5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S.

Held

This website uses cookies

This site uses cookies and other technologies to provide you with a more responsive and personalized service. We may share
information with third-parties to assist us in these efforts. To learn more about cookies, including how to disable them, view our
Online Privacy Notice. By clicking Accept or using this site you consent to our use of cookies.

Necessary (Required) Statistics Marketing

DETAILS ACCEPT

MENU

INSIGHTS

A Naturalistic Study of Vehicle Acceleration and

Deceleration at an Intersection

J.S. Held Strengthens Construction Advisory Practice in EMEA with Aquila Forensics Acquisition
READ MORE

HOME · INSIGHTS · ARTICLES

By David Danaher & Sean McDonough & Drew Donaldson & Reece Cochran

 
https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 1/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

The material in this paper was researched, compiled, and written by J.S. Held. It was originally
published by SAE International.

Abstract

Approaching an intersection and braking to a stop, as well as accelerating from a stop, is a common
occurrence in daily life. While the experience is routine, the actual rate of deceleration and acceleration
has not been analyzed from an orthogonal aerial perspective. The aerial perspective provides video
footage that allows for accurate planar motion tracking and does not influence the drivers’ actions in any
way. This paper examines the behavior of drivers at two separate signal light controlled intersections to
determine both the rate at which they slow down to a stop, and also the rate at which they accelerate
through the intersection after a signal change. The paper will also address the acceleration rate
differences of vehicles who are first to reach the intersection in comparison to those that are directly
behind another vehicle, as well as the lag in reaction between vehicles as they begin to accelerate from a
stop.

Introduction

In the field of accident reconstruction, determination of a stopped vehicle’s acceleration from an

intersection is often required for analysis of time-space relationships relating to a collision. Similarly, the
deceleration rates of a vehicle approaching a yellow or red light at an intersection may be required. A
single acceleration value is commonly assigned to a vehicle throughout its entire acceleration sequence,
which may under- or overestimate the vehicle’s speed at points through the motion analysis. The
acceleration of passenger vehicles from a stop has been extensively studied in peer reviewed literature,
however there is little research relating to the acceleration and deceleration of passenger vehicles at
controlled intersections in a naturalistic style. In addition, no studies were found which compare the
accelerations of vehicles in either the first or second row position at an intersection, nor the lag time in
acceleration for second row vehicles.

Fugger, et al., [1] conducted a study in which the “real world” perception reaction times and accelerations
for stationary vehicles reacting to a signal change were analyzed. For this study, data was captured of 100
vehicles at two intersections with a 120Hz digital camcorder placed at ground level. The timing of vehicles
crossing each intersection were used to determine the driver perception/reaction times and vehicle
accelerations. They found that the acceleration could be modeled with two phases: the initial phase
lasted for a mean time of 0.94 seconds with an acceleration rate of 0.06 g, and the secondary phase was
characterized by an acceleration rate of 0.22 g.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 2/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Chang, et al., [2] studied driver behavior to changes in traffic signals when approaching an intersection. In
this study, Chang utilized data from 579 stopping vehicles with varying roadway and traffic conditions and
found that the mean deceleration rate for stopping vehicles was 9.5 ft/sec2, or approximately 0.295 g.
Chang noted that the observed deceleration rate “is primarily the result of the driver’s selection of
comfort. It is not an indication of whether they can perform certain deceleration rates.”

Haas, et al., [3] analyzed driver acceleration and deceleration behavior at stop sign-controlled
intersections, which included 299 deceleration events from 24 drivers and 214 acceleration events from
22 drivers. Haas concluded that deceleration rates varied based on the stopping vehicle’s initial speed,
ranging from 0.1 g for initial speeds between 20 and 25 mph to an average deceleration rate of 0.18 g for
initial speeds between 35 and 40 mph. The study concluded an average acceleration rate of 0.14 g
regardless of final speed.

Wortman, et al., [4] studied driver behavior associated with a traffic signal at six intersections during both
day and night. Data was collected using a super 8 mm movie camera recording at 18 frames per second
placed near the intersection. The study reported mean deceleration rates ranging from 7.0 ft/sec2 (0.22 g)
to 13.9 ft/sec2 (0.43 g) for the various intersections. The mean deceleration for all intersections was 11.6
ft/sec2 (0.36 g). Additionally, the study found the difference in daytime and nighttime deceleration rates
was not statistically significant.

