Research Article
Metaheuristic Optimisation for Radio Interface-Constrained Channel Assignment in a Hybrid Wi-Fi–Dynamic Spectrum Access Wireless Mesh Network
@INPROCEEDINGS{10.1007/978-3-030-98002-3_4, author={Natasha Zlobinsky and David Johnson and Amit Kumar Mishra and Albert A. Lysko}, title={Metaheuristic Optimisation for Radio Interface-Constrained Channel Assignment in a Hybrid Wi-Fi--Dynamic Spectrum Access Wireless Mesh Network}, proceedings={Cognitive Radio Oriented Wireless Networks and Wireless Internet. 16th EAI International Conference, CROWNCOM 2021, Virtual Event, December 11, 2021, and 14th EAI International Conference, WiCON 2021, Virtual Event, November 9, 2021, Proceedings}, proceedings_a={CROWNCOM \& WICON}, year={2022}, month={3}, keywords={Channel Assignment Dynamic Spectrum Access DSA Wireless Mesh Networks WMN CBRS Wi-Fi 6E TVWS Genetic Algorithm Simulated Annealing}, doi={10.1007/978-3-030-98002-3_4} }- Natasha Zlobinsky
David Johnson
Amit Kumar Mishra
Albert A. Lysko
Year: 2022
Metaheuristic Optimisation for Radio Interface-Constrained Channel Assignment in a Hybrid Wi-Fi–Dynamic Spectrum Access Wireless Mesh Network
CROWNCOM & WICON
Springer
DOI: 10.1007/978-3-030-98002-3_4
Abstract
Channel Assignment (CA) in wireless mesh networks (WMNs) has not been well studied in scenarios where the network uses Dynamic Spectrum Access (DSA). This work aims to fill some of this gap. We compare metaheuristic algorithms for optimising the CA in a WMN that has both Wi-Fi and DSA radios (where DSA could be Television White Spaces or 6 GHz). We also present a novel algorithm used alongside these metaheuristic algorithms to ensure that the CA solutions are feasible. Feasible solutions meet the interface constraint, i.e. only as many channels are allocated to a node as it has radios. The algorithm also allows the topology to be preserved by maintaining links. Many previous studies tried to ensure feasibility and/or topology preservation by using two separate steps. The first step optimised without checking feasibility and the second step fixed infeasible solutions. This second step often negated the benefits of the previous step and degraded performance. Other CA algorithms tend to use simple on/off interference models, instead of models that more realistically reflect the physical layer environment, such as the Signal to Interference plus Noise Ratio (SINR). We present our more realisticSINR-based model and optimisation objective. Simulated Annealing (SA) and Genetic Algorithm (GA) are applied to the problem. Performance is evaluated and verified through simulation. We find that GA outperforms SA, finding higher quality solutions faster, although both metaheuristics are better than random allocations. GA can be used daily to find good CAs in changing conditions.
