SMART EDGE COMPUTING INTEGRATION FOR LATENCY-CRITICAL APPLICATIONS IN NEXT-GENERATION WIRELESS NETWORKS

One of the efforts in wireless networks (5G and beyond) is anticipated to serve an enormous amount of latency-sensitive applications, including autonomous driving, augmented/virtual reality (AR/VR), and industrial automation in real-time. Nevertheless, these applications have strict latency and availability demands that are very problematic to traditional cloud-based infrastructures. The purpose of this paper is to demonstrate an intelligent edge computing system that will be used to dynamically assign the computing resources and network bandwidth to tasks, which are latency-sensitive. The suggested system will combine Deep Reinforcement Learning (DRL) with Smart Edge Computing to provide adaptive task execution, efficient bandwidth usage and lower end-to-end latency in heterogeneous network spaces. Python is used to create a realistic simulation model, which will capture the variation of bandwidth with time, loss of packets, queuing behavior, and the prioritization of tasks under various load conditions. The DRA agent keeps learning the best task to node assignment strategies, which would allow reducing delay and keeping Service Level Agreement (SLA) compliance. The experiments have shown that the proposed framework has an average latency of sub-second (0.7-0.9 s), SLA violation of almost zero and can perform load balancing between edge and cloud nodes. These results prove that the performance and the reliability of the applications with a high latency level in 5G/6G networks can be significantly improved with the use of the DRL-based edge intelligence.