Roenitz, et al, [5] studied the characteristic behavior of drivers decelerating for known hazards. Data was
collected using a battery-operated radar gun with raw data being stored to an interfacing computer. A
total of 313 decelerations were recorded at a variety of stop sign controlled intersections in varying
weather conditions. Roenitz concluded that vehicle deceleration could be modeled with three phases:
The “Initiation Phase,” lasting approximately 3 seconds, was characterized by a deceleration rate of 0.064
g. An average deceleration rate of 0.216 g was determined for the “Active Braking Phase,” which included
most of a vehicle’s deceleration. The final phase discussed is the “Completion Phase,” in which vehicles
dissipate their final 3.4 km/h of speed, lasting approximately 1.7 seconds.

Kodsi and Muttart, [6] conducted a study in which the naturalistic acceleration profiles of passenger
vehicles were observed at two-way-stop controlled intersections. Data was captured using a video
camera recording at 30 frames per second placed perpendicular to the vehicle’s direction of travel.
Vehicles were tracked until they were either out of camera frame or up to 15 meters from the stop line. 244
vehicles were tracked, with 179 of them coming to a complete stop before proceeding through the
intersection. The authors found that the acceleration could be modeled with two phases: the initial phase
lasted for approximately 0.90 seconds with an acceleration rate of 0.07 g, and the second phase lasted for
approximately 2.1 seconds with an acceleration rate of 0.25 g.

This paper evaluates data collected from two signal light controlled intersections to determine the
acceleration and deceleration profiles of passenger vehicles in a naturalistic scenario. This analysis is

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 3/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

intended to expand on the testing referenced above and derive acceleration and deceleration profiles for
accident reconstruction purposes. This study also expands on previous research by including factors
relevant to the field of accident reconstruction, such as whether the vehicle is the first in line or the
vehicle directly behind, and the vehicle’s speed when approaching the controlled intersection.

A data collection method for vehicles was presented in the 2019 SAE publication by Carter, et al., in which
the authors developed a video tracking technique for automated tracking of left turning vehicles from
overhead video footage, which was recorded with a small unmanned aerial system (sUAS) [7]. The 2019
paper validated the video tracking process with VBOX data to verify its accuracy. This analysis method
was also used in a 2020 paper published by the authors of this paper, which presented acceleration
models for left-turning heavy trucks [8].

The aerial video tracking process was utilized for this research to collect data from two urban
intersections, which captured 178 accelerating and 212 decelerating vehicles that were used for analysis
of time, distance, speed, and longitudinal acceleration and deceleration. The acceleration tests include 95
vehicles which were first to reach the intersection and 83 vehicles which were second to reach the
intersection. Additionally, 85 of the 178 accelerating vehicles were positioned on the inside lane (lane 1).
The deceleration tests include 111 vehicles which were first to reach the intersection, and 101 vehicles
which were second to reach the intersection. 104 of the 212 decelerating vehicles were positioned on the
inside travel lane (lane 1). The rest of the tracked vehicles were positioned in the outside travel lane (lane
2), as shown in Figure 6.

Data Collection

Multiple intersections in the greater Denver metropolitan area were inspected for suitability for data
collection; two were selected using the following criteria: sufficient flow of passenger vehicle traffic,
multiple travel lanes, and no airspace restrictions for longer duration sUAS flights for video capture. The
two intersections that were selected for data collection are listed below in Table 1.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 4/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Table 1 - Two intersections selected for data collection

Figure 1 and Figure 2 show aerial images of each intersection taken from a DJI Mavic 2 Pro sUAS during
data collection.

Figure 1 - 56th Ave and Quebec St.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 5/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 2 - 92nd Ave and Federal Blvd.

56th Ave and Quebec St form a mid-sized, controlled intersection in Commerce City, CO, which has a
posted speed limit of 45 mph for northbound and southbound traffic. There are three through travel lanes
for both northbound and southbound directions, separated by a concrete median, and there are double
left turn lanes for both directions at 56th Ave. Areas where data was collected for accelerating and
decelerating vehicles are shown outlined in Figure 3, along with a label for the intersection width, which
was measured between the stop lines of the innermost southbound travel lane. The red section indicates
the tracking area for decelerating vehicles while the green area indicates the tracking area for
accelerating vehicles. The orange area represents the overlap between acceleration and deceleration
zones.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 6/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 3 - 56th Ave and Quebec St test area

92nd Ave and Federal Blvd is a mid-sized, controlled intersection in Westminster, CO, which has a posted
speed limit of 45 mph for northbound and southbound traffic. There are three straight travel lanes for
both northbound and southbound directions, separated by a concrete median, and there are double left
turn lanes for both directions at 92nd Ave. Areas where data was collected for accelerating and
decelerating vehicles are shown in Figure 4, along with a label for the intersection width, which was
measured between the crosswalks of the innermost southbound travel lane. The red section indicates the
tracking area for decelerating vehicles while the green area indicates the tracking area for accelerating
vehicles. The orange area represents the overlap between acceleration and deceleration zones.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 7/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 4 - 92nd Ave and Federal Blvd Test Area

Data was collected at each intersection with a DJI Mavic 2 Pro sUAS. All flights were operated in Class G
airspace in accordance with the Federal Aviation Administration (FAA) Small Unmanned Aircraft
Regulations, Part 107, by a licensed remote pilot. The sUAS was flown at 400 feet above ground level, and
video was recorded with the camera facing 90 degrees downward to generate video as close to
perpendicular to traffic flow as possible.

