A Deep Reinforcement Learning Approach For LoRa WAN Energy Optimization

In almost-majority of IoT (Internet of Things) networks; the wireless sensor nodes are small and low-cost portable devices. Besides, they are typically battery-powered with a finite lifespan duration. In most emerged applications, these nodes are required to remain for the maximum time possible with a standalone performance for large-scale coverage. Therefore, it is important to understand the node's processing before real implementation of any massive long-range IoT networks to assign accurate settings for each node on its positions along with estimating the approximate mean duration of all networks. LoRa (Long Range) is one of the recently emerged Low Power Wide Area Networks IoT networks that is knowing high interest in various applications. In this paper, we propose a Markov Decision Process-based model that aims to estimate the lifespan of LoRa class A end nodes regarding the efficiency in resources allocation and transmission parameter settings. The proposed model has been tested and showed interesting results on the accuracy of lifetime estimation and prediction under different scenarios.