Content uploaded by Yunjie Zhao
Author content
All content in this area was uploaded by Yunjie Zhao on Nov 12, 2015
Content may be subject to copyright.
Page 1 of 26
Integrated Traffic-Driving-Networking Simulator for the Design of
Connected Vehicle Applications: Eco-Signal Case Study
Yunjie Zhao1, Aditya Wagh2, Yunfei Hou2,
Kevin Hulme3, Chunming Qiao2, Adel W. Sadek1
1Department of Civil, Structural and Environmental Engineering,
2Department of Computer Science and Engineering,
3 NYSCEDII,
University at Buffalo, the State University of New York, Buffalo, New York, USA
Address correspondence to Adel W. Sadek, Department of Civil, Structural and Environmental Engineering,
233 Ketter Hall, Buffalo, NY 14260, USA. Phone: (716) 645-4367. E-mail: asadek@buffalo.edu
ABSTRACT
This paper describes an on-going trans-disciplinary research initiative aimed at developing an
integrated traffic-driving-networking simulator (ITDNS) for the design and evaluation of
Cyber Transportation Systems (CTS) and Connected Vehicle (CV) applications. The ITDNS
allows a human driver to control a subject vehicle in a virtual environment which is capable
of communicating with other vehicles and the infrastructure with CTS messages. The
challenges associated with the integration of the three simulators, and how those challenges
were overcome, are discussed. As an application example, an eco-signal system, which
recommends the approach speed for vehicles approaching the intersection so as to minimize
fuel consumption and emissions, was implemented in the ITDNS. Test drivers were then
asked to drive through the intersection twice, one time with the eco-signal system in place
and another without the system. By doing so, the research was able to evaluate the likely
benefits of the eco-signal system in a fashion that considers the likely response of human
drivers to the recommended speed profiles of the eco-signal. In terms of the ITDNS, the
study demonstrates the unique advantages of the simulator in the evaluation and design of
CTS applications. Regarding the specific eco-signal case study considered, the preliminary
results demonstrate the potential of the concept to result in tangible reductions in fuel
consumption and emissions at signalized intersections of around 9% for energy consumption,
18% for carbon monoxide, and 25% for nitrogen oxides emissions. Besides its
environmental benefits, the eco-signal application eliminated hard accelerations and
decelerations maneuvers, and thus appears to have the potential to improve safety.
Keywords: Integrated Simulator; Traffic simulation; Driving Simulators; Networking
simulators; Connected Vehicles.
Page 2 of 26
LIST OF FIGURES
Figure 1. Motion Simulator...................................................................................................... 18
Figure 2. Architecture of the ITDNS ....................................................................................... 19
Figure 3. Eco-signal Display Panel .......................................................................................... 20
Figure 4. Eco-driving Distance-Time Diagram ....................................................................... 21
Figure 5. Paramics Arterial Test Bed, Connecting UB South and North Campus .................. 22
Figure 6. Energy, NOx, CO w/ and w/o Eco-Driving ............................................................. 23
Figure 7. Speed-Time Diagrams w/ and w/o Eco-Driving ...................................................... 24
Figure 8. Accel-Time Diagrams w/ and w/o Eco-Driving……………...…………… ……. .25
LIST OF TABLES
Table 1. Average Energy, NOx, CO, Time, Max Accelerations and Decelerations and Speeds
.................................................................................................................................................. 26
Page 3 of 26
I. Introduction
The paper describes a trans-disciplinary research initiative aimed at developing an integrated
traffic-driving-networking simulator (ITDNS) for the design and evaluation of Cyber
Transportation Systems (CTS) and Connected Vehicle (CV) applications. The ITDNS is
architected to allow a human driver to control a subject vehicle in a virtual environment
which is capable of communicating with other vehicles and infrastructure with CTS messages
as well as sending warning messages to the driver. ITDNS combines the main features of a
traffic simulator (TS), a networking simulator (NS) and a driving simulator (DS), and
therefore may be referred to as an integrated 3-in-1 simulator. The key advantage of the
ITDNS compared to previous efforts on the topic is its ability to take into account human
responses to proposed applications in a controlled environment. Given that the primary
motivation behind developing the ITDNS is to provide a unique facility to support the
evaluation of novel CTS and CV applications, following the description of the ITDNS and
the integration challenges, the paper discusses how the ITDNS is used to evaluate the likely
environmental benefits of eco-signals. Eco-signals are systems designed to recommend a
speed profile to vehicles approaching an intersection, which allows them to proceed through
the intersection with minimal energy consumption or emissions (e.g., Wu et al., 2010).
Historically, the transportation community has used several distinct simulators, but with no
true integration. On one hand, traffic simulation (TS) models (e.g., PARAMICS by
Quadstone , 2010) were used to evaluate the operational efficiency of transportation
networks. Driving Simulator (DS), on the other hand, were used to examine the behavior of
individual human subjects within a virtual environment. Finally, in recent years and with the
interest in CV applications, transportation researchers have also begun to utilize
communications network simulations (NS) such as NS-2 (NS-2 Simulator Website). Each
simulator type, when used individually, has its own set of strengths and limitations as
described next.
TS models capture the dynamics of large-scale traffic networks, but they lack driver
behavioral realism because vehicle movements are idealistic. For example, drivers in a
traffic simulator do not run a red light or do not get too close to the vehicle in front so as to
pose an accident risk. This, as a result, precludes their use for traffic safety analyses which
require models that reflect driver error. On the other hand, DSs are quite effective in studying
driver behavior in a controlled environment, but they lack network authenticity since
background traffic is often pre-programmed and does not react to the real-time actions of the
human subject. Both TS and DS naturally lack the ability to model the performance of
communication systems, which is captured by NS. With respect to the communications
network simulators, although they are capable of simulating wireless channels and exchange
CTS messages among connected nodes (i.e., the running applications on the vehicles), they
are incapable of simulating the realistic motion of the vehicular traffic itself, a task which is
best left to a dedicated TS.