Data Processing

Overhead video captured from each intersection was reviewed and clips were exported of each
accelerating and decelerating vehicle for input into the video tracking process. Video footage was
reviewed and tracked vehicles which could not be classified as passenger vehicles were removed from
the analysis. These excluded vehicles include box trucks, heavy trucks, and motorcycles. For acceleration,
the start of a given tracked run was identified as when the vehicle first began to move forward across the
intersection from its initial stopped position. The end of a run was determined by when the vehicle began
exiting the frame after proceeding through the intersection. For deceleration, the start of a given tracked
run began when the vehicle first entered the frame and ended when it first reached a complete stop. The
video tracking process outputs are a video labeled with tracked objects and a text file containing those
vehicles’ X and Y coordinates for each video frame, in pixels. The point at which passenger vehicles are

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 8/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

tracked is approximately centered over the vehicle. Figure 5 shows a frame from the video output with
labeled tracked objects from the 56th Ave and Quebec St intersection.

Figure 5 - Video frame from video tracking output video with object labels

The tracking data for each passenger vehicle was then processed utilizing the same techniques as
presented in SAE 2019-01-0421 and 2020-01-0882, such as a two-pass Butterworth filter and scaling from
pixels to feet. Vehicle data was categorized based on the intersection, whether the vehicle was
accelerating or decelerating, and if the vehicle was in the first or second row at the intersection.

Data Analysis

All runs were processed, and calculations were performed to generate the average and peak longitudinal
accelerations and decelerations for each tracked vehicle. Vehicles were categorized based on whether
they were leaving or approaching the intersection and whether they were the first or second car to reach
the intersection in their lane. Vehicles in the left and right turn lanes were excluded from the analysis due
to inconsistent interference caused by other turning vehicles. Figure 6 shows lane and row designations
for tracked vehicles at the 56th and Quebec intersection, with the analyzed travel lanes highlighted in
green and omitted lanes highlighted in red. For this research paper, only first and second row vehicles
were tracked and analyzed at each intersection.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 9/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 6 - Vehicle Row and Lane Designations

Once the tracked vehicles’ accelerations and decelerations were determined for each dataset, the values
were plotted against all other runs to analyze the effects of row position on the average and peak
longitudinal accelerations and decelerations. Distinctions were also made for variations between the two
intersections. The following sections outline the data analysis process.

Acceleration

Speed data for all tracked accelerating vehicles was plotted with respect to time, starting with the frame
the vehicle began moving. Figure 7 shows the average tracked vehicle speeds for all acceleration runs.
Individual runs are shown in gray, the average of all tracked runs is indicated with a solid black line, and
the area between the dotted black lines represent 70 percent of the data, with the tests outside of the
dotted black lines representing the top and bottom 15 percent of the data.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 10/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 7 - Average Speed vs. Time (Acceleration)

56th and Quebec

Tracked runs were divided into two datasets based on the intersection in which the data was recorded,
preventing any influence of the intersection on effects caused by differences in vehicle row. Figure 8
shows the average speeds of all tracked accelerating vehicles over time at the 56th and Quebec
intersection.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 11/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 8 - Average Speed vs. Time for All Vehicles (56th and Quebec, Acceleration)

Runs from the 56th and Quebec intersection were then subdivided into two groups: vehicles that were
first to reach the intersection, and vehicles directly behind the lead vehicle at the intersection. Division of
runs based on row allows for analysis of the effect of row on the acceleration of vehicles from an
intersection, such as the authors’ hypothesis that the second row vehicles’ acceleration and deceleration
rates are tied to the rates of the vehicles preceding them in the first row. Figure 9 and Figure 10 show the
speed versus time plots for first and second row vehicles, respectively.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 12/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 9 - Average Speed vs. Time for First Row Vehicles (56th and Quebec, Acceleration)

Figure 10 - Average Speed vs. Time for Second Row Vehicles (56th and Quebec, Acceleration)

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 13/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

When analyzing the motion of a vehicle accelerating from a stop, an accident reconstructionist typically
uses the vehicle’s distance traveled to determine its speed at impact. In this study, vehicles positioned in
the second row were tracked for a greater distance than first row vehicles due to their starting position in
the video frame. For further analysis of the effect of row on acceleration, plots were made which compare
the distance traveled for tracked vehicles to their acceleration time. Figure 11 shows the average distance
traveled for first row vehicles based on their acceleration time. Figure 12 shows a similar plot for all
second row vehicles.

