A New Distance Metric Based on Bias-Variance Tradeoff for Wireless Fingerprinting

This article proposes a new distance metric for indoor positioning based on wireless fingerprinting, referred to as the weighted bias-variance (WBV) metric. The WBV metric is parameterized by a single parameter that is adjusted based on the concept of bias-variance tradeoff, making it highly effective in scenarios where there is a network reconfiguration or change in the environmental layout. The optimal setting was found to be minimally affected by the number of nearest neighbors in the estimation algorithm as well as the access point (AP) selection schemes. In an experiment at a university library, the switch from the Manhattan metric to the WBV metric after a network reconfiguration achieved a reduction in the mean position error (MPE), standard deviation and median of the error distribution of 34.9%, 22.1%, and 41.5%, respectively, with negligible changes to the floor accuracy (FA). When the number of APs was reduced, the corresponding reduction was 42.7%, 35.5%, and 46.3%, respectively. In another experiment at a university faculty building, the switch from the Manhattan metric to the WBV metric after a layout change attained a reduction in the MPE, standard deviation, and median of the error distribution of 55.1%, 62.6%, and 23.1%, respectively. The FA improved by 22.2%. The drastic improvement in accuracy has the potential to significantly advance indoor positioning for the ubiquitous Internet of Things.