A Deep Learning for Predicting Possibility of Bluetooth Low Energy Tag Signals

With the rapid growth of IoT technology, using Bluetooth Low Energy (BLE) gateways to track the trajectory of elderly individuals can help prevent them from getting lost. However, BLE tag signals do not persist, and BLE gateways often enter standby mode for extended periods, resulting in unnecessary energy waste and reducing the lifespan of the devices. To address this issue, predicting the likelihood of BLE tag signals is essential to conserve energy. This paper presents a deep learning approach to enhance the operational efficiency and lifespan of BLE gateways deployed in an IoT system. The system consists of two major modules: the Number of BLE Tags Estimation Module and the Next Accessed BLE Gateway Prediction Module. The Number of BLE Tags Estimation Module leverages a CNN-GRU (Convolutional Neural Network - Gated Recurrent Unit) architecture to predict the time lapse between the sensing of signals from BLE tags. The Next Accessed BLE Gateway Prediction Module employs a self-attention mechanism to predict which specific BLE gateway will receive the BLE tag signal, further optimizing system performance. Experimental results show that the proposed prediction model effectively reduces unnecessary energy consumption, thereby improving the efficiency and service life of the entire IoT system.