This article presents the first field deployment of a multi-agent reinforcement-learning (MARL) based variable speed limit (VSL) control system on the I-24 freeway near Nashville, Tennessee. We describe how we train MARL agents in a traffic simulator and directly deploy the simulation-based policy on a 17-mile stretch of Interstate 24 with 67 VSL controllers. We use invalid action masking and several safety guards to ensure the posted speed limits satisfy the real-world constraints from the traffic management center and the Tennessee Department of Transportation. Since the time of launch of the system through April, 2024, the system has made approximately 10,000,000 decisions on 8,000,000 trips. The analysis of the controller shows that the MARL policy takes control for up to 98% of the time without intervention from safety guards. The time-space diagrams of traffic speed and control commands illustrate how the algorithm behaves during rush hour. Finally, we quantify the domain mismatch between the simulation and real-world data and demonstrate the robustness of the MARL policy to this mismatch.

Abstract-- The deep neural networks (DNNs)-based autonomous driving systems (ADSs) are expected to reduce road accidents and improve safety in the transportation domain as it removes the factor of human error from driving tasks. The DNN-based ADS sometimes may exhibit erroneous or unexpected behaviours due to unexpected driving conditions which may cause accidents. Therefore, safety assurance is vital to the ADS. However, DNN-based ADS is a highly complex system that puts forward a strong demand for robustness, more specifically, the ability to predict unexpected driving conditions to prevent potential inconsistent behaviour. It is not possible to generalize the DNN model's performance for all driving conditions. Therefore, the driving conditions that were not considered during the training of the ADS may lead to unpredictable consequences for the safety of autonomous vehicles. This study proposes an autoencoder and time series analysis-based anomaly detection system to prevent the safety-critical inconsistent behaviour of autonomous vehicles at runtime. Our approach called DeepGuard consists of two components. The first component-the inconsistent behaviour predictor, is based on an autoencoder and time series analysis to reconstruct the driving scenarios. Based on reconstruction error (e) and threshold (θ), it determines the normal and unexpected driving scenarios and predicts potential inconsistent behaviour. The second component provides on-the-fly safety guards, that is, it automatically activates healing strategies to prevent inconsistencies in the behaviour. We evaluated the performance of DeepGuard in predicting the injected anomalous driving scenarios using already available open-sourced DNN-based ADSs in the Udacity simulator. Our simulation results show that the best variant of DeepGuard can predict up to 93 % on the CHAUFFEUR ADS, 83 % on DAVE-2 ADS, and 80 % of inconsistent behaviour on the EPOCH ADS model, outperforming SELFORACLE and DeepRoad. Overall, DeepGuard can prevent up to 89% of all predicted inconsistent behaviours of ADS by executing predefined safety guards. I. INTRODUCTION Autonomous vehicles are one of the most promising applications of artificial intelligence. This would be a technological revolution in the transportation industry in the near future. Autonomous driving systems (ADSs) use sensors such as cameras, radar, Lidar, and GPS to automatically produce driving parameters such as vehicle velocity, throttle, brakes, steering angles, and directions. Advancements in deep learning have made progress in autonomous systems, such as autonomous vehicles and unmanned aerial vehicles.