WIFI empowerment: Towards Flexible, Adaptable and Reliable QoS in IoT Networks

IoT networks are expanding at an unprecedented rate, with new use cases and applications being developed every day. The IoT sensors and devices in these networks rely heavily on their wireless connectivity to deliver data reliably, particularly in smart healthcare and industrial IoT networks. These connectivity requirements vary significantly in terms of throughput, latency, and packet loss, depending on the use case.

WiFi, being the most widely used technology for IoT connectivity, lacks the necessary QoS diversity to support diverse IoT applications. This thesis addresses QoS architectural improvements required in WiFi technology to meet the diverse QoS demands of healthcare and industrial IoT use cases. Moreover, network slicing technology has been employed to develop a flexible QoS delivery system for WiFi-enabled IoT networks.

Since wireless network conditions and QoS requirements change over time, deep reinforcement learning (DRL) has been utilized to develop an autonomous and adaptable system that manages QoS in a continuously evolving wireless environment. To enhance the reliability of the system, a Cross-Layer Design (CLD) approach is adopted alongside DRL-based optimization methods, creating a fully flexible, adaptable, and reliable QoS architecture for WiFi-based IoT networks.