![Ford Autonomous Vehicle in an urban environment [21]](https://www.researchgate.net/profile/Hesham-Alghodhaifi/publication/341491372/figure/fig1/AS:892941379833857@1589905223701/Ford-Autonomous-Vehicle-in-an-urban-environment-21_Q320.jpg)
Content uploaded by Hesham Alghodhaifi
Author content
All content in this area was uploaded by Hesham Alghodhaifi on May 19, 2020
Content may be subject to copyright.
PROCEEDINGS OF SPIE
SPIEDigitalLibrary.org/conference-proceedings-of-spie
Simulation-based model for surrogate
safety measures analysis in
automated vehicle-pedestrian conflict
on an urban environment
Alghodhaifi, Hesham, Lakshmanan, Sridhar
Hesham Alghodhaifi, Sridhar Lakshmanan, "Simulation-based model for
surrogate safety measures analysis in automated vehicle-pedestrian conflict
on an urban environment," Proc. SPIE 11415, Autonomous Systems:
Sensors, Processing, and Security for Vehicles and Infrastructure 2020,
1141504 (18 May 2020); doi: 10.1117/12.2558830
Event: SPIE Defense + Commercial Sensing, 2020, Online Only, California,
United States
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Simulation-based Model for Surrogate Safety Measures
Analysis in Automated Vehicle-Pedestrian Conflict on an
Urban Environment
Hesham Alghodhaifi and Sridhar Lakshmanan
Dept. of Electrical & Computer Engineering, University of Michigan-Dearborn, 4901
Evergreen Rd, Dearborn, MI USA 48128
ABSTRACT
Conflict analysis using surrogate safety measures (SSMs) has turned into an efficient way to investigate safety
issues in autonomous vehicles. Previous studies largely focus on video images taken from high elevation. However,
it involves overwhelming work, high expense of maintenance, and even security limitations. This study applies
a simulation-based model for surrogate safety analysis of pedestrian-vehicle conflicts in urban roads. We show
how an automated vehicle system that utilizes a radar and a camera as an input to a Pedestrian Protection
System (PPS) can be used for surrogate safety analysis under uncertain weather conditions. Different scenarios
for surrogate safety measures were built and analyzed. The detection and tracking systems for vehicle and
pedestrian trajectory are modeled. Three SSMs, namely, Pedestrian Classification Time to Collision (PCT),
Total Braking Time to Collision (TBT), and Total Minimum Time to Collision (TMT) are employed to represent
how spatially and temporally close the pedestrian-vehicle conflict is to a collision. The simulation is built using
PreScan, and the software reproduces the test scenarios accurately as well as incorporates vehicle control and
logic. The results from our analysis highlight the exposure of pedestrians to traffic conflict both inside and outside
crosswalks. The findings demonstrate that simulation-based models can support urban roads safety analysis of
autonomous vehicles in an accurate and yet cost-effective way.
Keywords: Autonomous Vehicle, Conflict analysis, surrogate safety measures, simulation, pedestrian, weather
condition, Time-to-Collision, classification, PreScan
1. INTRODUCTION
Interaction in autonomous driving has a wide definition and can include tasks such as recognizing road users,
analyzing their behavior, interacting with them, predicting their future actions, and reacting with an appropriate
response [2]. Predictable social interaction between humans helps avoid many potential problems in urban traffic.
In contrast, studies have shown that the lack of social interaction in autonomous cars can cause collisions [3]
or have them behave unpredictably towards pedestrians [4]. Autonomous vehicles estimation of pedestrians’
behavior is limited because of the lack of social understanding. According to studies in the field of behavioral
psychology, many factors can possibly affect the behavior of road users [5]-[7]. These factors are pedestrians’
demographics [8], road conditions [7], social factors [6], traffic characteristics [9], and the interrelation between
these factors [2]. Intention estimation algorithms in intelligent driving have been established to predict future
actions of pedestrians [10] and drivers [11]. Recent developments in vehicle-to-everything (V2X) communications
have enabled more sharing of data and warning between autonomous vehicles and road users [12]-[16]. V2X
also helps collect more data to improve intention estimation and pedestrian safety [17]-[20]. In this paper, we
are focused on filling some gaps in the interaction between pedestrians and autonomous vehicles in an urban
environment (figure 1). Urban safety analysis has been relying on (old) collected accidents data. However, this
data has limitations: first, traffic collisions are infrequent, unpredictable and need extensive time to observe and
record; second, the process of identifying collisions within the data is time consuming; third, the quality of the
collected data is inadequate to make quantitative inference [1].
