An Experimental EACI-Based Localization Framework Using LQI and CNN for Consumer IoT

Precise indoor localization remains a challenge in wireless sensor networks (WSNs) due to multipath fading, interference, and signal fluctuations in different environments. Traditional methods depend on Received Signal Strength (RSS) also often struggle with accuracy in indoor scenario. This study presents an experimental localization framework that utilizes Link Quality Indicator (LQI) values and Convolutional Neural Networks (CNNs) within an Edge Computing-Assisted Consumer IoT (EACI) model. The proposed approach segments the network using a pyramid-loop algorithm and employs LQI-based measurements for more stable and accurate distance estimation. A CNN classifier is trained on normalized LQI data, including statistical features such as kurtosis, to predict node locations. The system is authenticated by a real-world testbed using Zigbee XB24C nodes. The experimental results show an overall localization error of 0.12m at zone 1 with a standard deviation of 0.89m. This reflects an improved localization accuracy and reduced error compared to RSS-based and existing CNN-based methods. The proposed technique effectiveness is observed for indoor localization in consumer IoT environments.