Half-Meter Accuracy in Localization Using Frequency Diversity and/or Spatial Diversity

When it comes to indoor localization, is spatial diversity better than frequency diversity? Ranging with respect toWi-Fi access points is useful for indoor localization, but has been thought to be limited to 1–2 m accuracy when using the IEEE 802.11mc protocol. It has been shown that line-of-sight measurements of the distance between a phone and aWi-Fi access point can be more accurate when measurements taken at different frequencies are combined. Here, we show that this frequency-diversity advantage also applies to the determination of location—even when the signals are subject to multipath, diffraction and refractive retardation. The newer IEEE 802.11az protocol, not yet widely implemented, provides sub-meter accuracy more directly, in part because of spatial diversity. We compare 1these two approaches. In previous work, the Bayesian grid update method has been shown to be effective in determining locations from range measurements. This method uses an observation model—the conditional probability distribution of the measured distance versus the actual distance. Good results can be obtained with a simple generic observation model, but the accuracy can be increased by adjusting the observation model. The best model parameters can be found using the maximum-likelihood method. We demonstrate that in a typical cluttered indoor environment 1/2 m accuracy in localization can be achieved using either (i) IEEE 802.11mc with frequency diversity or (ii) IEEE 802.11az with spatial diversity. Even higher accuracy is possible if spatial diversity and frequency diversity are combined—albeit at the cost of slower update rates.