Q-Net: Kalman-Based Neural Network for Queue Length Estimation
Researchers have developed Q-Net, a queue estimation framework for signalized intersections that integrates aggregated vehicle counts from loop detectors and floating car data (aFCD) using a state-space formulation. The framework addresses challenges like partial observability and violations of traffic conservation assumptions by following a Kalman predict-update structure with physical interpretability. Q-Net employs an AI-augmented Kalman filter to fuse data with differing spatial and temporal resolutions. The work is detailed in arXiv preprint 2509.24725.
Key facts
- Q-Net is a framework for estimating queue lengths at signalized intersections.
- It integrates aggregated vehicle counts from loop detectors and aFCD average speed measurements.
- The framework uses a state-space formulation and Kalman predict-update structure.
- It maintains physical interpretability in state evolution and measurement models.
- Q-Net employs an AI-augmented Kalman filter.
- The research addresses partial observability and violations of traffic conservation assumptions.
- The paper is available on arXiv with ID 2509.24725.
- The announcement type is replace-cross.
Entities
Institutions
- arXiv