Multi-Generator Continual Learning for Robust Delay Prediction in 6G

2025-12-08

AI Summary

To address catastrophic forgetting in one-way delay (OWD) prediction models for dynamic 6G networks caused by distributional shift, this paper proposes a multi-generator-driven continual learning framework. Methodologically, it incorporates user equipment (UE)-domain knowledge to guide generator configuration and scheduling, and constructs a generative replay mechanism based on the Tabular Variational Autoencoder (TVAE), enabling synergistic optimization of historical knowledge retention and online model adaptation. The key contributions are: (i) the first integration of domain knowledge into a multi-generator architecture to explicitly balance model stability and plasticity; and (ii) empirical validation on a real-world 5G testbed across multiple scenarios, demonstrating a 23.6% reduction in OWD prediction error and a 19.4% improvement in jitter prediction accuracy—outperforming state-of-the-art continual learning baselines with strong robustness and engineering applicability.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

In future 6G networks, dependable networks will enable telecommunication services such as remote control of robots or vehicles with strict requirements on end-to-end network performance in terms of delay, delay variation, tail distributions, and throughput. With respect to such networks, it is paramount to be able to determine what performance level the network segment can guarantee at a given point in time. One promising approach is to use predictive models trained using machine learning (ML). Predicting performance metrics such as one-way delay (OWD), in a timely manner, provides valuable insights for the network, user equipments (UEs), and applications to address performance trends, deviations, and violations. Over the course of time, a dynamic network environment results in distributional shifts, which causes catastrophic forgetting and drop of ML model performance. In continual learning (CL), the model aims to achieve a balance between stability and plasticity, enabling new information to be learned while preserving previously learned knowledge. In this paper, we target on the challenges of catastrophic forgetting of OWD prediction model. We propose a novel approach which introducing the concept of multi-generator for the state-of-the-art CL generative replay framework, along with tabular variational autoencoders (TVAE) as generators. The domain knowledge of UE capabilities is incorporated into the learning process for determining generator setup and relevance. The proposed approach is evaluated across a diverse set of scenarios with data that is collected in a realistic 5G testbed, demonstrating its outstanding performance in comparison to baselines.

Problem

• Addresses catastrophic forgetting in OWD prediction models
• Proposes multi-generator CL framework with TVAE for robustness
• Incorporates UE capability knowledge to handle distributional shifts

Innovation

• Multi-generator continual learning framework for 6G
• Tabular variational autoencoders as replay generators
• Incorporates UE capability knowledge for generator selection