Beyond Static Forecasting: Unleashing the Power of World Models for Mobile Traffic Extrapolation

2026/04/30 by Xiaoqian Qi, Haoye Chai, Yue Wang +1
Computer Science · #cs.NI

paper · pdf · doi:10.48550/arxiv.2604.08199

arxiv created 2026/08/06 · arxiv updated 2026/08/07

Abstract

Mobile traffic prediction is a fundamental yet challenging problem for wireless network planning and optimization. Conventional models mainly learn static long-term temporal patterns and cannot capture the dynamics under network-parameter adjustments. Leveraging the advantage of world models in learning underlying dynamics, we propose MobiWM, a mobile network world model that treats cell traffic as states and antenna parameters as actions. MobiWM combines factorized spatio-temporal modelling with multimodal environmental context aligned through shared spatial semantics. Its learned action-state transitions enable iterative rollout over specified adjustment trajectories for counterfactual planning. Extensive experiments on massive variable-parameter mobile traffic datasets demonstrate that MobiWM outperforms baselines by at least 16.40% on average. A model-based Actor-critic case study further demonstrates its potential as a learned surrogate for network optimization.

Citations

• HiSTM: Hierarchical Spatiotemporal Mamba for Cellular Traffic Forecasting
• MobiWorld: World Models for Mobile Wireless Network
• Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts
• TD-MPC2: Scalable, Robust World Models for Continuous Control
• Sionna RT: Differentiable Ray Tracing for Radio Propagation Modeling
• Mastering Diverse Domains through World Models
• TransDreamer: Reinforcement Learning with Transformer World Models
• CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation
• Mastering Atari with Discrete World Models
• Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting
• Open Graph Benchmark: Datasets for Machine Learning on Graphs
• Dream to Control: Learning Behaviors by Latent Imagination
• Graph WaveNet for Deep Spatial-Temporal Graph Modeling
• Recurrent World Models Facilitate Policy Evolution
• Deep Learning in Mobile and Wireless Networking: A Survey
• Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting
• Proximal Policy Optimization Algorithms
• Inductive Representation Learning on Large Graphs
• Deep Spatio-Temporal Residual Networks for Citywide Crowd Flows Prediction
• Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting