A Family of Open Time-Series Foundation Models for the Radio Access Network

2026-04-05

AI Summary

This work addresses the fragmentation, limited knowledge sharing, poor generalization, and high system complexity inherent in current radio access network (RAN) intelligence approaches that rely on task-specific models. To overcome these challenges, we propose TimeRAN—the first unified multi-task foundation model framework tailored for RAN time series—featuring a lightweight backbone and task-specific heads to learn transferable representations under limited supervision. Our contributions include releasing TimeRAN DataPile, the largest open RAN time series dataset to date, and establishing the first open foundation model for RAN time series analysis. Experiments demonstrate that TimeRAN achieves state-of-the-art performance across diverse tasks—including anomaly detection, classification, forecasting, and imputation—with minimal or no fine-tuning, while real-world deployment on a 5G testbed confirms its efficiency under low-resource constraints.

Abstract

The Radio Access Network (RAN) is evolving into a programmable and disaggregated infrastructure that increasingly relies on AI-native algorithms for optimization and closed-loop control. However, current RAN intelligence is still largely built from task-specific models tailored to individual functions, resulting in model fragmentation, limited knowledge sharing across tasks, poor generalization, and increased system complexity. To address these limitations, we introduce TimeRAN, a unified multi-task learning framework for time-series modeling in the RAN. TimeRAN leverages a lightweight time-series foundation model with few task-specific heads to learn transferable representations that can be efficiently adapted across diverse tasks with limited supervision. To enable large-scale pretraining, we further curate and open-source TimeRAN DataPile, the largest time-series corpus for RAN analytics to date, comprising over 355K time series and 0.56B measurements across diverse telemetry sources, protocol layers, and deployment scenarios. We evaluate TimeRAN across a comprehensive set of RAN analytics tasks, including anomaly detection, classification, forecasting, and imputation, and show that it achieves state-of-the-art performance with minimal or no task-specific fine-tuning. Finally, we integrate TimeRAN into a proof-of-concept 5G testbed and demonstrate that it operates efficiently with limited resource requirements in real-world scenarios.

Problem

• Radio Access Network
• model fragmentation
• knowledge sharing
• generalization
• system complexity

Innovation

• time-series foundation model
• multi-task learning
• Radio Access Network (RAN)
• transferable representation
• open-source dataset