Probabilistic Delay Forecasting in 5G Using Recurrent and Attention-Based Architectures

TL;DR

This work tackles the challenge of probabilistic end-to-end latency forecasting in 5G networks by predicting the full delay distribution over multiple future packets. It introduces a three-component pipeline: learnable tokenization of historical network context, temporal models (LSTM or Transformer) to produce distribution parameters, and a Gaussian Mixture MDN to produce conditional delay distributions. Empirical results on a real SDR-based OpenAirInterface 5G testbed show that multi-step Transformer forecasts achieve the best negative log-likelihood and MAE across horizons, with robust calibration and favorable training efficiency. The approach enables proactive network management for adaptive scheduling and resource allocation, contributing to more reliable QoS in evolving 5G/6G systems.

Abstract

With the emergence of new application areas such as cyber-physical systems and human-in-the-loop applications ensuring a specific level of end-to-end network latency with high reliability (e.g., 99.9%) is becoming increasingly critical. To align wireless links with these reliability requirements, it is essential to analyze and control network latency in terms of its full probability distribution. However, in a wireless link, the distribution may vary over time, making this task particularly challenging. We propose predicting the latency distribution using state-of-the-art data-driven techniques that leverage historical network information. Our approach tokenizes network state information and processes it using temporal deep-learning architectures-namely LSTM and Transformer models-to capture both short- and long-term delay dependencies. These models output parameters for a chosen parametric density via a mixture density network with Gaussian mixtures, yielding multi-step probabilistic forecasts of future delays. To validate our proposed approach, we implemented and tested these methods using a time-synchronized, SDR-based OpenAirInterface 5G testbed to collect and preprocess network-delay data. Our experiments show that the Transformer model achieves lower negative log-likelihood and mean absolute error than both LSTM and feed-forward baselines in challenging scenarios, while also providing insights into model complexity and training/inference overhead. This framework enables more informed decision-making for adaptive scheduling and resource allocation, paving the way toward enhanced QoS in evolving 5G and 6G networks.

Figures (16)

• Figure 1: Illustration of packet transmission cycles over a wireless link, showing a broad reception time profile caused by the stochastic nature of wireless delays, which may change over time from cycle $n$ to $n+L$
• Figure 2: A 5G uplink system model illustrating the end-to-end data path and the real-time delay monitoring & prediction module gathering relevant information to estimate the one-way delay.
• Figure 3: Temporal-based learning and prediction approach overview
• Figure 4: Recurrent-based neural prediction
• Figure 5: Transformer-based neural prediction