Maximum-Likelihood Acceleration Estimation From Existing Roadway Vehicle Detectors

Abstract

Transportation agencies have invested in extensive infrastructure for vehicle detection and speed estimation. Although knowledge of vehicle speeds helps characterize traffic flow, vehicle accelerations can lead to better characterization. Vehicle accelerations are important in designing signal timings with respect to yellow intervals and green extensions for dilemma zone protection. Vehicle acceleration models are also used in studies of vehicle emissions. This paper develops an algorithm that uses existing inductive loops and magnetometers in speed trap configurations to measure acceleration. The algorithm chosen is the maximum-likelihood estimator, given an additive white Gaussian noise model for noise. The algorithm is found to have an error of about 0.02 g.