Traffic Counts-based Origin-Destination Matrix Estimation using a Traffic Simulator and Machine Learning

Accurate origin-destination (OD) matrices, which represent the travel patterns of individuals or groups between different locations, are vital for traffic planning and logistics. Traditionally, they’re determined through costly surveys, often resulting in outdated data. A newer method involves using traffic sensors, but mapping these counts to OD matrices is challenging, especially when sensor coverage is limited.To tackle this, we propose using traffic simulators like SUMO to create synthetic datasets. Machine learning models are then trained on this data to predict OD matrices from traffic counts. Different machine learning models have been investigated in this paper, we focus on Random Forest models, which consistently outperform existing methods like DFrouter, a SUMO tool for OD matrix estimation, and evaluated on the same test data using the root mean squared error (RMSE) and mean absolute error (MAE) metrics. The proposed Random Forest-based method is shown to consistently outperform DFrouter, achieving average errors three times lower than DFrouter.