Adaptive Indoor Area Localization for Perpetual Crowdsourced Data Collection

Abstract

The accuracy of fingerprinting-based indoor localization correlates with the quality and up-to-dateness of collected training data. Perpetual crowdsourced data collection reduces manual labeling effort and provides a fresh data base. However, the decentralized collection comes with the cost of heterogeneous data that causes performance degradation. In settings with imperfect data, area localization can provide higher positioning guarantees than exact position estimation. Existing area localization solutions employ a static segmentation into areas that is independent of the available training data. This approach is not applicable for crowdsoucred data collection, which features an unbalanced spatial training data distribution that evolves over time. A segmentation is required that utilizes the existing training data distribution and adapts once new data is accumulated. We propose an algorithm for data-aware floor plan segmentation and a selection metric that balances expressiveness (information gain) and performance (correctly classified examples) of area classifiers. We utilize supervised machine learning, in particular, deep learning, to train the area classifiers. We demonstrate how to regularly provide an area localization model that adapts its prediction space to the accumulating training data. The resulting models are shown to provide higher reliability compared to models that pinpoint the exact position.

DeepLocBox: Reliable Fingerprinting-Based Indoor Area Localization

“…Fingerprinting is still an attractive approach for device-based—especially smartphone-based—indoor localization, since it does not require dedicated infrastructure to be installed. The advent of deep learning has resulted in increasing interest in transfering its success to indoor localization [ 14 , 15 , 16 , 17 , 18 ]. Still, the nature of the fingerprinting method limits the accuracy that can ideally be reached, such that models are required that nevertheless provide a reliable space estimation while sacrificing as little expressiveness as possible.…”

“…Area classification : Whereas [ 48 ] Liu et al estimated the probability over predefined areas, Laska et al [ 18 ] proposed a framework for adaptive indoor area localization using deep learning to classify the correct segment of a set of predefined segments. Njima et al [ 49 ] constructed 3D input images that consist of the RSS data and the kurtosis values derived from the RSS data.…”

“…This problem formulation is a natural fit for supervised machine learning, which is why the recent popularity of deep learning led to an increasing amount of publications that successfully apply deep learning for fingerprinting-based indoor localization [ 14 , 15 , 16 , 17 , 18 ]. RSS results from the superposition of multipath components and, as a consequence, fluctuates even at a static detection point [ 6 ], which limits the theoretically achievable localization performance.…”

Feature Fusion Using Stacked Denoising Auto-Encoder and GBDT for Wi-Fi Fingerprint-Based Indoor Positioning

“…The solution for fingerprint matching usually requires a preliminary system calibration procedure [26] (called off-line phase) for constructing the indoor floor plan by specialists [27]. This offline phase consists of manual collections for the received signal strength (RSS) observations at certain reference points in predefined locations, regions or grid cells [28]. Then, the collected RSS vectors at each reference point will be stored in a database as training fingerprints for further pattern matching during the localization procedure (called online phase) [26].…”

“…As an alternative to the schemes that aim at locating exact positions, the area classification has recently started to attract attention on estimating the current area of a user, such as the room in a building or the shop inside a mall. This is especially applicable to large-scale deployments with low cost or crowdsourced data with low quality that do not allow for accurate localization [28]. Liu et al [31] proposed an area estimation algorithm by using indoor map information and user trajectories.…”