A multi-path data exclusion model for RSSI-based indoor localization

Positioning a device with the only help of an RF transmitter in an indoor environment is difficult because of the complexity and of the unpredictable nature of radio propagation in such a scenario. The effects of fading, multipath, shadowing make it difficult to infer distance between two points from a blind measurement of the signal attenuation. However, the Received Signal Strength Indicator (RSSI) remains a popular ranging technique when it comes to the Internet of Things, as it does not require dedicated or expensive hardware. The variability of the RSSI is often addressed by modeling channel attenuation by a parametric model like the log-normal shadowing. Such model parameters are generally evaluated by maximum likelihood estimation (MLE). In this paper, we confront this technique to an indoor realistic testbed and show that it achieves a low accuracy. We propose to use an alternate model named biased log-normal shadowing model that is able to alleviate the effects of multipath and show that MLE on this biased model achieves a better precision.