Joint Estimation of Signal Timings for all Directions of an Intersection Using Low-Frequency GPS Data

Abstract

This paper studies the signal timing estimation problem for intersections using sparse GPS data. Judging the signal status by the GPS points after the stop line, we first apply Fast Fourier Transform to obtain the cycle length in the frequency domain. Then, we put all GPS data in one cycle length to deal with the shortage of GPS data. Considering the conflicts of signal phases, we jointly minimize the total error to get the actual signal timing of the intersection. This problem is formulated as a Binary-Mix-Integer-Linear-Program, which can be solved efficiently by the standard branch-and-bound method. Finally, the effectiveness and the robustness of our approach are validated by the numerical experiments.