To address the limitations of the stand-alone simulators and to leverage the advantages
unique to each type, this paper develops an integrated simulator which consists of: (1)
PARAMICS, which serves as the traffic simulator; (2) NS-2, to serve as the communications
network simulator; and (3) the University at Buffalo’s (UB) driving simulator. While there
have been numerous studies that have attempted to develop integrated TS/NS simulators
Page 4 of 26
(e.g., Vuyyuru and Oguchi 2007) and a few others that attempted to integrate TS and DS(e.g.
Ikeuchi et al. 2005), to the best of our knowledge, none has attempted to integrate all three
types of simulators in the fashion this study has accomplished.
The ITDNS, by virtue of its integration of the three types of simulators, provides a unique
testing and evaluation facility for CTS applications, such as the eco-signal application
considered in this paper. Specifically, ITDNS’s unique advantage steps from its human-in-
the-loop design which allows for observing and understanding the drivers’ perceptions and
responses to a given CTS application, and the implications of the observed response with
respect to improving (or worsening) the transportation system safety or sustainability. For
example, observing human subjects’ response in the ITDNS can provide answers to questions
such as, (1) what kind of alert or warning messages is more effective and easy to perceive by
drivers? (2) To what level are drivers willing to comply with the alerts or advisory messages
(e.g. slow down when receiving an incident alert without physically seeing the incident scene
for themselves)? (3) Are eco-driving tips and feedback messages truly effective in changing
driver patterns to make them more eco-friendly? Addressing these very important human
factors issues appear to be missing in many of the previous simulation-type studies.
There is a second important reason which underscores the importance of understanding and
correctly accounting for human factors issues when designing and evaluating CTS
applications, which is the evolutionary nature of the likely deployment path of CTS.
Transitioning from a driving environment in which the human driver is the primary
responsible agent for controlling the vehicle, as is the case today, to an environment in which
vehicles are fully autonomous cannot happen immediately. Instead, the transition would have
to take place in phases, in which incremental levels of human control are relinquished to
automation over time. This means that in the short- and medium- term, CTS applications are
still likely to involve a human-in-the-loop, since initially CTS systems would be designed to
basically provide human drivers with advisory and alert messages, leaving the final driving
control decision to the human driver.
Consider for example the case of an eco-signal system which is designed to reduce fuel
consumption and/or emissions for approaching vehicles. In its initial deployment phase, the
system may just recommend an approach speed to the driver, who would be responsible for
controlling the vehicle’s speed based upon the system’s recommendation, if he/she desires.
As a next stage in the evolution of the system, the vehicle’s speed may be automatically
controlled similar to what takes place in an adaptive cruise control system, and drivers may
still maintain steering control and override capability. Finally, in a number of years, there
may be a transition to fully automated vehicles.
Moreover, in terms of market penetration of CTS applications, in the early stages, only a
small portion of vehicles will be equipped for CTS applications via wireless communications
and on-board driving assistance devices. It thus becomes essential to study the interactions
between equipped and non-equipped vehicles at both the microscopic and macroscopic
levels. Once again, the ITDNS may be ideal for studying such situations (e.g., how human
drivers in non-equipped vehicles interact with equipped vehicles).
The third unique advantage of including the human element in the ITDNS is the ability it
provides for modeling the salient features of CTS. To the authors’ knowledge, there are
currently no commercial simulation packages that explicitly model the impact of CTS
Page 5 of 26
messages, communicated via vehicle-to-vehicle (V2V) and Vehicle-to-Infrastructure (V2I)
communications for example, on driver behavior. With the ITDNS, the response of human
subjects to various CTS warning or advisory messages may be observed, analyzed and finally
used to build new driver behavior models that explicitly accounts for the impact of CTS
messages on driving behavior. These models could then be integrated within the ITDNS
framework.
The rest of the paper is organized as follows. Section 2 first presents background information
on traffic, driving and networking simulators, and then briefly surveys the few previous
studies aimed at developing integrated simulators. The design of the ITDNS is described in
section 3, and the challenges encountered during the integration effort are discussed in the
section 4. The eco-signal application is then introduced in section 5, and used as an example
to show how the ITDNS may be used to realistically evaluate CTS applications. The paper
concludes by summarizing the study’s main conclusions and providing suggestions for future
research.
II. Background
II.1. Traffic Simulator
Recently, there has been a growing interest among the transportation community in the use of
microscopic traffic simulation modeling for evaluating the system performance of
transportation networks, as for example endorsed by the Highway Capacity Manual 2000
(Transportation Research Board, 2000). Microscopic simulation models simulate the
movement of individual driver-vehicle units (DVUs) based on car-following and lane-
changing theories. Interactions among those DVUs then help define the overall transportation
system performance, in terms of measures such as average speed, delay, queue lengths, etc.
Over the years, a number of car following models have been proposed and evaluated,
typically using test vehicles on test tracks (Gartner et. al., 2005). Examples of state-of-the-art
microscopic simulation models include PARAMICS (Quadstone, 2010), VISSIM (PTV-AG,
2004), and AIMSUN (Barcelo and Ferrer, 1998).
While the use of microscopic traffic simulation has historically focused on the analysis of the
transportation system efficiency and operations, recently there has been an increased interest
in its use in traffic safety and traffic conflict analysis (Cunto and Saccomanno, 2006). One
challenge, however, with this is the complex and multi-disciplinary nature of road-user
behavior. As pointed out by Archer (2000), existing microscopic simulation model based on
available car following, gap-acceptance, and lane change models may lack the level of detail
required for safety evaluations, which demand models that reflect errors in drivers’
perception, decision-making and action.
II.2. Driving Simulator
In the specific case of road vehicles, a typical driving simulator will consist of driver controls
that dictate the driven course of the human operated vehicle. These controls could range
from a simple keyboard/mouse on a desktop PC, to a more sophisticated and authentic game-
based steering wheel and pedals. The display environment for such a simulator might be in
the form of a single desktop monitor, or multiple “life size” display screens that surround and
“immerse” the human driver. Many realistic driving simulators also provide audio cues
Page 6 of 26
resulting from sounds that originate both interior and exterior to the vehicle cabin. In terms
of the simulator motion platform, a basic driving simulator could be of the fixed-base variety
(i.e., it is “stationary”). More advanced simulators, however, will include a moving base
within which the human driver is seated. Such a motion platform, parallel manipulator, or
Stewart Platform (Stewart, 1965), is a powered, mechanical, self-contained system for the
execution of motion-based simulation and is commonly described by referring to the number
of degrees-of-freedom (DOF) that can be simulated by the hardware. These systems are
commonly used for motion simulation and training activities in various applications that may
go beyond road driving simulation, including flight simulation (e.g., Baret, 1978), spacecraft
training (e.g., Claudinon and Lievre, 1985), and amusement rides and location-based
entertainment (e.g., SMS, 2003).
Advanced driving simulators are present at a moderate number of universities across North
America. Examples include the National Advanced Driving Simulator (NADS) at the
University of Iowa, and the driving simulators at the University of Michigan’s Transportation
Research Institute (UMTRI), Virginia Tech’s Transportation Institute (VTTI), the University
of Massachusetts, Penn State, and the University at Buffalo (which is utilized in this study).
Major commercial driving simulators that include system hardware and software include:
DriveSafety, STISIM, and VirtualDriver ( DriverSafety, STISIM and VirtualDriver
Websites). One major limitation with a majority of existing driving simulators is that they
lack network realism. Background traffic is often non-intelligent and pre-programmed, and
therefore does not respond to the actions (or reactions) of the human driver, in real-time.
This primary deficiency limits the application of these driving simulators to a subset of
scenarios that are limited to single site or local situations (e.g., a particular intersection)
without substantial neighboring traffic vehicles, such as would be required by transportation
system-level evaluations.
II.3. Integrated Simulator
The concept of linking two simulators: driving and traffic simulation, or traffic and network
simulator – has been attracting increased attention. The next sections will briefly review
attempts at integrating traffic and driving simulators on one hand, and traffic and
communications network simulators on the other.
II.3.1. Traffic-Driving Simulator
In an early research attempt, Pursula (1999) described the concept of integrating traffic and
driving simulators and its advantages, but did not elaborate on any technical details. Ikeuchi
et al. (2005) described an ambitious research program in Japan aimed at constructing a
“Mixed Reality Traffic Experiment Space”, which indeed involved linking driving simulators
with traffic simulation models. However, the paper did not offer much in the way of specific
technical details, specifically those relating to implementation. Jin and Lam (2003) proposed
integrating VISSIM with a driving simulator to study the impacts of Intelligent
Transportation Systems (ITS). Jenkins (2005) described a case study of integrating the
PARAMICS model with a driving simulator, and discussed the shortcomings of the approach.
Finally, a paper by Ciuffo et al. (2009) discussed the challenges of integrating driving and
traffic simulation, and presents some first solutions to some of the problems. Despite various
efforts over the last decade, however, there are still several research questions that need to be
addressed.
Page 7 of 26
II.3.2. Traffic-Network Simulator
A number of research groups have explored the idea of integrating traffic and
communications network simulators in order to better model the Vehicular ad-hoc network
(VANET) environment. De Marco et al. (2007) proposed CAVENET, a toolbox for
combining traffic and networking simulators. However, the two simulators run sequentially
with the mobility traces of the traffic simulation being used to control the nodes in the
networking simulation. Since there is no feedback in the opposite direction (i.e. from network
simulation to traffic simulation), this tool cannot be used to test applications such as those for
safety improvement that can affect the mobility of nodes.
Barolli et al. (2010) addressed the vehicular mobility model and tested the performance of
AODV, OLSR and DYMO but did not add a feedback loop between the simulators.
Mangharam et al. (2006) proposed GrooveNet, which included an interaction between the
network and traffic simulators. However, the traffic model did not include realistic lane
changing and car following behaviors. NCTUns (Wang et al. 2010) is another tool that
closely combines traffic and network simulation and provides a good simulation of the
networking stack by using kernel re-entering. However, the tight coupling of mobility
simulation and network simulation makes it difficult to extend their model to include driving
simulation and human factors models.
TraNS (Piorkowski et al. 2008) and Veins (Sommer et al. 2011) connect the network
simulators ns-2 and OMNeT++ to the SUMO traffic simulator respectively. Like our work,
they both implement the TraCI (Wegener et al. 2008) interface and directly use the network
simulators to suggest mobility changes to SUMO where required. None of the above
simulators involve human-in-the-loop feedback via a driving simulator, a feature which our
unique 3-in-1 ITDNS includes, as described next.
III. INTEGRATED TRAFFIC-DRIVING-NETWORK SIMULATOR
The following sections will first introduce the specific traffic, driving, and networking
simulators used in this study, followed by how the three simulation types were integrated.
III.1. Traffic Simulator
The model that was selected for this study is PARAMICS v6.0, a suite of microscopic traffic
simulation software for modeling freeway and arterial networks. The suite consists of the
following three modules: (1) Modeler; (2) Processor; and (3) Analyzer (Quadstone, 2010).
Modeler is used to define the network of nodes, links and junctions (intersections), as well as
the specifics of the traffic demand and vehicle profiles. Simulations can be performed in a
high speed non-graphic mode, as well as in 2-D or 3-D visual modes. Processor allows
multiple simulations for testing various configurations or alternative situations. Finally,
Analyzer provides various tools to review several measures of effectiveness (MOEs) such as
vehicle counts, speed, delay, travel time, and queues.