Further author information:
First Author: E-mail: halghodh@umich.edu, Telephone: +1 313-485-1902
Corresponding Author: E-mail: lakshman@umich.edu, Telephone: +1 313-593-5516
Autonomous Systems: Sensors, Processing, and Security for Vehicles and Infrastructure 2020,
edited by Michael C. Dudzik, Stephen M. Jameson, Proc. of SPIE Vol. 11415, 1141504
© 2020 SPIE · CCC code: 0277-786X/20/$21 · doi: 10.1117/12.2558830
Proc. of SPIE Vol. 11415 1141504-1
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Figure 1: Ford Autonomous Vehicle in an urban environment [21]
Therefore, surrogate safety measures (SSMs) have been introduced as a solution to tackle these challenges
and to produce more analysis without relying on accidents statistics alone. In addition, the hope is that these
safety analysis measures will also be used to create algorithms to improve autonomous driving. Many current
approaches to SSMs rely on traffic conflict analysis [22]-[25]. Traffic conflict is defined as "two or more road
users approach each other in space and time to such an extent that a collision is imminent if their movements
remain unchanged” [26]. Compared to collisions, conflicts are more likely [27]. Therefore, estimation of average
accident frequency using TCTs is more efficient compared to collisions based analysis [28]. The problem is, many
existing Traffic Conflict Techniques (TCTs) require a team of professional observers to investigate how critical
and frequent the traffic conflicts are at an intersection or other road section [29]. This manual process is inefficient
and prone to (subjective) error. Automated video analysis is found to be another good option to deal with the
limitations of manual traffic conflicts data gathering process, because traffic cameras are already placed on many
intersections and urban roads [30]. Automated video analysis method was applied extensively to vehicular traffic
analysis [31], [32]. The problem of automated detection and tracking of pedestrians is a complicated process
that is at the cutting-edge of computer vision research. The complexity in pedestrian detection and tracking
comes from stochastic dynamic movements, grouping, varied appearance, non-rigid nature, and the generally
less organized nature of pedestrian traffic as compared to vehicular traffic. The previous discussions make the
case for simulation-based traffic conflict analysis, the focus of this paper:
is used to both mimic the realistic and stochastic nature of the testing environment. Traffic conflict data
is then collected in a time-efficient manner by running the model.Specific traffic scenarios are used to
simulate conflicts between the pedestrian and autonomous vehicle. These conflicts could potentially lead
to accidents. SSMs are used for traffic conflict analysis.
Proc. of SPIE Vol. 11415 1141504-2
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
2. DATA COLLECTION FOR DEVELOPING SSMS-VPC MODEL
Six different test scenarios (figure 2) are used to model conflict between the pedestrian and autonomous vehicle.
These simulation-based models were developed within PreScan, Matlab and Simulink.
Figure 2: Test Scenarios
1.2.
The development of a simulation-based model for surrogate safety measures of an autonomous vehicle-
pedestrian conflict (SSMs-VPC) is based on testing environment data and SSMs-VPC performance data. The
testing environment data are vehicle speed, pedestrian direction and size, light and weather conditions, and the
forward distance. The SSMs-VPC performance data includes detection distance to target, pedestrian detection
distance to target, detection time to collision, pedestrian classification time to collision, braking time to collision,
and minimum time to collision.
3. DEVELOP THE SSMS-VPC MODEL
3.1 Structure of the SSMs-VPC Model
The Pedestrian Protection System (PPS) model from PreScan is used to establish the SSMs-VPC system. PPS
is a forward looking, predictive safety model that allows a safe navigation of autonomous vehicles by reducing the
danger of collisions with vulnerable road users. The PPS model uses a radar and a camera sensor for automated
recognition and collision avoidance of the vulnerable road users. Collision avoidance comes from total control of
the car’s brakes [33]. Our SSMs-VPC model comprises of the following sub-models:
•Radar Perception Model - Based on the radar characteristics, this model can detect an object and calculate
the time to collision (TTC). If the TTC drops below Estimated Min Safe TTC, this object will be considered
to be dangerous.
