Q-Net: Queue Length Estimation via Kalman-based Neural Networks

Q-Net is a queue length estimation framework designed to integrate aggregated vehicle counts and floating car data using a Kalman-based neural network architecture. It employs a state-space formulation and an AI-augmented Kalman filter to learn time-varying dynamics, achieving real-time performance and improved spatial transferability by grouping measurements into fixed-size local groups. This approach enhances the accuracy of queue estimation at signalized intersections, making it a valuable tool for traffic management without the need for expensive sensing infrastructure.