The primary reason for selecting this particular software to serve as the microscopic traffic
simulator within the ITDNS framework is the fact that it has an add-on module called
Programmer, which is a comprehensive development Application Programmer Interface
(API). Programmer allows the user to retrieve output values, assign input parameters, and
Page 8 of 26
augment the core simulation with new functions and driver behavior. This capability was
critical for integrating the traffic, driving and networking simulators together.
III.2. Driving Simulator
UB’s driving simulator which was utilized in the current study consists, primarily, of a six
degree-of-freedom electrically actuated motion platform. Two passengers are accommodated
in a front-seat vehicle passenger cabin (a 1999 Ford Contour). The driver supplies inputs to
the simulator using a steering wheel (force feedback, with a 900˚ rotational stroke), three
pressure modulated/adjustable floor pedals (gas, brake, and clutch), and a console gear-shifter
with programmable buttons. Additional simulation hardware includes an Emergency-STOP
switch, a four-screen (Front, Left, Right, and Rear-view, hexagonally arranged), front-
projected XVGA+ visualization system (4:3, 8’ x 6’, 1400x1050 pixel resolution), and a 2.1
channel stereo sound system. Figure 1 is an image of the exterior of the vehicle cabin and
frontward simulation screens.
III.3. Network Simulator
The open source simulator ns-2 was chosen to provide networking simulation function in the
ITDNS, because it allows users to examine and modify its internal components, and is thus
already widely used by the academic community. In the majority of the previous studies
reported in the literature, however, ns-2 was used for the simulation of VANETs in one of
two ways: (1) by reading in a trace file generated by a traffic simulator and mimicking the
movements or (2) through simple logic that was written into ns-2 itself. However, neither of
these techniques can simulate a realistic VANET where drivers follow certain driving
patterns on the road and also respond intelligently to CTS messages in real time. This has led
researchers to explore how ns-2 can be integrated to run with other specialized traffic
simulators (Wegener et al. 2008) in parallel. This study builds upon this third integration
approach. Moreover, the study adds the ability to consider human factors issues, through the
integration with the driving simulator.
III.4. ITDNS
Fig. 2 shows the overall architecture of our ITDNS. As can be seen, the integration of
PARAMICS and DS is implemented via a two-way data exchange. The actions of the human
driver in the DS are first translated, via the DS’s vehicle dynamics model, into several vehicle
state outputs (e.g., position, orientation, velocity, acceleration etc.). At each time step, those
state outputs are delivered to PARAMICS such that, one chosen vehicle’s (referred to herein
as the subject vehicle) speed and position/orientation are overridden by the actions of the
human driver in the DS. Overriding the default behavior of the TS is achieved using a plug-in
written in C++ and which utilizes several of PARAMICS custom API’s. Meanwhile, the
positions, speeds and accelerations of the other vehicles in the vicinity of the subject vehicle
are exported from PARAMICS to the DS, to represent the background traffic which the
human driver observes and reacts to. Because the positions and speeds of the background
traffic are determined by PARAMICS, background traffic in the DS is intelligent and reacts
to the actions of the human driver.
Simultaneously, PARAMICS and NS-2 are integrated to allow them to run in parallel, by
adopting the TraCI interface developed by the EPFL team (Wegener et al., 2008), as a
Page 9 of 26
starting point in our work. In doing so, a complete feedback loop is implemented to send
results from the NS back to the TS for further action. This will allow for eventually
implementing CTS-responsive driver models within the TS, which mimics how drivers act
upon the CTS warning messages they receive.
IV. INTEGRATION CHALLENGES
Achieving the desired integration among the three simulators involved overcoming several
challenges. The following section will briefly describe those challenges and how they were
addressed in the current study. The challenges of integrating the TS and DS are discussed
first, followed by the TS-NS integration challenges.
IV.1. TS-DS Integration Challenges
IV.1.1. Co-ordinate System
Paramics represents the traffic network as a link-node system, and thus describes a vehicle’s
location by tracking the link it is on, and the distance to the link’s stop line. In UB’s DS, the
simulation world has no knowledge of links and nodes, but rather tracks coordinates of the
vehicle’s location. Routines were thus developed to allow for location translation. After
receiving the coordinates from the DS, Paramics API pinpoints the best match for the
distance to the link end and current lane, and then updates the subject vehicle’s location.
IV.1.2. Driving Environment
The driving environment refers to the virtual world that human driver sees on the four-screen
projection display, which includes building models, trees, road signs, etc. Through the course
of this study, two rendering approaches were implemented and evaluated, i.e. a generic
approach and a custom-tailored approach. Initially, the research’s focus was to evaluate the
feasibility of the ITDNS concept, and therefore, the in-built 3-D visualization model in
Paramics was utilized to reproduce the buildings, trees, and road signs of the virtual world.
Obviously, the advantage of this approach is that it cut down on the development effort, but
naturally that comes at the cost of fidelity of the virtual environment in which the driver is
immersed.
In order to make the visual environment more realistic and to allow for full control over that
environment, the researchers later decided to re-create the full virtual environment, from
scratch, in the DS. In doing so, the researchers took advantage of a number of publically
available data sources including Google Map and Google Sketchup. When building models
were not already available, the researchers created their own models based on real world
photographs. By building their own virtual environment, the researchers were able to
represent what the human driver would see in the rear mirror. They were also able to make
the environment more realistic than that provided by PARAMICS built-in animation and
visualization routines. For example, in the implemented ITDNS, the braking lights and the
turn signals can be displayed.
