Modeling Small-Granularity Expressway Traffic Volumes With Quantum Walks

At small granularity (e.g., 10-minutes to hourly), expressway traffic volumes rely heavily on drivers’ driving habits heterogeneity and decision randomness, making it challenging for accurate modeling. In this paper, we propose a small granularity simulation model named Small-Granularity Expressway Traffic Volumes with Quantum Walks (SGETV-QW). The proposed model adopts quantum walks to generate probability patterns of the exiting time of drivers from the expressway. Then, we refine and map the generated probability patterns to empirical traffic-volume data via a stepwise regression and quantify the modeling accuracy in both the time and frequency domain. We validate SGETV-QW for traffic volume data from seven stations along the Nanjing-Changzhou Expressway in China and compare it with Autoregressive Integrated Moving Average Model (ARIMA) and Long and Short-Term Memory (LSTM) networks. The results show that SGETV-QW improves the simulation accuracy at small granularity. In addition, traffic volumes simulated by SGETV-QW have almost the same frequency spectrum as observed traffic volumes. Finally, we conduct a sensibility analysis and show that SGETV-QW can adapt its parameters to model traffic volumes at different granularities.