Adaptive Duty-Cycle Scheduling using Bi-Directional Long Short-Term Memory (BiLSTM) for Next Generation IoT Applications

Internet of Things (IoT) is a group of interconnected devices which contain sensors to collect useful data. Artificial Intelligence (AI) techniques like Machine learning (ML) and Deep Learning (DL) are very much popular in many applications including IoT. The main goal of Artificial Intelligence of Things (AIoT) is to perform more effective IoT functions, enhance the man-to-machine interactions and enhance the capacity for data management. In this paper, an adaptive duty-cycle scheduling using BiDirectional Long Short-Term Memory (BiLSTM) algorithm is proposed for IoT networks. In this algorithm, initially, the Jaccard Similarity Index (JSI) between two sensors are estimated. Then sensors with highest similarity (ie highest JSI values) are selected for sleep scheduling. Once the eligible sensors for sleep scheduling are selected based on the JSI values, their traffic loads and Energy levels are estimated from the previous traffic patterns using BiLSTM. The duty cycle of sensors are by estimated by adjusting the length of sleep duration according to the estimated traffic load and energy levels of the selected sensors. By experimental results, we show that the proposed BiLSTM-ADS technique achieves higher prediction accuracy and residual energy with lesser computational cost.