Offline Reinforcement Learning Dataset from a Field Deployed Variable Speed Limit Control System

This dataset consists of VSL operation data deployed on Interstate 24 (I-24), containing a total of 67 VSL controllers operating on both directions of the freeway. The dataset contains ~100 million data samples covering 18 months of continuous operation from March 9, 2024, to September 9, 2025. The deployed VSL control system is based on reinforcement learning, specifically multi-agent reinforcement learning (MARL), and has been successfully field-tested in real-world traffic conditions.

The collected dataset has been converted to offline RL format for further research, enabling the development and evaluation of offline RL algorithms for traffic control applications. This real-world dataset provides valuable insights for researchers working on offline reinforcement learning for intelligent transportation systems.