Figure 11 - Average Speed vs. Distance for First Row Vehicles (56th and Quebec, Acceleration)

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 14/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 12 - Average Speed vs. Distance for Second Row Vehicles (56th and Quebec, Acceleration)

92nd and Federal

Acceleration data collected from 92nd and Federal was separated and processed in the same manner as
data the data collected from 56th and Quebec. The average speed versus time plot for all tracked vehicles
from 92nd and Federal is shown in Figure 13.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 15/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 13 - Average Speed vs. Time for All Vehicles (92nd and Federal, Acceleration)

Runs from 92nd and Federal were split into first and second row vehicles as described with the data from
56th and Quebec. Figure 14 and Figure 15 show the average speed versus time plots for the accelerating
first and second row vehicles, respectively.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 16/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 14 - Average Speed vs. Time for First Row Vehicles (92nd and Federal, Acceleration)

Figure 15 - Average Speed vs. Time for Second Row Vehicles (92nd and Federal, Acceleration)

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 17/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Average speed versus distance plots were also created from the data for accelerating first and second
row vehicles at 92nd and Federal, which are shown in Figure 16 and Figure 17.

Figure 16 - Average Speed vs. Distance for First Row Vehicles (92nd and Federal Acceleration)

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 18/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 17 - Average Speed vs. Distance for Second Row Vehicles (92nd and Federal, Acceleration)

Deceleration

Speed data for all tracked decelerating vehicles was plotted with respect to time, starting when the
vehicle entered frame and ending when the vehicle came to a stop at the intersection. Figure 18 shows
the average tracked vehicle speeds for all deceleration runs over time. As with the acceleration plots,
individual runs are shown in gray, the average of all tracked runs is indicated with a solid black line, and
the area between the dotted black lines represent 70 percent of the data, with the tests outside of the
dotted black lines representing the top and bottom 15 percent of the data.

Figure 18 - Average Speed vs. Time for All Vehicles (Deceleration)

56th and Quebec

Tracked runs were divided into two sets based on the intersection in which the data was recorded,
preventing any influence of the intersection on effects caused by differences in vehicle row. Figure 19
shows the average speeds of all tracked decelerating vehicles at the 56th and Quebec intersection.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 19/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Deceleration runs from the 56th and Quebec intersection were then subdivided into two plots for
vehicles which were first to reach the intersection (row 1) and vehicles which were second to reach the
intersection (row 2). Figure 20 and Figure 21 show the speed versus time plots for first and second row
vehicles, respectively.

Figure 19 - Average Speed vs. Time for All Vehicles (56th and Quebec, Deceleration)

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 20/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 20 - Average Speed vs. Time for First Row Vehicles (56th and Quebec, Deceleration)

Figure 21 - Average Speed vs. Time for Second Row Vehicles (56th and Quebec, Deceleration)

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 21/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

92nd and Federal

Deceleration data collected from 92nd and Federal was separated and processed in the same manner as
data the data collected from 56th and Quebec. The average speed versus time plot for all tracked vehicles
from 92nd and Federal is shown in Figure 22.

Figure 22 - Average Speed vs. Time for All Vehicles (92nd and Federal, Deceleration)

Runs from 92nd and Federal were split into first and second row vehicles as described with the data from
56th and Quebec. Figure 23 and Figure 24 show the average speed versus time plots for the decelerating
first and second row vehicles, respectively.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 22/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 23 - Average Speed vs. Time for First Row Vehicles (92nd and Federal, Deceleration)

Figure 24 - Average Speed vs. Time for Second Row Vehicles (92nd and Federal, Deceleration)

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 23/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Discussion

Acceleration

Inspection of the acceleration plot for all tracked runs show a wide range of speeds reached during
tracking, ranging from approximately 25-45 mph; however, the data shows a clear trend of acceleration
with multiple different phases. A few of the vehicles fell well outside the norm or showed erratic
accelerations mostly due to the software losing the vehicle for a period of time before reacquiring them.
Therefore, in analyzing the data, these vehicles were removed from the data sets.

Separation of the data based on the intersection from which the data was collected produces a narrower
range of speeds respective to the size of the intersections. Further separating the data into groups based
on vehicles stopped in the first row versus vehicles stopped in the second row at the intersection
demonstrates the differences in actions from leading drivers versus following drivers, who are
constrained by the leading vehicle. While separating the data into subsets generates specific trends for
each classification at a given intersection, the overall acceleration profile of all tracked runs closely
follows those of the subsets.

