A Contextual Aware Enhanced LoRaWAN Adaptive Data Rate for mobile IoT applications

Abstract

Long range wide area network (LoRaWAN) utilize Adaptive Data Rate (ADR) for static Internet of Things (IoT) applications such as smart parking in smart city. Blind ADR (BADR) has been introduced for end devices to manage the resources of mobile applications such as assets tracking. However, the predetermined mechanism of allocating the spreading factors (SFs) to mobile end devices is not adequate in terms of energy depletion. Recently, AI-based solutions to resource allocation have been introduced in the existing literature. However, implementing complex models directly on low-power devices is not ideal in terms of energy and processing power. Therefore, considering these challenges, in this paper, we present a novel Contextual Aware Enhanced LoRaWAN Adaptive Data Rate (CA-ADR) for mobile IoT Applications. The proposed CA-ADR comprises two modes offline and online. In offline mode, we compile a dataset based on successful acknowledgments received by the end devices. Later, dataset is modified by implementing contextual rule-based learning (CRL), following which we train a hybrid CNN-LSTM model. In the online mode, we utilize pre-trained model for efficient resource allocation (e.g., SF) to static and mobile end devices. The proposed CA-ADR has been implemented using TinyML, recommended for low-power and computational devices, which has shown improved results in terms of packet success ratio and energy consumption.