A hybrid and self-adaptive QoS and QoE-driven RAT selection strategy for IoT devices

In recent years, the Internet of Things (IoT) has witnessed an explosion, which enables new applications based on connecting numerous devices to the Internet. However, the increasing number of IoT devices utilizing various Radio Access Technologies (RATs), each with distinct communication requirements, presents significant challenges in ensuring reliable Quality of Service (QoS) in such a heterogeneous, multi-technology, multi-user, and multi-application environment. Therefore, for a proper RAT selection, this paper proposes a novel hybrid and adaptive approach based on Deep Reinforcement Learning (DRL) that combines QoS and Quality of Experience (QoE) factors. Unlike previous research, our approach integrates QoS parameters and an adaptive network trust factor that reflects the user's QoE based on reliable past device experiences, ultimately providing users with consistently better connectivity. Within our framework, an IoT device dynamically selects its RAT in a distributed manner based on service requirements and the trust score it assigns to RATs, reflecting its satisfaction with past RAT behavior. The proposed hybrid and adaptive RAT selection approach meets QoS requirements, for various applications at any time as indicated by simulations run using Network Simulator 3 (NS3). Compared to other methods, the energy consumption was reduced by 64.12%, network handovers were decreased by 37.73%, while the average packet delivery ratio was improved by 17.85%. This demonstrates the effectiveness of the proposed approach in optimizing both energy efficiency and network performance, ensuring a more reliable and sustainable system for IoT applications.