Enhancing Path Flow Estimation on Signalized Arterials with a Hybrid Model: Integrating Sparse Vehicle Data and Automatic Vehicle Identification under Low Coverage

Abstract

The integration of multiple data sources for concurrent traffic flow estimation has garnered significant attention in recent years. Probe vehicle (PV) trajectory data offer complete path information for sampled vehicles, representing partial path flows. Automatic vehicle identification (AVI) data provide precise timestamps and vehicle identities for all recorded vehicles. In this study, we propose a hybrid model, namely the EGLS–EPF model, which combines these two data sources to estimate path flows on urban arterials. This model comprises two sub-models, namely the extended generalized linear square (EGLS) and the extended particle filtering (EPF) models, operating within a novel computational framework. The EGLS submodel leverages both data sources and extends the conventional generalized linear square (GLS) framework, incorporating path flow and travel time as objective terms to iteratively update path flow estimates. The EPF submodel reconstructs individual vehicle paths through probabilistic filtering, using both data sources to establish filtering criteria. The computational framework is designed to improve global estimates by iteratively updating the parameters of both submodels. This approach effectively harnesses the complementary characteristics of the two data sources, enhancing estimation accuracy. Empirical and simulation tests demonstrate that the proposed model consistently achieves more accurate and stable estimations, particularly under conditions of low AVI device coverage and limited penetration rates of probe vehicles, outperforming traditional GLS and particle filtering (PF) models.