Abstract : Traffic flow dynamics in the vicinity of urban arterial intersections is a complex and nonlinear phenomenon, influenced by factors such as signal timing plan, road geometry, driver behaviors, etc. Predicting such flow dynamics is an important task for urban traffic signal control and planning. Current methods use microscopic simulation for studying the impact of a large number of signal timing plans at each of the intersections. A major drawback of microscopic simulation is that they are they are based on source destination traffic generation models and cannot incorporate the high resolution loop detector data such as that are provided by automated traffic signal performance measures (ATSPM) based systems. The arrival (or departure) information of each vehicle on a detector can be thought of as a time series waveform. Given the high granularity of ATSPM data, this waveform can be used to several interesting analyses. The waveforms can be used to derive information on platoon dispersion as vehicles progress across the corridor. Also, these waveforms can be modelled to understand how the vehicles progress across the corridor for a variety of signal timing plans. In this paper, we show that deep neural networks based machine learning systems can be used to effectively leverage the waveforms collected at multiple sensors (stopbar and advanced) on the intersection to model the traffic dynamics both at an intersection and across intersections. We show that modelling of these waveforms can be useful to understand traffic flow dynamics under different signal timing plans and can be potentially integrated into signal timing optimization software. Further, these methods are three to four orders of magnitudes faster than using microscopic simulations.
The arrival (or departure) information of each vehicle on a detector can be thought of as a time series waveform . Given the high granularity of ATSPM data, this waveform can be used to derive information about platoons (multiple vehicles passing without significant distance) or gaps (no vehicles passing through for a duration). In this paper, we show that deep neural networks based machine learning systems can be used to effectively leverage the waveforms collected at multiple sensors (stopbar and advanced) on the intersection to model the traffic dynamics both at an intersection and across intersections. In particular, using these waveforms at stopbar and advanced detectors:
- We develop models to both impute the traffic waveform from each inbound direction to an intersection as well as the traffic waveform to each outbound direction from the intersection conditional on the signal timing plan. The input waveforms in all the directions can be used to understand if the current signal timing is near optimal. The models can be used to model the vehicle progression to downstream intersections and estimate performance measures for different signal timing plans. However, since they use data that is directly measured based on the traffic sensors in the network, they are a much more accurate indicator of traffic movement than imputed origin destination pairs.
- We develop models that can predict the dispersion along a road segment (exit from one intersection to entrance of the neighboring intersections) more accurately than the Robertson model that is traditionally used. This is because our models, like microscopic simulation models, can effectively capture non-uniform velocities of vehicles as well variation in driver behaviors.
- We develop models that can predict the impact of signal timing of the downstream intersection on platoons as they arrive close to a downstream intersection. The output waveform from a given intersection along with the output waveform on the downstream also be used to understand the leakage or addition of traffic during a short time period.
Thus, our models can capture both the local (i.e., near an intersection) traffic flow dynamics as well as coupled traffic flow dynamics (i.e., between two consecutive intersections) and are significant extensions of the prior work on platoon dispersion models. We develop and provide multi-scale error measures to demonstrate that our predictions are accurate and comparable to microscopic simulation.
Our models use novel deep learning based architecture with attention layers and teacher forcing that is specifically designed to model and predict the behavior of input and output behavior at an intersection using advance and stop bar high resolution loop detector data and signal timing information. The use of our GPU implementation of deep neural networks can generate accurate predictions at three to four orders of magnitude faster than using microscopic simulations for this purpose.