Signal timing estimation based on low frequency floating car data

The objective of this paper is the development of a staged methodology that allows the estimation of signal timing information like cycle length, green and red time intervals for time-dependent fixed-time controlled and actuated intersections based on floating car data (FCD). To be able to infer signal timing information based on low frequency FCD, the approach assumes as a basic condition the daily repetition of signal plans, whereby similar daytime and workdays are aggregated to reach a sufficient data density. The established concept utilizes only a very small number of trajectories that covers typical sampling intervals between 15-45 secs. The explained methodology considers three processing stages. Firstly, map matching, data decomposition and stop line estimation are realized in a basic data preparation. Secondly, the method calculates for each trajectory the specific moment, where each trajectory has crossed the inferred stop line position, whereby crossing times are projected by the application of a modulo operation into the time scale of different cycle lengths. A statistical data analysis allows the identification of daytime slices, where signal program stays constant. Finally, the last stage considers the precise estimation of red and green time intervals based on a histogram analysis. The basic applicability of the developed concept has been demonstrated using a simulated trajectory dataset and could be also successfully tested on a real world case study. Data source used in this paper allowed finally the exact estimation of cycle length, whereas red and green time interval could be estimated with an accuracy of +/- 1 secs when comparing estimates against reference signal program documents. (C) 2017 The Authors. Published by Elsevier B.V.