AI-Driven Edge Intelligence for Bandwidth- and Energy-Efficient Cellular IoT Networks Toward 6G

Abstract

The extreme proliferation of the IoT and the development of 5G and now 6G networks have created new bandwidth, spectrum-efficiency and energy costs issues. The traffic patterns that take place in the IoT are dynamic and heterogeneous, and these cannot be tackled using the traditional, non-dynamic procedures of resource allocation. In order to conserve bandwidth consumption of the cells, and guarantee Quality of Service (QoS) at the Internet of Things networks, the present paper suggests an artificial intelligence (AI)-based optimization framework. Learning the best transmission policy is a side effect of the reinforcement learning (RL), that is, Deep Q-Networks (DQN) and Proximal Policy Optimization (PPO). The structure has brought in edge intelligence, elastic packet accumulation and predictable time schedule in the instant decision-making. The publicly accessible simulation tool (e.g., NS-3 and OMNeT++) will simulate the traffic of the IoT not only on top of LTE-M and NB-IoT but also on top of the 5G mMTC. Some of the examples of evaluation metrics include end-to-end latency, energy efficiency, ratio of packet delivery and savings in bandwidth. It is projected that the result will improve the design of smart, resource efficient and scalable cellular IoT networks that can be upgraded to 6G.