IV.1.3. Achieving acceptably fluid motion of subject vehicle
Within TS, only the “gross” motion of vehicles is of concern, with update rates of 1-30 Hertz,
typical. Within a DS, much more fluid vehicle motion is essential to ensure an authentic
experience for the human participant, with update rates of 60-100Hertz, typical. Simple yet
Page 10 of 26
efficient measures (e.g. prediction, extrapolation, & interpolation) were employed to account
for this discrepancy between update rates.
IV.1.4. “Smooth” Turning
Due to the unavailability of an API function which can retrieve and override the vehicle
orientation in PARAMICS, the vehicles’ turning movements were a bit “awkward” in the
virtual world. This is because PARAMICS’s approximation of the vehicle heading direction
while turning was somewhat abrupt when viewed by the human subject in the virtual world.
To address this and to allow for a smoother turning maneuver, an algorithm was designed to
allow for the gradual interpolation between the orientations of the exiting link and the target
link.
IV.1.5. Rapid Communication between simulators
Various data (e.g., vehicle identifiers, positions, speeds, and orientations, and intersection
traffic signals status) had to be rapidly transferred between TS and DS. Due to the constraints
stated above, this transfer had to take place at 30 Hertz minimum, with latency minimized to
an acceptable degree. TCP and UDP Socket protocols were designed to achieve this.
Moreover, to improve efficiency, the system was designed to only communicate the
information of those vehicles that are located in the vicinity of the subject vehicle (i.e. the
vehicle controlled by the human driver). Information pertaining to only those vehicles that
are located within a circle (whose radius is a user-specified parameter) would be
communicated.
With respect to information other than the vehicles’ locations, such as infrastructure-based
information (e.g. traffic signal status), such information did not require 30 Hertz rate
communication. Instead, those data exchange rates were downgraded to 1 Hertz, also for
better system performance. Moreover, the system was designed to only query the
intersections that are close enough to the subject vehicle, once again to improve efficiency.
IV.1.6. “Staging” Drivers’ Errors
Simulated vehicles in PARAMICS strictly operate under car-following and lane changing
model, which leads to an ideal, error-free driving world, and therefore “zero” incidents. To
allow for testing CV applications, it was thus necessary to manipulate the model in a way that
“running red light” and “rear end collision” can be reproduced in the simulation world. This
was achieved via the use of PARAMICS different API functions to allow for overriding the
default, error-free behavior of vehicles.
IV.2. TS-NS Integration Challenges
IV.2.1. Fast-moving Nodes
The first challenge encountered stemmed from the fact that the default implementation of
TraCI (the interface used in this study to integrate the TS and NS) did not work correctly on
the NS side, once vehicles started moving at higher speeds (e.g., 45 mph). Another challenge
was that while NS-2 maintained an ordered list of the nodes that is used to simulate wireless
transmission, TraCI made changes to the node positions without maintaining the ordering and
this led to the messages being delivered incorrectly. The TraCI implementation was thus
slightly modified to preserve the ordering and to account for the high speeds.
Page 11 of 26
IV.2.2. The large number of nodes
It was soon recognized that NS-2 was not capable of creating nodes continuously to match
the rate at which PARAMICS creates them. Declaring a large number of wireless nodes in
NS-2 at the very beginning and then activating them as new vehicles are created in
PARAMICS also degraded NS-2’s performance rapidly. To solve this problem, the NS-2
simulation was kept constrained to a dynamic circular area around the subject vehicle which
moved with it. Functionality was also added to force NS-2 to recycle unused nodes as new
vehicles enter and old vehicles exit the circle in PARAMICS. Other issues that were
addressed included timing the two simulators to run in sync, and managing node/vehicle IDs
between the two simulators.
V. ITDNS Application
To demonstrate the capabilities of the developed ITDNS, and its unique advantages in terms
of evaluating CV applications, the integrated simulator was then used to evaluate the likely
benefits of a recently proposed CV application known as eco-vehicle speed control at
signalized intersections (referred to herein as eco-signals for brevity) . The basic premise of
the eco-signals concept is that if drivers of vehicles approaching a signalized intersection
have accurate information about the upcoming signal status (e.g. the time remaining for the
green phase or when the green phase is to start), the vehicle speed may be adjusted
accordingly so as to avoid idling and/or hard deceleration and acceleration maneuvers,
operating modes that are associated with increased fuel consumption and emissions rates.
The eco-signal concept has been attracting increased attention among transportation
researchers in the last few years. For example, Wu et al. (2010) researched the impact of two
types of advanced driving alert systems (ADAS), providing Traffic Signal Status (TSS)
information to drivers, on CO2 emissions. Mandava et al. (2009) developed arterial velocity
planning algorithms based on signal phase and timing information communicated from the
signal controller to equipped vehicles, and evaluated likely environmental benefits using
simulation and the University of California-Riverside CMEM model. Dobre et al. (2012)
presented algorithms for determining the approaching vehicle speed profile which would
minimize fuel consumption. They then evaluated the likely benefits of their algorithm using
an integrated vehicular traffic and network communication simulator called VNSim
(Gradinescu et al., 2007). In all those studies, however, human factors issues related to the
eco-signal concept of operations do not appear to have been explicitly addressed.
Specifically, those studies primarily depended on using traffic simulation (without
considering a human in the loop), and hence could not assess driver response or acceptance.
Thanks to the 3-in-1 integrated simulator developed in this study, it is now possible to
explicitly study human factors related issues as briefly described below.
V.1. ITDNS Setup for Evaluating the Eco-signal Application
Given that the eco-signal application is primarily utilized in this study to demonstrate the
capabilities of the ITDNS, the focus is on a simple application where the goal is to control the
speed of the approaching vehicle, assuming that the signal parameters are kept unchanged.
