Speed Advisory System Using Real-Time Actuated Traffic Light Phase Length Prediction

Mikhail Burov

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2019-184
December 24, 2019

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-184.pdf

A speed advisory system (SAS) for connected vehicles (CVs) on urban streets is based on the estimation of green (or red) light duration at signalized intersections. A particular challenge is to predict the signal phases of semi- and fully-actuated signals.

In this paper, we introduce an algorithm predicting whether a given CV will be able to make it through the next intersection with an actuated signal on green or not, based on available traffic measurement data. Our mechanism processes traffic data collected from advanced detectors on incoming links and assigns ``PASS"/``WAIT" labels to vehicles according to their estimated ability to go through the intersection within the current phase. Additional computations provide an estimate for the duration of the current green phase that can be used by SAS to minimize fuel consumption.

Simulation results show 95\% prediction accuracy, which yields up to 30\% reduction in fuel consumption when used in SAS. Traffic progression quality also benefits from our mechanism demonstrating an improvement of 20\% at peak for medium traffic demand, reducing delays and idling at intersections.

Advisor: Murat Arcak


BibTeX citation:

@mastersthesis{Burov:EECS-2019-184,
    Author = {Burov, Mikhail},
    Title = {Speed Advisory System Using Real-Time Actuated Traffic Light Phase Length Prediction},
    School = {EECS Department, University of California, Berkeley},
    Year = {2019},
    Month = {Dec},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-184.html},
    Number = {UCB/EECS-2019-184},
    Abstract = {A speed advisory system (SAS) for connected vehicles (CVs) on urban streets is based on the estimation of green (or red) light duration at signalized intersections. A particular challenge is to predict the signal phases of semi- and fully-actuated signals.

In this paper, we introduce an algorithm predicting whether a given CV will be able to make it through the next intersection with an actuated signal on green or not, based on available traffic measurement data. Our mechanism processes traffic data collected from advanced detectors on incoming links and assigns ``PASS"/``WAIT" labels to vehicles according to their estimated ability to go through the intersection within the current phase. Additional computations provide an estimate for the duration of the current green phase that can be used by SAS to minimize fuel consumption.

Simulation results show 95\% prediction accuracy, which yields up to 30\% reduction in fuel consumption when used in SAS. Traffic progression quality also benefits from our mechanism demonstrating an improvement of 20\% at peak for medium traffic demand, reducing delays and idling at intersections.}
}

EndNote citation:

%0 Thesis
%A Burov, Mikhail
%T Speed Advisory System Using Real-Time Actuated Traffic Light Phase Length Prediction
%I EECS Department, University of California, Berkeley
%D 2019
%8 December 24
%@ UCB/EECS-2019-184
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-184.html
%F Burov:EECS-2019-184