Examination of the speed versus time plot for all tracked acceleration runs shows multiple distinct
phases of acceleration. In order to most closely approximate the average acceleration presented in the
plots, the acceleration was broken into four separate time ranges: 0-1 seconds, 1-3 seconds, 3-5 seconds,
and 5-7 seconds. The initial phase of acceleration, taking place within the first second of movement, is
characterized by a 0.12 g acceleration rate. During this time, the driver is likely transferring from the brake
pedal to the gas pedal, which allows the vehicle to roll forward with minimal acceleration at or near
engine idle speed for some portion of the first second of acceleration. The second phase of acceleration
(1-3 seconds) is characterized by an acceleration of 0.26 g. Most vehicles tracked during this study
recorded their peak acceleration during the second phase. As time progresses beyond the second phase,
the rate of acceleration for tracked vehicles begins to decrease, which first becomes evident in the third
phase of acceleration (3-5 seconds). During this phase, tracked vehicles had an average acceleration rate
of 0.22 g, which is less than the peak acceleration seen during the second phase. In the fourth phase of
acceleration (5-7 seconds), vehicles began approaching the speed limit, resulting in a drop in the average
rate of acceleration to 0.17 g. Figure 25 shows the speed versus time data for all tracked runs, with the four
phase acceleration model trendline shown with different colors for each phase.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 24/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 25 - Acceleration Model Trendline

The acceleration model trendline is shown in Figure 26 with plots of individual test runs removed for
clarity.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 25/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Figure 26 - Acceleration model trendline, individual tests removed.

Comparison of the acceleration profiles for vehicles in the first row versus vehicles in the second row
show slight variations in acceleration over all four phases. As seen in the plots, vehicles accelerating from
the first row tended to have a higher rate of acceleration than those starting from the second row. The
authors hypothesize that this change in acceleration results from the first row drivers having a clear view
and lane for accelerating, while drivers in the second row are limited by the acceleration of the vehicle in
front of them and do not have a clear view of the road ahead. As vehicles began approaching the speed
limit, the variation in accelerations became negligible. When analyzing the data from both intersections,
vehicles in the first row accelerated to approximately 32 mph within the first 7 seconds of tracking, while
second row vehicles accelerated to approximately 28 mph. As can be seen in Figure 14, many of the first
row vehicles at 92nd and Federal were not tracked for a full 7 seconds. This is likely due to the shorter
distance required for first row vehicles to clear the intersection in comparison to second row vehicles. To
account for the loss of data, it was determined that using a time range of 5-6.4 seconds for the final phase
of acceleration most accurately represented the data for those vehicles.

Separating data by intersection reveals similar trends as seen when analyzing all tracked runs. For the
56th and Quebec intersection, first row vehicles accelerated to approximately 36 mph during the first 7
seconds, and second row vehicles accelerated to approximately 31 mph. Fewer vehicles at the 92nd and
Federal intersection were tracked for a full seven seconds of acceleration, however variations in
acceleration profiles for first and second row vehicles can be seen during the first five seconds. At 5
seconds, first row vehicles had accelerated to approximately 23 mph, while second row vehicles had
reached approximately 19 mph. Tables detailing acceleration rates for all four phases are provided in
Appendix A.

It should be noted that driver actions can contribute to outlier data points. The accelerations calculated
in this study are not representative of the maximum accelerations these vehicles are capable of, but
instead demonstrate the naturalistic acceleration of drivers operating within their comfort zone. Higher
rates of acceleration are possible depending on the actions of the driver; drivers may also accelerate at a
lower rate if outside factors are affecting their concentration or comfort level. Should a reconstructionist
have knowledge of the driving style of the driver and any roadway factors such as construction zones,
darkness, or traffic, such outlier data can be considered as valid should the situation require.

Deceleration

Inspection of the deceleration plot for all tracked runs show a wide range of initial speeds, ranging from
approximately 20-50 mph. This wide range of initial speeds is partially the result of inconsistencies in the
beginning of deceleration for vehicles tracked during this study. Some vehicles began their deceleration
prior to entering the frame, and therefore were not tracked during the first portion of their deceleration.
However, this inconsistency does not detrimentally affect the data, as many vehicles began their

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 26/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

deceleration after entering the frame and the overall range of speeds narrows after the first second of
tracking. Furthermore, outlier data as previously discussed was removed from analysis. For example, one
test vehicle began decelerating from approximately 65 mph, far exceeding the road speed limit. This test
was removed as it did not accurately depict a naturalistic braking scenario for most drivers.

Analysis of the data shows a linear rate of deceleration during the first 6 seconds of tracking. After 6
seconds, many vehicles had nearly come to rest, with vehicles continuing to decelerate at a lower rate.
For the first 6 seconds of tracking, the dataset shows an average deceleration rate of 0.19 g. The phase
from 6 to 8 seconds is characterized by a deceleration rate of 0.09 g. While some test vehicles continued
to decelerate after 8 seconds, restrictions in the effectiveness of the data tracking process at very low
speeds mean that the data collected may be unreliable. Therefore, data after 8 seconds was not used for
determining deceleration profiles for this study.