•Pedestrian Classification Model - If a dangerous object is detected by the Radar Detection Model, this
model continues to check whether it is a pedestrian, and provides a warning if a pedestrian is classified.
•Decision Making Model - If an imminent crash to a pedestrian is identified by the Pedestrian Identification
Model, this model decides the warning and braking behavior of the vehicle.
Proc. of SPIE Vol. 11415 1141504-3
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
3.2 SSMs-VPC sub-Model
3.2.1 Radar Perception Model
Radar Perception Model mimics the radar behavior for detecting possible severe conflicts to objects. Radar
systems provide essential sensor input for safe and well-founded autonomous driving system performance. Au-
tonomous radar employs millimeter wave frequencies for long-range or short-range object and obstacle detection,
as well as for tracking the velocity and direction of the various road users such as pedestrians, other vehicles, etc.,
in the surrounding environment of the autonomous vehicle [34]. Moreover, the radar sensor can determine the
distance to or from any road users or object around the autonomous vehicles. In our case, when the autonomous
vehicle and the pedestrian are moving in a straight direction, the absolute velocity vector allows us to determine
the point where the severe conflict will happen. Then, time to collision (TTC) is calculated and compared with
set values [37].
Our Radar Perception Model (figure 3) was created using PreScan’s PPS model with some modifications such
as adding more safety measures parameters. Radar range, Doppler velocity, and Azimuth angle are inputted
to the Radar Perception Model. Moreover, the host vehicle’s actual velocity and actual yaw rate are used as
additional inputs to the Radar Perception Model. Time-to-Collision and Radar Detection Flag are the two
outputs that we use. Based on the TTC values, signal flags that are used as system triggers are set. An obstacle
is considered to be in a severe conflict if the predicted time to collision with the host (autonomous) vehicle is less
than 2 seconds (T T C < 2s) [37]. Based on the TTC values further decisions will be made to avoid a (deadly)
collision. Radar Detection Flag is used to show if a dangerous object is detected by the radar system. The Radar
Detection Flag is used to trigger the pedestrian classification model.
Figure 3: Radar Perception Model
3.2.2 Pedestrian Classification Model
After receiving information from the radar that an object is detected and predicted to cause a potential accident,
the pedestrian classification algorithm using the camera data is involved to decide if this object is a pedestrian
or not. The radar data and camera data are brought together to form a single image of the environment around
Proc. of SPIE Vol. 11415 1141504-4
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
a vehicle. Thus, it is challenging for the scientists to determine what specific image processing and sensor fusion
algorithms to use in the SSMs-VPC system. Rather than focus on a specific image processing and sensor fusion
algorithm for pedestrian classification, our work here focuses on studying the behavior of the SSMs-VPC system.
Moreover, this study aims to determine the conditions that determine whether a pedestrian could be detected
by the SSMs-VPC system and how much time the SSMs-VPC would take to classify the pedestrian. Generally,
using the pedestrian classification model (figure 4) with different image processing and sensor fusion algorithms
in diverse environments produces significant various outputs. That means the accuracy and the efficiency for
each algorithm will be different. As a result, we modeled the environment behavior through the environmental
factors to capture the performance of real image processing algorithms. The input parameters of the Pedestrian
Classification Model are AV speed, pedestrian speed, size of the pedestrian, cloth color of the pedestrian, and
weather condition. The Pedestrian Detection Flag is the output of the Pedestrian Classification Model and
is used to indicate if a pedestrian can be classified by the SSMs-VPC and when it would be identified. The
performance of SSMs-VPC models can be different under various test scenarios. From this model, the total
Pedestrian Classification Time to Collision is calculated as a SSM.
