on further computational enhancements using physics-informed cuts.
ER -