Infinite-Term Memory Classifier for Wi-Fi Localization Based on Dynamic Wi-Fi Simulator

Abstract

Wi-Fi localization is an active research topic, and various challenges are not yet resolved in this field. Researchers develop models and use benchmark datasets for Wi-Fi or fingerprinting to create a quantitative comparative evaluation. These benchmarking datasets are limited by their failure to support dynamical navigation. As a result, Wi-Fi models are only evaluated as usual classifiers without including actual navigation maneuvers in the evaluation, which makes the models incapable of handling the actual navigation behavior and its impact on the performance. One common navigation behavior is the cyclic dynamic behavior, which occurs frequently in the indoor environment when a person visits the same place or location multiple times or repeats the same trajectory or similar one more than once. For this purpose, we developed two models: a simulation model for generating time series data to support actual conducted navigation scenarios and a Wi-Fi classification model to handle dynamical scenarios generated by the simulator under cyclic dynamic behavior. Various testing scenarios were conducted for evaluation, and a comparison with benchmarks was performed. Results show the superiority of our developed model which is infinite-term memory online sequential extreme learning machine (OSELM) to the benchmarks with a percentage of 173% over feature adaptive OSELM and 1638% over OSELM.

Keywords

• Indoor localization
• extreme learning machine
• Wi-Fi cyclic dynamics
• feature adaptive