To implement the eco-signal application in the ITDNS, a speed panel was projected at the
lower corner of the front DS display as shown in Figure 3. The panel has the following
features. First, the orange dial rotates to reflect the current driving speed. The number in the
Page 12 of 26
lower box reflects the recommended speed that the driver is advised to maintain. Moreover,
the box background color references three statuses: (a) green for when the driver is advised
to speed up (i.e. the recommended speed is higher than the current driving speed); (b) red for
when the recommendation is to slow down; and (c) yellow for when the recommendation is
to maintain the current speed. In setting these color indications, an eco-threshold of 10% of
the current speed is utilized (i.e. if the recommended speed is within 10% of the current
speed, the driver is asked to maintain the speed he/she is driving at).
With this setup, as soon as a connected vehicle (i.e. equipped vehicle) entered the
communication range of the eco-signal, the signal controller would send the signal phasing
and timing information to the vehicle. Based on this, the system would examine the
following three options in the following order:
Option (i): If the connected vehicle can arrive at the intersection within the green light
window at current speed, then it shall cruise at the same speed (Figure 4.a);
Option (ii): If option i is not available, the connected vehicle examines whether speeding up,
while remaining below the posted speed limit for the road, would allow it to catch up the
green window ahead of the Estimated Time of Arrival at the intersection (ETA) at the current
speed (Figure 4.b).
Option (iii): In this case, the connected vehicle has to slow down, so that it can make the
arrival at the next green window after the original ETA (Figure 4.b);
Two additional extensions of the ITDNS were required before evaluating the likely benefits
of eco-signals. The first extension involved implementing a customized collision avoidance
algorithm to ensure that the recommended speed would not cause the connected vehicle to
crash into the vehicle in front. To do this, the connected vehicle would examine the speed
difference and space headway relative to the lead vehicle, and determine the maximum safe
operating speed. The second extension involved using PARAMICS API functions to export
the second-by-second speed trajectory of the connected vehicle. With the speed trajectory
exported, the latest Environmental Protection Agency’s (EPA) MOVES2010 model was used
to calculate the emissions and energy consumption of the vehicle. The details on how the
MOVES2010 model may be used to calculate emissions and fuel consumption based on
second-by-second vehicle speeds, the reader is referred to the paper by Zhao and Sadek
(2013).
V.2. Experimental Setup
In order to evaluate the likely environmental benefits of the eco-signal concept, a 2.5-mile
long segment of an arterial corridor consisting of ten signalized intersections, connecting the
University at Buffalo’s South and North Campuses, was modeled in the ITDNS (as shown in
Figure 5). Field counts of the turning movements at the intersections were collected, and
translated into the Origin-Destination demand matrix required to run the PARAMICS model,
using the PARAMICS model Estimator (an add-on module for PARAMICS which allows for
estimating O-D demand matrices from turning movement counts). In addition, exact signal
Page 13 of 26
timing and phases were collected and coded in the model to reproduce the realistic traffic
signal system settings.
Five drivers were then solicited and asked to drive the 2.5 mile segment; that number was
deemed adequate given that the purpose here is to basically demonstrate the capabilities of
the ITDNS (for a more thorough investigation of the likely benefits of eco-signals, a larger
experiment involving more drivers is needed). Each driver was asked to perform two test
runs, one with the eco-signal app and one without. For the base run, the test drivers were
instructed to following their typical driving pattern. As for the eco-signal run, they were
requested to follow the speed recommended by the eco-signal system. The second-by-second
speed trajectories of their drives were then imported to MOVES2010’s input database for
energy consumption and emission calculations. The comparison involved comparing the
estimates for the energy consumption, Carbon Monoxide (CO) and Nitrogen Oxides (NOx)
for the drivers when using the eco-signal application, compared to the base case without the
system.
V.3. Results
Figure 6 compares the energy consumption and emissions rates for the ten runs (five with
eco-driving on and the other five off). Specifically, the solid lines represent the runs with eco-
signal application turned on, while the dashed ones depicted the off cases. As previously
mentioned, the analysis involved comparing energy or fuel consumption levels (the lines with
the squares), the CO emissions (the circles), and the NOx emissions (the triangles).
As can be seen, while the exact savings vary from one driver to another, for all five drivers,
the energy consumption and the emissions rates are lower for the eco-signal case compared to
the base case. Depending upon the aggressiveness level of each driver, the savings ranged
between 4% to 14% percent for the energy consumption, between 6% and 35% for CO, and
between 6% and 42% for NOx. The average values (average for the five drivers) for the
energy and emissions savings are shown in Table 1, where it can be seen that the savings
were 9% for energy consumption, 25% for NOx, and 18% for CO.
Besides the environmental benefits of the eco-signal application, the application appears to
have the potential to improve the overall safety of traffic operations along the modeled
corridor as well. The safety benefits stem from the reductions that the application seems to
bring in the magnitudes of the maximum accelerations and deceleration, and in maximum
speed, as can also be seen from Table 1. As can be seen, the values for the maximum
accelerations and deceleration seem to have dropped by more than 1/3 of their original
values. At the same time, the maximum speed was reduced by almost 20% (note that the
speed limit along the modeled corridor varied between 30, 35 and 40 mph). Lastly, it should
be noted that the environmental and safety benefits came at a modest expense in terms of a
5% increase in travel time.
A look at Figures 7 and 8, which compare the speed and acceleration/deceleration profiles for
the vehicles with the eco-signal application activated and deactivated, also clearly confirms
the potential for the eco-signal application to improve sustainability as well safety. As can be
seen from Figure 7, the eco-signal application, by controlling the approaching vehicle speed,
managed to prevent the vehicles from fully stopping at time steps 45, 260 and 290 seconds.
Page 14 of 26
On the other hand, Figure 8 clearly shows the ability of the eco-signal application to
smoothen the vehicle trajectory by eliminating hard acceleration and deceleration maneuvers.