Comparison of the deceleration profiles for first and second row vehicles show similar linear trends,
however variations can be noted in the initial speed and overall rate of deceleration. Vehicles decelerating
from the first row tend to have a higher initial speed than vehicles which are second to reach the
intersection, resulting in a higher rate of deceleration for first row vehicles. Vehicles in the second row
have a slightly lower initial speed, and thus have a lower rate of deceleration. The deceleration rates for
first and second row vehicles were 0.22 g and 0.17 g, respectively. Figure 27 shows the speed versus time
plot for all tracked decelerating vehicles, with the linear trendline shown in black.

Figure 27 – Deceleration Model Trendline

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 27/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

The deceleration model trendline is shown in Figure 28 with plots of individual test runs removed for
clarity.

Figure 28 - Deceleration model trendline, individual tests removed.

It should be noted that driver actions can contribute to outlier data points and variations in deceleration
rates between vehicles in the first and second rows. First row vehicles encountering a changing traffic
signal are forced to decide whether to stop or proceed through the intersection. Vehicles positioned in
the second row typically have more time to react to the changing traffic signal and are also influenced by
the actions of the vehicle in front of them. Should a reconstructionist have knowledge of the traffic light
condition at the time of the incident along with the driver’s actions resulting from the signal change, such
outlier data can be considered as valid.

Analysis of previous vehicle acceleration and deceleration research has shown that data can typically be
extrapolated to include speeds and distances that may not have been included in the study, assuming the
roadway conditions and other variables remain consistent. This extrapolation is possible due to the
comfort level of drivers. While a passenger vehicle is capable of decelerations far exceeding those
determined in this paper, drivers in a naturalistic scenario are most likely to decelerate at a rate which
they feel comfortable. Therefore, a driver will most likely decelerate at approximately the same rate
independent of their initial speed, allowing for the deceleration data to be extrapolated. It is important to
note that this extrapolation will only influence the time and travel distance of the first deceleration phase.
The second deceleration phase (6-8 seconds) includes the final adjustments in braking required for a
https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 28/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

driver to come to a controlled stop either at the stop line or behind another vehicle, which remained
consistent regardless of the tracked vehicle’s initial speed in the captured runs.

Acceleration Lag Time

As a secondary consideration, data was analyzed to determine the time elapsed from when the first
vehicle at an intersection initiates its acceleration to when the vehicle directly behind begins
accelerating. The lag time between the two vehicles accelerations can assist a reconstructionist in
completing a time-space analysis for a vehicle stopped at an intersection. 93 vehicles from the dataset
were compared to determine the lag time in accelerations. The average calculated lag time before second
row vehicles began accelerating was approximately 1.1 seconds. Large variations in the calculated lag
time resulted in a standard deviation of 0.69 seconds.

Summary/Conclusions

Trends relating to both straight line accelerations and decelerations were evident upon analysis of the
passenger vehicle video tracking data, as shown in this paper.

The average accelerations for passenger vehicles were divided into four phases to accurately follow the
acceleration profile determined from the test data. The average accelerations calculated for each phase
are listed in Table 2. Also listed in Table 2 is the percent decrease in acceleration row 2 vehicles displayed
in comparison to row 1 vehicles. Tables of calculated acceleration and deceleration values are provided in
Appendix A.

Table 2 - Average Accelerations For Each Phase

Analysis of the processed data shows trends relating to the accelerations for first and second row
vehicles. First row vehicles accelerated at a greater rate than second row vehicles throughout the entire
tracking period. Additionally, comparison of data collected at the two intersections show slightly higher
average rates of acceleration for vehicles at 56th and Quebec than vehicles from 92nd and Federal. The
average of all processed runs resulted in an acceleration rate of 0.12 g for the first second, 0.26 g from one
to three seconds, 0.22 g from three to five seconds, and 0.17 g from five to seven seconds.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 29/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

The average decelerations for passenger vehicles were determined to follow a linear trend. The
deceleration averages calculated for all runs are listed in Table 3. Also listed in Table 3 is the percent
decrease in deceleration row 2 experienced compared to row 1.

Table 3 - Deceleration Data

Analysis of the processed data shows trends similar to those found from the captured acceleration data.
First row vehicles decelerated at a greater rate than those in the second row. Differences in deceleration
were noted when comparing data from the two separate intersections, however the differences in
deceleration rate were minor. The average of all processed runs resulted in a deceleration rate of 0.19 g
from 0 to 6 seconds, and a deceleration rate of 0.09 g for 6 to 8 seconds.