Figure 4: Pedestrian Classification Model
3.2.3 Decision Making Model
The Decision Making Model (figure 5) is the final sub-model in the SSMs-VPC model. The aim of the Decision
Making Model is to determine the behavior of the autonomous vehicle. We use the PreScan PPS Actuation
Model with some modifications. The Warning Flag within the Actuation Model is set if a pedestrian has been
classified and TTC value drops below 1.6 s. The Braking Flag is then set if the TTC value drops below 0.6
s. Moreover, the braking pressure is set in advance to a specific value. The Warning TTC, Braking TTC, and
Braking Pressure differ significantly in various test scenarios. AV speed, TTC, Radar Perception Model output,
and Pedestrian Classification Model output are the input parameters of the Decision Making Model. The warning
flag, braking flag, and the braking pressure are the three outputs of this model. If the Warning Flag is set, the
action model will turn on the warning lights and beeps. If the warning flag is turned on, the Braking Flag is
set and a specified braking pressure will be applied to the vehicle. The output actions will be displayed by the
displayed model shown in Figure 5. The Decision Making Model continuously receives and examines the input
Proc. of SPIE Vol. 11415 1141504-5
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
parameters, Radar Detection Flag and Pedestrian Detection Flag. The Warning Flag is triggered right away
when the other two flags are raised.
Figure 5: Decision Making Model
4. SIMULATION RESULTS
AUDI A8 Sedan with SSMs-VPC capability were used in PreScan and MATLAB/Simulink to test the SSMs-VPC
model. The Pedestrian Protection System uses two sensors (Table 1 and 2).
Table 1: Mid Range Radar (MRR)
Range 60 m
Beam width 60 deg
Table 2: Camera Sensor
Monocular Monochrome
CCD Chip Size 1/2” (6.4 mm ×4.8mm)
Proc. of SPIE Vol. 11415 1141504-6
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
The following table describes the system model specifications.
Table 3: System Model
Host Vehicle Audi A8
Sensing Radar, Camera Sensor
Controller Time-to-Collision (TTC), Pedestrian Detection
Actuation Brakes
Thirty nine scenarios were conducted and 100 test runs for each scenario. Pedestrian Classification Time to
Collision is calculated. The Pedestrian Classification Time to Collision is defined as the time required to classify
a pedestrian by the Pedestrian Classification Model. Six basic directions for pedestrians with different pedestrian
types are defined and discussed. Table 4 describes the relationship between the Pedestrian Classification Time
to Collision and the pedestrian direction and type.
Table 4: Pedestrian Classification Time vs Pedestrian Direction and Type
Pedestrian Direction - Type Pedestrian Classification Time (PCT) (s)
Farside - Adult 0.0422
Nearside - Bicyclist 0.0771
Longitudinal- Adult 0.12
Nearside - Adult Obstructed 0.158
Nearside - Adult 0.4407
Longitudinal - Bicyclist 1.3688
The relationship between the Pedestrian Classification Time to Collision and the contrast of the pedestrian to
the background is studied. The type of contrast is identified based on the light condition, Weather condition, and
the cloth color of the pedestrian. The results show that high contrast produces a small Pedestrian Classification
Time to Collision. This helps to prevent any severe conflicts between the autonomous vehicles and vulnerable
road users.
Finally, the relationship between Pedestrian Classification Time to Collision (PCT), Total Braking Time to
Collision (TBT), and Total Minimum Time to Collision with the forward distance is studied. Figures 6, 8 and 10
show how these safety margins are changing while changing the forward distance. Higher vehicle speed, higher
pedestrian speed, higher forward distance, smaller pedestrian size, or poor visibility due to lighting and weather
conditions are responsible for producing longer Pedestrian Classification Time to Collision. Longer Pedestrian
Classification Time to Collision will result in a shorter time period for SSMs-VPC warning and braking which
will cause a severe conflict.
Figures 6, 8, and 10 show the modeling result of PCT, TBT, TMT based on the simulation data of AUDI
A8 sedan. The x axis is the forward distance in meters. The y axis is the PCT in seconds for figure 6, TBT for
figure 8, and TMT for figure 10. 39 test scenarios and the average of 100 runs for each test scenario are shown
in these figures. Using the polynomial fit with one degree of freedom, a straight line representing PCT, TBT,
and TMT are plotted. A random offset is added to represent a model close enough to the real world scenario.
The offset range is from -Dfto Df. From figure 6, the PCT can be calculated as shown below.
