Robust CNN based approach for Indoor localization using RSSI

Indoor localization is vital for applications like navigation and asset tracking, but is challenged by the ineffectiveness of GPS indoors. This study presents an AI-based indoor localization system that utilizes machine learning to enhance accuracy and reliability by integrating wireless signals strength. The received signal strength indicator (RSSI) values are collected from the Bluetooth beacons installed in the indoor scenario to enable user localization. A CNN based model is proposed that is trained using the collected RSSI data. The proposed model estimates the location of the user or device with accuracy of 93.1%. Further, the paper gives a comparative analysis of proposed model with other existing models in terms of mean square error (MSE) in localization and average loss. The proposed method offers improved localization in complex indoor environments and adapts to dynamic changes, making it robust for real-world applications. This work offers a scalable solution for indoor localization, benefiting smart building management and IoT applications.