These calculations can be applied directly to accident reconstruction to determine the time, distance,
and speed required for passenger vehicles to either traverse or brake to a stop at a controlled
intersection.

Acknowledgments

We would like to thank David Danaher, Sean McDonough, Drew Donaldson & Reece Cochran for
providing insight and expertise that greatly assisted this research.

David Danaher is a Senior Engineer in J.S. Held’s Accident Reconstruction Practice. His casework
involves major mechanical systems, heavy equipment, and motor vehicles. As a professional mechanical
engineer, Mr. Danaher has experience in the areas of mechanical design and operation as well as failure
analysis. His experience involves product design evaluation, strength of materials analysis, and failure
mechanics. Mr. Danaher has also analyzed mechanical components that are part of larger industrial
systems to demonstrate their operation as well as any related failures, at locations both nationwide and
internationally. The analysis performed in these industrial systems is then illustrated in a three-
dimensional environment, breaking down individual components to visually show the mechanics of the
operation.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 30/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

David can be reached at [email protected] or +1 303 733 1888.

Sean McDonough is a Senior Engineer in J.S. Held’s Accident Reconstruction Practice. Sean evaluates
vehicular accidents to determine the sequence of events leading up to the accident, using 3D scanning
and surveying to generate a 3D environment to accurately reconstruct the accident. He also employs
accepted principles of crush energy, impact momentum, and planar mechanics to determine speeds,
principle direction of force (PDOF), and with the use of PC Crash, he evaluates time and spatial
relationships of vehicular accidents. In addition to the wide range of crashes that he has analyzed, Mr.
McDonough brings a wealth of automotive maintenance expertise.

Sean can be reached at [email protected] or +1 303 733 1888.

Drew Donaldson is an Accident Reconstructionist in J.S. Held’s Accident Reconstruction Practice. Drew
is dedicated to investigation and vehicular accident reconstruction. In this field, he analyzes evidence,
conducts vehicle and scene inspections, and performs calculations using principles of physics.
Additionally, Drew has extensive experience in machine shop fabrication and automotive racing safety,
design, and fabrication techniques. Drew is involved in vehicular accident reconstruction and analysis,
mechanical systems design and failure analysis, technical research and writing of publications furthering
the capabilities of accident reconstructionists, and investigating new methods of physical evidence and
terrain data capture through aerial surveying.

Drew can be reached at [email protected] or +1 303 733 1888.

Reece Cochran is an Accident Reconstructionist in J.S. Held’s Accident Reconstruction Practice.

Reece can be reached at [email protected] or +1 303 733 1888.

References

[1] Fugger, T., Wobrock, J., Randles, B., Stein, A. et al., “Driver Characteristics at Signal-Controlled
Intersections,” SAE Technical paper 2001-01-0045, 2001, doi:10.4271/2001-01- 0045.

[2] Chang, M., Messer, C., and Santiago, A., “Timing Traffic Signal Change Intervals Based on Driver
Behavior,” Transportation Research Record 1027, 1985.

[3] Haas, R., Inman, V., Dixson, A., and Warren, D., “Use of Intelligent Transportation System Data to
Determine Driver Deceleration and Acceleration Behavior,” Transportation Research Record 1899, 2004,

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 31/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

doi:10.3141/1899-01.

[4] Wortman, R., and Matthias, J., “Evaluation of Driver Behavior at Signalized Intersections,”
Transportation Research Record 904, 1983.

[5] Carter, N., Beier, S., and Cordero, R., “Lateral and Tangential Accelerations of Left Turning Vehicles from
Naturalistic Observations,” SAE Technical Paper 2019-01-0421, 2019, doi:10.4271/2019-01-0421.

[6] Danaher, D., McDonough, S., and Donaldson, D., “Two Phase Heavy Truck Acceleration Model,” SAE
Technical Paper 2019- 01-0411, 2019, doi:10.4271/2019-01-0411.

[7] Danaher, D., Donaldson, A., and McDonough, S., “Acceleration of Left Turning Heavy Trucks,” SAE Int. J.
Adv. & Curr. Prac. in Mobility 2(4):2019-2036, 2020, doi:10.4271/2020-01-0882.

Appendix A

Appendix A contains tables of average longitudinal acceleration and deceleration rates for tracked
passenger vehicles, classified by intersection, and the row in which the vehicle stopped at the
intersection. The tables for all accelerating and decelerating vehicles list the mean value for each phase,
and a range of one standard deviation above and below the mean. Row 2 values for the individual
intersection tables also include the calculated percentage difference in acceleration/deceleration rate
from those calculated for first row vehicles.

