Machine Learning-Based Real-Time Indoor Landmark Localization

Abstract

Nowadays, smartphones can collect huge amounts of data from their surroundings with the help of highly accurate sensors. Since the combination of the Received Signal Strengths of surrounding access points and sensor data is assumed to be unique in some locations, it is possible to use this information to accurately predict smartphones' indoor locations. In this work, we apply machine learning methods to derive the correlation between smartphones' locations and the received Wi-Fi signal strength and sensor values. We have developed an Android application that is able to distinguish between rooms on a floor, and special landmarks within the detected room. Our real-world experiment results show that the Voting ensemble predictor outperforms individual machine learning algorithms and it achieves the best indoor landmark localization accuracy of 94% in office-like environments. This work provides a coarse-grained indoor room recognition and landmark localization within rooms, which can be envisioned as a basis for accurate indoor positioning.

Conditional probability-based ensemble learning for indoor landmark localization

“…• Experiment results show that our new algorithm outperforms other predictors, including our previous work [13].…”

“…We propose to apply existing machine learning methods (including both individual and ensemble predictors) and to develop novel conditional probability model-based ensemble predictors to solve this task due to the large amount of features that are available in indoor environments, such as Wi-Fi RSS values, magnetic field values, illuminance level, etc. In our previous work [13], we have validated the performance of different ML algorithms to recognize indoor landmarks. The results show that the Voting ensemble predictor outperforms individual machine learning algorithms and it achieves the best indoor landmark localization accuracy of 90% in office-like environments.…”

“…Whereas our previous work applied existing ML algorithms to recognize indoor locations [13], this work presents a novel ensemble machine learning algorithm to achieve high and stable room prediction performance. Compared to [13], the contribution of this article can be summarized as follows:…”

“…• We conduct experiments in a large indoor office environment that covers an entire floor, which has the double size of the experiment areas in [13].…”

Bluetooth beacons based indoor positioning in a shopping malls using machine learning

“…d. Support vector machine classifier (SVMC) SVMs are supervised learning models and their corresponding learning algorithms are used in ML for regression and classification analyses. A SVM training method, a non-probabilistic binary linear classifier, creates a model that categorizes fresh data measurements into one of two groups [27]. e. Gradient boosting classifier A group of ML techniques known as GBC combine numerous weak learning models to create a powerful predictive model.…”