Search papers, labs, and topics across Lattice.
This paper introduces TimeRAN, a multi-task learning framework using a time-series foundation model for the Radio Access Network (RAN), designed to overcome limitations of task-specific models. They curate and open-source TimeRAN DataPile, a large time-series corpus for RAN analytics, comprising 355K time series and 0.56B measurements. Evaluations across anomaly detection, classification, forecasting, and imputation tasks demonstrate state-of-the-art performance with minimal fine-tuning, validated in a 5G testbed.
A unified foundation model can now handle diverse RAN analytics tasks like anomaly detection and forecasting with state-of-the-art performance, eliminating the need for fragmented, task-specific models.
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.