P CTDf= 0.0570 ∗Df−1.2625 + of fset (1)
Proc. of SPIE Vol. 11415 1141504-7
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
A goodness of fit test is conducted to summarize the discrepancy between the PCT values and the PCT values
expected under the PCT model (figure 7). The covariance matrix M, p-value, R squared, Adjusted R squared,
and RMSE are calculated as shown below.
MP C T =−171.0809 −6.2120
0−0.6414(2)
p−value =0.0006 ×10−11 0.2761 ×10−11 (3)
Table 5: PCT Goodness of fit test
R squared 0.8111
Adjusted R squared 0.8060
RMSE 0.0774
The p-value here is ≤0.05, which indicates that the resulting data are unlikely with a true null hypothesis. The
null hypothesis here indicates the lack of a difference between the PCT values and the PCT expected values
under this model. The R squared value is almost 81.11%. This value indicates that the model explains almost
all the variability of the response data around its mean. The adjusted R squared value is 80.06% which shows
the best approximation of the degree of relationship in the basic population. The RMSE value is 0.0774 which
indicates a better fit.
Figure 6: Forward Distance vs Pedestrian Classification Time to Collision
Proc. of SPIE Vol. 11415 1141504-8
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Figure 7: Linear Fit of PCT Data with 95% Prediction Interval
Figure 8: Forward Distance vs Total Braking Time to Collision
Proc. of SPIE Vol. 11415 1141504-9
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Figure 9: Linear Fit of TBT Data with 95% Prediction Interval
The modeling result of TBT is shown in figure 8 and the TBT can be calculated as shown below.
T BTDf= 0.2846 −0.0077 ∗Df+of f set (4)
A goodness of fit test is conducted to show the difference between the TBT values and the TBT expected values
under the TBT model (figure 9). The covariance matrix M, p-value, R squared, Adjusted R squared, and RMSE
are calculated as shown below.
MT B T =−171.0809 −6.2120
0−0.6414(5)
p−value =0.8492 ×10−80.0005 ×10−8(6)
Table 6: TBT Goodness of fit test
R squared 0.5964
Adjusted R squared 0.5855
RMSE 0.0178
Figure 10 describes the relationship between forward distance and Total Minimum Time to Collision (TMT).
The TMT can be calculated from figure 10 as shown below.
T M TDf= 0.0079 ∗Df+ 0.2922 + offset (7)
Proc. of SPIE Vol. 11415 1141504-10
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Figure 10: Forward Distance vs Total Minimum Time to Collision
Figure 11 shows the linear fit of the Total Minimum Time to Collision with 95% prediction interval. To show
the goodness of fit, the covariance matrix M, p-value, R squared, Adjusted R squared, and RMSE are calculated
as shown below.
MT M T =−171.0809 −6.2120
0−0.6414(8)
p−value =0.4256 ×10−80.0002 ×10−8(9)
Table 7: TMT Goodness of fit test
R squared 0.6110
Adjusted R squared 0.6005
RMSE 0.0177
According to the R squared, and adjusted R squared values, around 61% of the data fit the regression model.
The RMSE indicates how close the TMT observed data are to the TMT model’s predicted values. A small value
of the RSME shows a better fit. Moreover, p-value expresses that the TMT data are unlikely with a true null
hypothesis.
Proc. of SPIE Vol. 11415 1141504-11
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Figure 11: Linear Fit of TMT Data with 95% Prediction Interval
5. CONCLUSION AND FUTURE WORK
This work described a structured approach to develop a simulation-based model for Surrogate Safety Measures
Analysis in Automated Vehicle-Pedestrian Conflict on an Urban Environment using vehicle simulation data.
The results showed that Pedestrian Classification Time to Collision (PCT), Total Braking Time to Collision,
and Total Minimum Time to Collision can be used as safety margins for autonomous vehicle-pedestrian conflicts.
Our future work will involve testing this model in more real world scenarios, and will require model improvements
and additional data.
References
[1] “Pedestrian safety: a road safety manual for decision-makers and practitioners.,” World Health Or-
ganization, <https://www.who.int/publications-detail/pedestrian-safety-a-road-safety-manual-for-decision-
makers-and-practitioners> (27 April 2020 ).
