While using any existing method for estimating traffic queues at signalized intersections
that are in close proximity to railroad grade crossings, it is extremely important to know the
specific limitations of the method, and the factors that can have a significant impact on the
resulting queue estimate. An adequate analysis of traffic queues at such location is imperative in
order to implement safe design that will minimize the risk of crashes. Such queuing analysis
becomes even more critical where direct observations of traffic queues are not possible or where
To investigate the limitations of current queue estimation methods, their sensitivity to
various traffic factors, and to help select an appropriate method for determining reliable
estimates of queue lengths to be used for preemption evaluation, this research analyzed and
compared different currently used analytical and simulation based methods including HCS,
Syncho, Sim-Traffic and VISSIM. The comparison provided in this chapter can help
understanding the appropriate application of these methods in determining queue estimates at
signalized intersections near rail-highway grade crossing for preemption evaluation.
The queue length estimates at signalized intersections from all the above mentioned
methods were compared for various traffic volumes, vehicle mix and signal control parameters.
Impact of different lane configurations and signal phasing was also included in the analysis. The
queue length parameter used in this analysis is 95th percentile queue length. Description on the differences between in each method on defining and computing queue lengths, and other
assumptions on vehicle arrival pattern, car-following parameters, vehicle characteristics, delay,
capacity and LOS computations etc. are provided in detail in this chapter. A few findings are
presented below.
Comparing the values obtained from each method, HCS computed queue estimates are
generally higher than all other methods particularly at under-saturated traffic conditions and for
simple traffic operations (such as single thru lane and no turning traffic) where various traffic
factors such as vehicle mix, speed, interaction between different lanes, and overflow queues do
not have a significant impact on vehicular delays and queues. Similarly for such traffic
operations, values obtained from the other analytical models (i.e. Synchro and Railroad
Assessment Tool) were also on the higher side compared to those estimated by simulation
For left-turn queues, deterministic analytical models HCS, Synchro, and Railroad
Assessment Tool were found providing lower extent for left-turn queues than simulation models
particularly for 0.5>v/c>1.0; at such traffic conditions HCS estimates on left-turn queues were
lower than Sim-Traffic while Synchro estimates were lower than VISSIM. Sim-Traffic estimates
on left-turn queues were significantly higher than all other methods.
The values computed from HCS are generally higher than other analytical tools i.e.
Synchro and Railroad Assessment Tool however, HCS and other macroscopic analytical
methods are significantly inadequate in accounting for various traffic factors that can have a
significant impact on queue estimation. For example, the HCS procedure cannot account for
vehicle-mix, speeds, lane overflow/blockage, impact of residual queues, and complex traffic
operations. Synchro is also a macroscopic model and does not account for many of the above
factors including vehicular interactions between lanes, complex left turn phasing, and spill back
beyond turning bays. Microscopic-simulation models (Sim-Traffic, VISSIM) have the capability
to account for the aforementioned traffic factors in computing delays and obtain information on
individual vehicles position, speed, deceleration, acceleration etc. per small increment of time
(per second or less than a second). In addition, these modeling techniques provide calibration of
various parameters to capture the actual traffic conditions at a location.
Two simulation models Sim-Traffic and VISSIM were also analyzed and compared in
this research. The differences in queue output between both the methods are attributed to the
various differences in factors that impact the resulting queue lengths. These include differences
in functions describing arrival and discharge rates, headway distributions, acceleration functions,
default parameters used for driving behaviors, vehicle characteristics, and factors used to model
also attributed to the different percentile computation methods used in both the methods. It is
also important to know that the traffic flow model and parameters defaults describing the traffic
flow, driver behaviors, vehicle lengths and characteristics etc. in VISSIM are based on the
studies conducted in Germany which may not adequately represent (unless calibrated) US
driving behaviors and vehicle characteristics.
Sim-Traffic simulation models estimates are generally higher than those obtained from
VISSIM using the default values for the aforementioned factors. It was also found that using the
default model parameters for car-following, driver behaviors, and vehicle characteristics etc.,
VISSIM model provides higher capacity and lower queue estimates as compared to Sim-Traffic,
and all other analytical models HCS, Synchro particularly for through lane queus. The
comparison of various queue estimates obtained from each method in this chapter showed that
the difference in the queue estimates among these methods becomes more significant as the
demand reaches approximately at/above 50% of the capacity (0.5≥v/c<1).
A comparison of the maximum queues observed at a local site was performed as part of
this research using the data on intersection geometry, traffic volumes, signal control, speeds and
heavy vehicles. No other model calibration was done. The results showed Sim-Traffic queue
estimates slightly above the observed maximum no. of vehicles, up to 1 passenger car. The 95th percentile estimate from Sim-Traffic also found to be close to the maximum observed queue
(particularly left-turn queue). The no. of vehicles estimated through VISSIM were lower than the
4.20 Recommended Procedure for Estimating Queue Lengths for Preemption Evaluation