Adaptive Slice-Aware Multi-access Edge Computing Scheduling for Low-Latency 5G URLLC Communications

Abstract

Ultra-Reliable Low-Latency Communication (URLLC) is one of the most critical types of services in fifth generation (5G) networks for applications which have the need for high speed and reliable communication. Use-cases like autonomous transportation, industrial automation, robotics, and remote healthcare can be negatively affected even by small delay increments in communication. Multi-access Edge Computing (MEC) and network slicing can lower the delay created by centralized processing by bringing computation closer to users and application-specific resources on the network level. Nevertheless, predictable latency in the face of changing network conditions is difficult to achieve due to network congestion, queueing delays, propagation delay, and static resource allocation policy. This paper introduces an Adaptive Slice Aware MEC Scheduling (ASMS) framework that aims to solve the problem of low latency in 5G URLLC scenarios. The suggested solution considers MEC-based processing, network slicing, dynamic resource allocation, intelligent base-station selection, and traffic-aware scheduling. Priority-based decision-making takes place based on the evaluation of network conditions including the current situation in queues, communication latency, channel quality, slice requirements, and Service Level Agreement (SLA) conditions prior to allocating computational and communication resources. The scheduling procedure is ongoing so that resources could be allocated again in case of congestion and violation of SLA. The effectiveness of the proposed ASMS framework was tested in simulations in various conditions of traffic and scheduling using an integrated simulation environment which included MATLAB, NS-3, and Python/SimPy simulators. According to the results, the adaptive MEC-based approach managed to decrease the average communication latency from 2.41 ms to 0.47 ms while staying SLA compliant. At the same time, the improvements were observed in the parameters of throughput, Signal-to-Noise Ratio (SNR), and Bit Error Rate (BER) compared with centralized processing approach.