DAELocNet: An AI-Enabled Indoor Localization for Accurate Positioning in Modern Buildings using Wireless Communication Technologies

Global positioning systems (GPS) do not provide accurate localization for indoor built environments. For accurate indoor positioning, wireless communication technologies with sensing and computational techniques can be used. There are numerous traditional approaches for indoor position estimation using wireless communication technologies. However, each traditional location estimation technique has its pros, cons, and limitations. The evolution of artificial intelligence (AI) along with the various sensing and wireless communication technologies including the Internet of Things (IoT), expressed promising outcomes to mitigate the existing challenges and quickly develop the indoor positioning system. Towards this, in this research work, we proposed an AI-based framework for the accurate indoor positioning system using the fingerprinting of high measurement density received signals strength indicator (HMD-RSSI) of IEEE 802.11 wireless communication network. Initially, we utilized machine learning regressors-based techniques to estimate the longitude and latitude separately, and then we used multioutput with regressors to predict both coordinates simultaneously. Further, we proposed auto-encoders with neural networks based two deep learning techniques (1) IAELocNet, and (2) DAELocNet for accurate position estimation. We utilized the recently published dataset of HMD-RSSI for training and validation of the proposed methodology. Finally, we performed a comparative analysis of implemented techniques using performance evaluation metrics. Experimental results showed that DAELocNet got the lowest error score of mean localization error of 2.29 meters, variance of 0.96, and standard deviation of 0.98.