Modelling Broadband Wireless Technology Coexistence in the Unlicensed Bands

Increasingly more unlicensed bands accommodate multiple broadband technologies, e.g. Wi-Fi and LTE in the 5 GHz band and Wi-Fi 6E and 5G NR-U in the new 6 GHz unlicensed band. In these bands, the technologies coexist primarily via distributed MAC spectrum sharing mechanisms, so it is crucial to evaluate their performance using the right tools that reliably and accurately model the coexisting schemes. CSMA and duty cycle are representative schemes in this context, but the few modelling tools that capture their coexistence performance are yet to be properly evaluated in terms of accuracy and computational efficiency. This paper is the first to compare three such modelling tools – stochastic geometry modelling (SGM), ns-3, and our hybrid model – which comprise different time granularities and abstraction levels for the MAC mechanisms. We compare both the accuracy of their estimated SINR and throughput versus ns-3 and their computational efficiency, for CSMA and duty cycle coexistence. Furthermore, we make recommendations pertinent to future broadband coexistence cases. Our results show that ns-3 captures important effects of the CSMA sensing time on the SINR, but its computation time may be prohibitive for modelling coexisting dense networks. This tool is thus useful primarily for studying dynamic MAC features in small to medium-sized coexisting deployments. SGM is the least accurate and also requires a long computation time, so we argue it is the least attractive approach to model coexisting CSMA and duty cycle technologies. Finally, our hybrid model yields the time-average per-link SINR with reasonable accuracy and the throughput with good accuracy and is at least two orders of magnitude faster to compute than SGM and ns-3, making it very well suited for extensive Monte Carlo simulations and large networks.