[2] Rasouli, Amir, and John K. Tsotsos. “Autonomous Vehicles That Interact With Pedestrians: A Survey of
Theory and Practice.” IEEE Transactions on Intelligent Transportation Systems, vol. 21, no. 3, 2020, pp.
900–918., doi:10.1109/tits.2019.2901817.
[3] Anthony, Samuel English. “A Google Car’s Crash Shows the Real Limitation of Driverless Vehicles.”
Slate Magazine, Slate, 1 Mar. 2016, slate.com/technology/2016/03/google-self-driving-cars-lack-a-humans-
intuition-for-what-other-drivers-will-do.html.
[4] Richtel, Matt, and Conor Dougherty. “Google’s Driverless Cars Run Into Problem:
Cars With Drivers.” The New York Times, The New York Times, 1 Sept. 2015,
www.nytimes.com/2015/09/02/technology/personaltech/google-says-its-not-the-driverless-cars-fault-
its-other-drivers.html.
[5] M. M. Ishaque and R. B. Noland, “Behavioural issues in pedestrian speed choice and street crossing be-
haviour: A review,” Transp. Rev., vol. 28, no. 1, pp. 61–85, 2008.
Proc. of SPIE Vol. 11415 1141504-12
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
[6] A. Willis, N. Gjersoe, C. Havard, J. Kerridge, and R. Kukla, “Human movement behaviour in urban spaces:
Implications for the design and modelling of effective pedestrian environments,” Environ. Planning B, Plan-
ning Des., vol. 31, no. 6, pp. 805–828, 2004.
[7] W. A. Harrell, “Factors influencing pedestrian cautiousness in crossing streets,” J. Social Psychol., vol. 131,
no. 3, pp. 367–372, 1991.
[8] J. Cohen, E. J. Dearnaley, and C. E. M. Hansel, “The risk taken in crossing a road,” J. Oper. Res. Soc., vol.
6, no. 3, pp. 120–128, 1955.
[9] G. Jacobs and D. G. Wilson, “A study of pedestrian risk in crossing busy roads in four towns,” Road Res.
Lab., London, U.K.: Tech. Rep. LR106, 1967.
[10] A. Rasouli, I. Kotseruba, and J. K. Tsotsos, “Are they going to cross? A benchmark dataset and baseline
for pedestrian crosswalk behavior,” in Proc. ICCV, 2017, pp. 206–213.
[11] P. Molchanov, S. Gupta, K. Kim, and K. Pulli, “Multi-sensor system for driver’s hand-gesture recognition,”
in Proc. FG, vol. 1, May 2015, pp. 1–8.
[12] Eric Cheek, Hesham Alghodhaifi, Cristian Adam, Russell Andres, and Sridhar Lakshmanan "Dedicated
short range communications used as fail-safe in autonomous navigation", Proc. SPIE 11425, Unmanned
Systems Technology XXII, 114250P (23 April 2020); https://doi.org/10.1117/12.2558925
[13] Sridhar Lakshmanan, Yuedong Yan, Stan Baek, and Hesham Alghodhaifi ”Modeling and simulation of leader-
follower autonomous vehicles: environment effects”, Proc. SPIE 11021, Unmanned Systems Technology XXI,
110210J (15 May 2019); https://doi.org/10.1117/12.2520485
[14] H. Alghodhaifi, S. Lakshmanan, S. Baek, and P. Richardson, Autonomy Modeling and Validation in a Highly
Uncertain Environment, Proceedings of the 2018 Ground Vehicle Systems Engineering and Technology
Symposium (GVSETS).
[15] L. Hobert, A. Festag, I. Llatser, L. Altomare, F. Visintainer, and A. Kovacs, “Enhancements of V2X com-
munication in support of cooperative autonomous driving,” IEEE Commun. Mag., vol. 53, no. 12, pp. 64–70,
Dec. 2015.
[16] A. Hussein, F. García, J. M. Armingol, and C. Olaverri-Monreal, “P2V and V2P communication for pedes-
trian warning on the basis of autonomous vehicles,” in Proc. ITSC, Nov. 2016, pp. 2034–2039.
