Artificial Intelligence Approaches to Mitigating Network Congestion in IoT Systems

Abstract

The unprecedented explosion of Internet of Things (IOT) devices has elevated the requirements of the network infrastructures to unprecedented levels, causing severe congestion problems, especially in applications which demand low latency, high throughput, and real-time feedback. Static routing protocols, AQM, and TCP variants are some of the traditional mechanisms for congestion control that are unable to perform efficiently in dynamic and diverse IoT environments as they are reactive-based and inflexible. To this end, in this paper, we explore the promising ability of Artificial Intelligence (AI) methods such as Machine Learning (ML), Deep Learning (DL), Reinforcement Learning (RL), and their combination in natura for proactive and intelligent traffic management for IoT. A comparative review of strengths (e.g., adaptivity in RL, pattern recognition in DL) and weaknesses (in terms of its scalability, interpretability, resources) of each method is also discussed. Moreover, the paper indicates some crucial research challenges on model generalization, evaluation criterion and platform integration. Future possible research directions to bridge these gaps include the development of lightweight AI architectures, Explainable AI (XAI) frameworks, cross-platform model deployment, scalable FL, and standardized benchmarking datasets. This work also leads to a hybrid AI model for traffic congestion prediction and control with an application of simulation tool and real data. Simulation results show significant improvements in latency, packet loss, and energy consumption. Finally, the study presents a ground work for incorporating the scalable, secure and intelligent AI enabled congestion control systems in a wide area of IoT applications.