Linear Programming Model for Estimating High-Resolution Freeway Traffic States from Vehicle Identification and Location Data

Abstract

The estimation of traffic state on freeway segments is widely studied as a complex nonlinear and stochastic estimation problem. A unified representation with a parsimonious explanation for traffic observations under free-flow, congested, and dynamic transient conditions is developed by capturing the essential characteristics of forward and backward wave propagation through cumulative flow count variables. New formulations are presented to use Bluetooth vehicle identification records and GPS vehicle location data on a freeway corridor with a merge and diverge. With the addition of nonnegativity and maximum discharge rate constraints, a computationally efficient linear programming model is constructed to estimate traffic states (i.e., density and traffic flow) from cumulative flow counts at each second. The proposed model is implemented and tested systematically on the basis of a real-world next generation simulation (NGSIM) data set.