Enhancing Quality-of-Service in LoRa Low-Power Wide-Area Networks via Multi-Hop Optimized Radio Resource Management

In transport systems in particular, LoRa wireless communication has become an essential technology for intelligent critical infrastructures. Optimizing transmission policy parameters, including carrier frequency, bandwidth, coding rate, and spreading factor, is crucial for achieving excellent Quality of Service (QoS) for LoRa systems; however, current methods have not fully optimized all LoRa parameters. In order to efficiently share information, the proposed Multiobjective Q Learning based LoRa (MQ-LoRa) uses a hybrid topology consisting of a cluster root rotating tree, wherein IoT nodes adhere to a cluster topology inspired by cellular tissues, while gateways follow a tree topology. For cluster construction, a membrane-inspired clustering algorithm is created, and influence scores are used to determine which header node is best. The study presents Data QoS Ranking for IoT nodes, which uses the new LoRa frame structure field QoS Ranking (QRank) to classify data into priority and non-priority categories. The Soft Actor Critic (SAC) fast deep reinforcement learning model is used for optimal transmission policy enforcement, taking into account signal-to-interference-plus-noise ratio, QRank, and other environmental characteristics. After that, the transmission policy is optimized taking carrier frequency, bandwidth, coding rate, and spreading factor into account. Furthermore, a multi-hop routing method with concurrent optimization that uses shuffling shepherd and mayfly optimization rates routes according to fitness standards. A multi-weighted sum approach is used to establish a weighted duty cycle in LoRa IoT networks to minimize resource inefficiency and loss of information. Using the NS3.26 LoRaWAN module, performance evaluation measures energy consumption, latency, throughput, packet receipt ratio, and packet rejection ratio, among other parameters. Comparison with popular LoRa techniques shows that the suggested MQ-LoRa performs better energy consumption of 56 mJ than exisitng Reinforcement Learning based LoRA, according to experimental results.