[17] A. Tahmasbi-Sarvestani, H. Nourkhiz Mahjoub, Y. P. Fallah, E. Moradi-Pari and O. Abuchaar, "Imple-
mentation and Evaluation of a Cooperative Vehicle-to-Pedestrian Safety Application," in IEEE Intelligent
Transportation Systems Magazine, vol. 9, no. 4, pp. 62-75, winter 2017.
[18] J. J. Anaya, P. Merdrignac, O. Shagdar, F. Nashashibi and J. E. Naranjo, "Vehicle to pedestrian commu-
nications for protection of vulnerable road users," 2014 IEEE Intelligent Vehicles Symposium Proceedings,
Dearborn, MI, 2014, pp. 1037-1042.
[19] M. Bagheri, M. Siekkinen and J. K. Nurminen, "Cellular-based vehicle to pedestrian (V2P) adaptive commu-
nication for collision avoidance," 2014 International Conference on Connected Vehicles and Expo (ICCVE),
Vienna, 2014, pp. 450-456.
[20] Liu, Z., Liu, Z., Meng, Z., Yang, X., Pu, L., Zhang, L. (2016). Implementation and performance measure-
ment of a V2X communication system for vehicle and pedestrian safety. International Journal of Distributed
Sensor Networks. https://doi.org/10.1177/1550147716671267
[21] Moscaritolo, Angela. “Riding Around Miami in Ford’s Self-Driving Test Cars.” PCMAG, PCMag, 15 Nov.
2018, www.pcmag.com/news/riding-around-miami-in-fords-self-driving-test-cars.
[22] Traffic conflict standards for intersections. Sayed, T. and Zein, S. s.l. : Transportation Planning and Tech-
nology, Vol. 22, pp. 309–323.
Proc. of SPIE Vol. 11415 1141504-13
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
[23] Simulation of Traffic Conflicts at Unsignalised Intersections with TSC-Sim. Sayed, T., Brown, G. R. and
Navin, F. 5, s.l. : Accident Analysis Prevention, 1994, Vol. 26, pp. 593-607.
[24] Traffic Conflict Techniques and Other Forms of Behavioural Analysis: Application to Safety Diagnoses.
Muhlrad, Nicole. 1993.
[25] Perkins, S. and Harris, J. Criteria for Traffic Conflict Characteristics. s.l. : General Motors Corporation,
1967. Report GMR 632.
[26] Amundson, F. and and Hydén, C. Oslo : Proceedings of First Workshop on traffic Conflicts, Institute of
Economics, 1977.
[27] Development of Conflicts Analysis Technique for Pedestrian Crossings. Cynecki, M. J. 1980, Transportation
Research Record, Vol. 743.
[28] Migletz, D.J., Glauz, W.D. and Bauer, K.M. Relationships between traffic conflicts and accidents. s.l. : US
Department of Transportation, Federal Highway Administration, 1985. FHWA/RD-84/042.
[29] Fitzpatrick, Kay, et al. TCRP Report 112/NCHRP Report 562. Improving Pedestrian Safety at Unsignalized
Crossings. Transportation Resarch Board. 2006.
[30] Estimating the severity of safety related behaviour. Svensson, Å. and Hydén, C. 2006, Accident Analysis
and Prevention, Vol. 38, pp. 379-385.
[31] Pedestrian Walking Speed, Step Size, and Step Frequency from the Perspective of Gender and Age: Case
Study in Beijing, China. Hui, Xiong, et al. 2007. Source Data: Transportation Research Board Annual
Meeting 2007 Paper No. 07-1486.
[32] Messelodi, S. and Modena, C. M. A computer vision system for traffic accident risk measurement: A case
study. s.l. : Technical Report, ITC-irst , ITC, 2005. T05-06-07.
[33] Simcenter Prescan User Manual (Version 8.6.0), Siemens PLM Software, Plano, TX, USA, 2018. Accessed
on: April. 21, 2020.
[34] Carpenter, Shawn. “Autonomous Vehicle Radar: Improving Radar Performance with Simulation.” Engi-
neering Simulation amp; 3D Design Software, www.ansys.com/about-ansys/advantage-magazine/volume-
xii-issue-1-2018/autonomous-vehicle-radar.
Proc. of SPIE Vol. 11415 1141504-14
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 19 May 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