Table 4 - All Accelerating Passenger Vehicles

Table 5 - Accelerating Passenger Vehicles – 56th and Quebec

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 32/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Table 6 - Accelerating Passenger Vehicles – 92nd and Federal

Table 7 - All Decelerating Passenger Vehicles

Table 8 - All Decelerating Passenger Vehicles – 56th and Quebec

Table 9 - All Decelerating Passenger Vehicles - 92nd and Federal

Find your expert. ℠

This publication is for educational and general information purposes only. It may contain errors and is provided as is.
It is not intended as specific advice, legal, or otherwise. Opinions and views are not necessarily those of J.S. Held or
its affiliates and it should not be presumed that J.S. Held subscribes to any particular method, interpretation, or
analysis merely because it appears in this publication. We disclaim any representation and/or warranty regarding the
accuracy, timeliness, quality, or applicability of any of the contents. You should not act, or fail to act, in reliance on
this publication and we disclaim all liability in respect to such actions or failure to act. We assume no responsibility
for information contained in this publication and disclaim all liability and damages in respect to such information.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 33/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

This publication is not a substitute for competent legal advice. The content herein may be updated or otherwise
modified without notice.

YOU MAY ALSO BE INTERESTED IN

WHITE PAPERS & RESEARCH REPORTS

Accuracy of Aerial Photoscanning with Real-Time Kinematic Technology

Photoscanning photogrammetry is a method for obtaining and preserving three-
dimensional site data from photographs. This photogrammetric method is commonly
associated with small Unmanned Aircraft Systems (sUAS) and is particularly beneficial
for large area site documentation...

WHITE PAPERS & RESEARCH REPORTS

Aerial Photoscanning with Ground Control Points from USGS LiDAR

This paper presents a methodology for utilizing publicly available LiDAR data from the
United States Geological Survey (USGS) in combination with high-resolution aerial
imagery to establish GCPs based on preexisting site landmarks. This method is...

WHITE PAPERS & RESEARCH REPORTS

A Comparison of Mobile LiDAR Capture and Established Ground-Based 3D

Scanning Methodologies
Ground-based Light Detection and Ranging (LiDAR) using FARO Focus 3D scanners
(and other brands of scanners) are repeatedly shown to accurately capture the
geometry of accident scenes, accident vehicles, and exemplar vehicles, as well as...

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 34/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

INDUSTRY INSIGHTS

Keep up with the latest research and

announcements from our team.

First Name Last Name

Email *

SUBMIT

BACK TO TOP

J.S. HELD

Headquarters
50 Jericho Quadrangle
Suite 117
Jericho, N.Y. 11753
+1 516 621 2900
[email protected]

FOLLOW US

LinkedIn

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 35/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

MEDIA RELATIONS

Contact:
[email protected]

FOLLOW US

LinkedIn

PARTNER FIRMS

J.S. Held Engineering Services PLLC

SUBSIDIARY FIRMS

Contract Administration Associates, a J.S. Held Company

DBSConsult, a J.S. Held Company
JS Held Malaysia Sdn. Bhd. | Registration No. 202001033278 (1389599-P)
J.S. Held Singapore Pte. Ltd. | Registration No. 202032700W
J.S. Held UK Limited
Ocean Tomo, a part of J.S. Held

AREAS OF EXPERTISE

Risk & Advisory

Financial
Technical & Scientific

INSIGHTS

White Papers & Research Reports

Perspectives
Educational Seminars & Events

MARKETS SERVED

Aerospace & Aviation

Agriculture
Cannabis
https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 36/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

Construction
Consumer Products & Services
Cyber
Education
Energy & Power
Financial Services
Food & Beverage
Government
Healthcare
Industrial & Manufacturing
Insurance
Private Equity
Real Estate
Retail, Restaurant & Hospitality
Technology, Media & Telecommunications
Transportation Infrastructure

ABOUT J.S. HELD

Culture & Careers

Our Experts
Our Offices
Leadership
News
Contact Us

MORE

Submit Assignment
Accessibility Statement
Complaints Handling Procedure
Do Not Sell My Personal Information
Online Privacy Notice
Service Organization Control (SOC) 2 Type II

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 37/38
5/17/23, 9:17 PM A Naturalistic Study of Vehicle Acceleration and… | J.S. Held

J.S. Held, its affiliates and subsidiaries are not certified public accounting firm(s) and do not provide audit, attest,
or any other public accounting services. J.S. Held, its affiliates and subsidiaries are not law firms and do not
provide legal advice. Securities offered through our affiliate, Ocean Tomo Investment Group, LLC, member
FINRA/SIPC. All rights reserved.

https://jsheld.com/insights/articles/a-naturalistic-study-of-vehicle-acceleration-and-deceleration-at-an-intersection 38/38