VI. Conclusions and Future Work
In this paper, an integrated 3-in-1 simulator was developed. The integrated simulator
combines the main features of a traffic simulator, a driving simulator, and a communications
networking simulator. Several challenges were encountered during the integration process
which necessitated devising a number of innovative techniques in order to overcome them.
To demonstrate the advantages of the developed simulator in evaluating CV and CTS
applications, it was used to evaluate and quantify the likely environmental benefits of the
eco-signal concept. Among the main conclusions of the study are:
(1) The integration of traffic, driving and networking simulator is technically feasible,
although a number of challenges need to be addressed to achieve full integration;
(2) The unique advantage of the ITDNS developed in this study stems from its ability to
account for human factors related issues in the design and evaluation of CTS applications;
(3) The eco-vehicle speed control concept at signalized intersections appears to have the
potential to result in tangible environmental benefits. Specifically, for the simple case study
considered in this research, the concept resulted in average reductions of about 9% for energy
consumption, 25% for NOx emissions, and 18% for CO emissions. This came at the expense
of only about 5% increase in travel time. More comprehensive tests however are required to
validate this conclusion;
(4) Besides its environmental benefits, the eco-signal concept appears to be also capable of
yielding safety benefits. The experiments performed herein seem to indicate that the concept
has managed to smoothen the vehicles’ trajectories, by eliminating hard accelerations and
decelerations maneuvers.
The researchers are currently working on a number of different future research initiatives to
complement the work presented herein. First, with respect to the ITDNS, the researchers are
planning a field validation of the integrated simulator by compared the behavior of the human
subjects in the virtual environment to their behavior in the real-world. Second, with respect
to the eco-signal concept which was primarily used herein for demonstration, the researchers
are planning a more comprehensive evaluation involving a much larger sample of drivers.
They are also planning to refine the algorithm used to calculate the recommended approach
speed.
Acknowledgements
This work was supported in part by National Science Foundation of USA CPS Program
(Grant No. NSF-CPS-1035733) and the Cisco University Research Program.
Page 15 of 26
References
Archer, J. (2000). “Developing the potential of micro-simulation modeling for traffic safety
assessment”, 13th ICTCT Workshop. Corfu, Greece.
Barcelo, J. and Ferrer, J. L. (1998). “AIMSUN2: Advanced Interactive Microscopic
Simulation for Urban Networks.” Departmento de Estadistica e Investigation Operative,
Faculdad de Informatica, Universidad de Cataluna.
Baret, M., (1978). “Six degrees of freedom large motion system for flight simulators”,
AGARD Conference Proceedings, No. 249, Piloted Aircraft Environment Simulation
Techniques, pp. 22-1/22-7.
Barolli, L., Mino, G., Xhafa, F., Durresi, A., Koyama, A., (2010) Improvement and
Performance Evaluation of CAVENET: A Network Simulation Tool for Vehicular Networks,
Distributed Computing Systems Workshops (ICDCSW), 2010 IEEE 30th International
Conference on , vol., no., pp.118-125, 21-25 June 2010
Ceylan, H., Bell, M., (2004) Traffic signal timing optimisation based on genetic algorithm
approach, including drivers’ routing, Transportation Research Part B: Methodological,
Volume 38, Issue 4, May 2004, Pages 329-342
Ciuffo, B., Punzo, V., and Torrieri, V. (2009). “Integration of Driving and Traffic
Simulation: Issues and First Solutions”, 88th Annual Transportation Research Board
Meeting, Washington, D.C.
Claudinon, B. and Lievre, J., (1985). “Test facility for rendezvous and docking”, 36th
Congress of the IAF, Stockholm, pp. 1-6.
Cunto, F., and Saccomanno, F.F., (2006). “Evaluation of Safety Countermeasures at
Intersections Using Microscopic Simulation”, 85th Annual Transportation Research Board
Meeting, Washington, D.C.
De Marco, G., Tadauchi, M., Barolli, L, (2007) CAVENET: Description and analysis of a
toolbox for vehicular networks simulation, Parallel and Distributed Systems, 2007
International Conference on , vol.2, no., pp.1-6, 5-7 Dec. 2007
Dobre C., Szekeres, A., Pop, Florin, Cristea, V., (2012) Intelligent Traffic Lights to Reduce
Vehicle Emissions, Proceedings 26th European Conference on Modeling and Simulation.
DriveSafety Website: http://www.drivesafety.com/
Gradinescu, V., Gorgorin, C., Diaconescu, R., Cristea, V., Iftode, L., (2007) , Adaptive
Traffic Lights Using Car-to-Car Communication, Vehicular Technology Conference, 2007.
VTC2007-Spring. IEEE 65th , vol., no., pp.21-25, 22-25 April 2007
Hulme, K., Huang, S., Sadek, A.W., and Qiao, C., (2010). Next Generation Human-in-the-
loop Transportation Simulation Modeling with Application to Eco-Driving Evaluation,
Page 16 of 26
Proceedings of the 2010 Inter-service / Industry Training, Simulation, and Education
Conference (I/ITSEC), Orlando, FL.
Ikeuchi, K., Kuwahara, M., Suda, Y. and Tanaka, Y., (2005) Sustainable ITS Project
Overview: Mixed Reality Traffic Experiment Space under Interactive Traffic Environment
for ITS, Proceedings of IEEE Intelligent Vehicles Symposium, 2005.
Gartner, N., C. J. Messer, and A. K. Rathi (Editors). (2005). Traffic Flow Theory; A State-of-
the-Art Report. Available at http://www.tfhrc.gov/its/tft/tft.htm. Last retrieved on June 18,
2010.
Jenkins, J.M., (2005). “Integrating driving simulation and traffic simulation: A case study”,
Proceedings of the Annual Conference of the Canadian Society for Civil Engineering.
