A Comparative Study of Machine-Learning Algorithms for Indoor Localization Based on the Wi-Fi Fingerprint According to User Postures

With 5G coming and IoT exploding, indoor positioning is becoming increasingly important in our daily lives. Wi-Fi is the most promising technology in this respect due to its wide deployment. Compared to other techniques used by Wi-Fi, fingerprint offers good accuracy. However, dynamic Fingerprint and taking into account the nature of the components of the fingerprint are costly. Machine-learning (ML) algorithms can solve these problems. Another important issue is the consideration of user postures. An indoor localization study using fingerprints and ML was conducted in this work. Users may have different postures when using indoor localization systems. User posture can affect the accuracy of indoor positioning systems, as different postures can cause fluctuations in the received signal strength indicator (RSSI). To provide accurate positioning information, indoor positioning systems must therefore take account of the user’s posture. To this end, we conducted several reproducible real-world scenarios based on different user postures, such as phone to the left ear, phone in the pocket, phone on a stool vertically and horizontally, phone in front of you, and phone on the back left. We then compared the ML algorithms to see the effect of user posture. The experimental results show that the “phone on a stool vertically” user posture gives the best result, with the accuracy of the Random Forest (RF), eXtreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) algorithms being 1.75 m, 1.86 m, 2.03 m and 1.83 m respectively. This is an important factor to consider when improving the accuracy of indoor Wi-Fi fingerprint location relative to the state of the art and previous papers.