A Nonlinear Autoregressive Neural Network for Interference Prediction and Resource Allocation in URLLC Scenarios

TL;DR

This paper introduces a nonlinear autoregressive neural network model to predict interference in URLLC scenarios, enabling more efficient resource allocation and meeting strict reliability and latency requirements.

Contribution

The paper presents a novel NARNN-based interference prediction method tailored for URLLC, improving prediction accuracy and resource efficiency over existing algorithms.

Findings

Achieved 7.8% mean absolute percentage error in interference prediction.

Reduced resource usage by up to 15% compared to previous methods.

Demonstrated effectiveness in meeting URLLC reliability and latency demands.

Abstract

Ultra reliable low latency communications (URLLC) is a new service class introduced in 5G which is characterized by strict reliability $(1-10^{-5})$ and low latency requirements (1 ms). To meet these requisites, several strategies like overprovisioning of resources and channel-predictive algorithms have been developed. This paper describes the application of a Nonlinear Autoregressive Neural Network (NARNN) as a novel approach to forecast interference levels in a wireless system for the purpose of efficient resource allocation. Accurate interference forecasts also grant the possibility of meeting specific outage probability requirements in URLLC scenarios. Performance of this proposal is evaluated upon the basis of NARNN predictions accuracy and system resource usage. Our proposed approach achieved a promising mean absolute percentage error of 7.8 % on interference predictions and also reduced the resource usage in up to 15 % when compared to a recently proposed interference prediction algorithm.