Jin, M. and Lam, S.-H., (2003). “A Virtual-Reality Based Integrated Driving-Traffic
Simulation System to Study the Impacts of Intelligent Transportation Systems (ITS)”,
Proceedings of the 2003 International Conference of Cyberworlds, Nanyang, Singapore.
Li, C., Shimamoto, S., (2011) A real time traffic light control scheme for reducing vehicles
CO2 emissions, Consumer Communications and Networking Conference (CCNC), 2011
IEEE , vol., no., pp.855-859, 9-12 Jan. 2011
Mangharam, R., Weller, D., Rajkumar, R., Mudalige, P., Bai, F., (2006) GrooveNet: A
Hybrid Simulator for Vehicle-to-Vehicle Networks, Mobile and Ubiquitous Systems:
Networking & Services, 2006 Third Annual International Conference on , vol., no., pp.1-8,
July 2006
NS-2 Simulator website: http://nsnam.isi.edu/nsnam/index.php/User_Information
Piorkowski, M., Raya, M., Lugo, A., Papadimitratos, P. and Grossglauser, M. et al., (2008)
TraNS: Realistic Joint Traffic and Network Simulator for VANETs, in ACM SIGMOBILE
Mobile Computing and Communications Review, vol. 12, num. 1, p. 31--33, 2008.
PTV-AG., (2004). “Vissim Microscopic Traffic and Transit Simulation User Manual,
v.3.70.” Karlsruhe, Germany.
Pursula, M. (1999). “Simulation of Traffic Systems - An Overview”, Journal of Geographic
Information and Decision Analysis, Vol.3, No.1, pp. 1-8.
Rakha, H., Kamalanathsharma, R.K., (2011) Eco-driving at signalized intersections using
V2I communication, Intelligent Transportation Systems (ITSC), 2011 14th International
IEEE Conference on , vol., no., pp.341-346, 5-7 Oct. 2011
Sommer, C., German, R., Dressler, F., (2011) Bidirectionally Coupled Network and Road
Traffic Simulation for Improved IVC Analysis, Mobile Computing, IEEE Transactions on ,
vol.10, no.1, pp.3-15, Jan. 2011
Stevanovic, A., Stevanovic, J., Zhang K., and Batterman S., (2009) Optimizing traffic control
to reduce fuel consumption and vehicular emissions: an integrated approach of VISSIM,
Page 17 of 26
CMEM, and VISGAOST, Proceedings of the 88th Annual Meeting of the Transportation
Research Board, Washington, DC, 2009.
Stewart, D., (1965). “A platform with six degrees of freedom”, The Institution of Mechanical
Engineers, 180 (15), pp. 371-384.
STISIM Website: http://www.stisimdrive.com/
Transportation Research Board (TRB), (2000). “Highway Capacity Manual”, TRB, National
Research Council, Washington, D.C.
Quadstone PARAMICS, PARAMICS Version 6.6 User Manual Online, Edinburgh, Scotland,
2010.
VirtualDriver Website: http://www.driverinteractive.com/index.php
Vuyyuru, R., and Oguchi, K., (2007) Vehicle-to-Vehicle Ad Hoc Communication Protocol
Evaluation Using Realistic Simulation Framework, the Fourth IEEE/IFIP Wireless on
demand Networks and Services, Austria, 2007.
Wang, S., Lin, C., (2010) NCTUns 6.0: A Simulator for Advanced Wireless Vehicular
Network Research, Vehicular Technology Conference (VTC 2010-Spring), 2010 IEEE 71st ,
vol., no., pp.1-2, 16-19 May 2010
Wegener, A., Piórkowski, M., Raya, M., Hellbrück, H., Fischer, S. and Hubaux, J., (2008)
TraCI: an interface for coupling road traffic and network simulators, Proceedings of the 11th
communications and networking simulation symposium (CNS'08), 2008.
Wu, G., Boriboonsomsin, K., Zhang, W., Li, M., and Barth, M., (2010) Energy and Emission
Benefit Comparison between Stationary and In-Vehicle Advanced Driving Alert Systems,
89th Annual Meeting of the Transportation Research Board, Washington, D.C.: 2010.
Zegeye, S.K., De Schutter, B., Hellendoorn, H., Breunesse, E., (2009) Reduction of travel
times and traffic emissions using model predictive control, American Control Conference,
2009. ACC '09. , vol., no., pp.5392-5397, 10-12 June 2009
Zhao, Y., and Sadek, A. W., (2013) Performance Evaluations of Approaches to Integrating
Microscopic Traffic Simulators with Emissions Models for Project-level Emissions Analysis,
Accepted for presentation at 2013 Annual TRB Meeting.
Page 18 of 26
Figure 1. Motion Simulator
Page 19 of 26
Figure 2. Architecture of the ITDNS
Page 20 of 26
Figure 3. Eco-signal Display Panel
Page 21 of 26
(a) Option i (b) Option ii, iii
Figure 4. Eco-driving Distance-Time Diagram
Page 22 of 26
Figure 5. Paramics Arterial Test Bed, Connecting UB South and North Campus
Page 23 of 26
Figure 5. Energy, NOx, CO w/ and w/o Eco-Driving
Page 24 of 26
Figure 6. Speed-Time Diagrams w/ and w/o Eco-Driving
Page 25 of 26
Figure 8. Accel-Time Diagrams w/ and w/o Eco-Driving
Page 26 of 26
Table 1. Average Energy, NOx, CO, Time, Max Accelerations and Decelerations and Speeds
Eco On
Eco Off
Savings
Travel Time(s)
325.4
309
-5%
Energy (J)
1.44E+07
1.58E+07
9%
NOx (kg)
1.78E-03
2.36E-03
25%
CO (kg)
1.36E-02
1.66E-02
18%
Max Accel. (mphps)
5.002
7.826
36%
Max Decel. (mphps)
-8.862
-13.658
35%
Max Speed (mph)
36.42
44.974
19%
