Physics-informed Learning for Identification and State Reconstruction of Traffic Density

This paper deals with traffic density reconstruction using measurements from Probe Vehicles (PVs). The main difficulty arises when considering a low penetration rate, meaning that the number of PVs is small compared to the total number of vehicles on the road. Moreover, the formulation assumes noisy measurements and a partially unknown first-order model. All these considerations make the use of machine learning to reconstruct the state the only applicable solution. We first investigate how the identification and reconstruction processes can be merged and how a sparse dataset can still enable a good identification. Secondly, we propose a pre-training procedure that helps the hyperparameter tuning, preventing the gradient descent algorithm from getting stuck at saddle points. Examples using numerical simulations and the SUMO traffic simulator show that the reconstructions are close to the real density in all cases.