Frequency Assignment Problem with Net Filter Discrimination Constraints

TL;DR

This paper introduces randomized heuristics for the frequency assignment problem that incorporate net filter discrimination effects, significantly improving solution quality and efficiency compared to traditional greedy algorithms.

Contribution

It presents novel randomized algorithms for FAP that are simple to implement, parameter-free, and effectively consider real-world geographical and filter effects.

Findings

Algorithms achieve solutions close to lower bounds

Significant improvement over greedy algorithms

Effective on synthetic and real geographical data

Abstract

Managing radio spectrum resources is a crucial issue. The frequency assignment problem (FAP) basically aims to allocate, in an efficient manner, limited number of frequencies to communication links. Geographically close links, however, cause interference, which complicates the assignment, imposing frequency separation constraints. The FAP is closely related to the graph-coloring problem and it is an NP-hard problem. In this paper, we propose to incorporate the randomization into greedy and fast heuristics. As far as being implemented, the proposed algorithms are very straight forward and are without system parameters that need tuned. The proposed algorithms significantly improve, quickly and effectively, the solutions obtained by greedy algorithms in terms of the number of assigned frequencies and the range. The enhanced versions of proposed algorithms perform close to the lower bounds while running for a reasonable duration. Another novelty of our study is its consideration of the net filter discrimination effects in the communication model. Performance analysis is done by synthetic and measured data, where the measurement data includes the effect of the real 3-dimensional (3D) geographical features in the Daejeon region in Korea. In both cases, we observe a significant improvement by employing randomized heuristics.