Limitations of Stochastic Geometry Modelling for Estimating the Performance of CSMA Networks

Abstract

This letter considers stochastic geometry modelling (SGM) for estimating the signal-to-interference-and-noise ratio (SINR) and throughput of CSMA networks. We show that, despite its compact mathematical formulation, SGM has serious limitations in terms of both accuracy and computational efficiency. SGM often severely underestimates the SINR versus ns-3 simulations, yet as it neglects the sensing overhead when mapping SINR to throughput, SGM usually overestimates the throughput substantially. We propose our hybrid model for CSMA, which we argue is a superior modelling approach due to being significantly more accurate and at least one order of magnitude faster to compute than SGM.