A novel AVI sensor location model for individual vehicle path reconstruction on urban road networks

Abstract

Advances in traffic detection technology have enabled the transition from traditional counting detectors to vehicle identification detectors in many cities across the world. This shift has facilitated the collection of high-dimensional network traffic data through Automatic Vehicle Identification (AVI) systems, significantly advancing areas such as path flow estimation and individual vehicle path reconstruction. However, due to the high cost of fully deploying AVI sensors across urban road networks, reconstructing individual vehicle paths with limited sensor data remains a significant challenge. While existing research has focused primarily on Network Sensor Location Problem (NSLP) models for path flow estimation, their application to individual vehicle path reconstruction is less explored. Inspired by system entropy theory in information science, this study introduces a novel metric called Path Reconstruction Entropy (PRE) to quantify the uncertainty in vehicle path choices as observed by AVI sensors. Leveraging this metric, we propose a novel non-linear integer programming AVI-NSLP model that follows the Minimum PRE principle, optimizing the placement of AVI sensors for enhanced vehicle path reconstruction. The model incorporates a composite objective function that considers network benefits, including the uncertainty of vehicle path reconstruction for observed vehicles and coverage of observed vehicles, across two scenarios: with and without prior path flow information. Numerical studies conducted on two networks—a small-sized Sioux Falls network and a large-sized Xujiahui network in Shanghai—demonstrate that the proposed model consistently outperforms existing models in terms of PRE and other sensor deployment metrics on the smaller network and shows a modest advantage on the larger network. The experimental results demonstrate that the AVI sensor placement strategy developed by the proposed model can significantly improve the accuracy of individual vehicle path reconstruction on urban road networks.