Reconstruction of a Freeway Control Systems’ Algorithm based on Convolution Neural Networks

Freeway Control Systems (FCS) play a vital role in enhancing road safety and traffic efficiency by dynamically managing traffic through variable speed limits, overtaking restrictions, and warning messages. An accurate representation of FCS behavior in traffic simulations is essential for realistic modeling. However, the manual implementation of FCS logic in simulations is time-consuming and requires high customization. This study proposes a data-driven approach to automatically reconstruct FCS control algorithms from historical traffic and display data. The models achieved prediction accuracies of at least 87%, effectively capturing key behaviors such as congestion-related speed reductions. Among the architectures evaluated, the baseline Convolutional neural network offered the best balance of performance and computational efficiency. At the same time, more complex models showed promise for further accuracy gains with continued development. These findings demonstrate the feasibility and potential benefits of integrating FCS models based on neural networks into traffic simulations.