Content uploaded by Martin Margreiter
Author content
All content in this area was uploaded by Martin Margreiter on Nov 26, 2015
Content may be subject to copyright.
Effects of cooperative systems on traffic safety and efficiency – results of the German
sim
TD
project
Florian Schimandl
Chair of Traffic Engineering and Control, Technische Universität München, Arcisstraße 21,
80333 München, Germany
Matthias Baur
Chair of Traffic Engineering and Control, Technische Universität München, Arcisstraße 21,
80333 München, Germany
Silja Hoffmann
Chair of Traffic Engineering and Control, Technische Universität München, Arcisstraße 21,
80333 München, Germany
Sebastian Gabloner
Chair of Traffic Engineering and Control, Technische Universität München, Arcisstraße 21,
80333 München, Germany
Martin Margreiter
Chair of Traffic Engineering and Control, Technische Universität München, Arcisstraße 21,
80333 München, Germany
Abstract
In this paper the approach, the challenges and the results of determining the effects of
cooperative Intelligent Transportation Systems on traffic efficiency and traffic safety using the
example of the sim
TD
project are presented. It is shown that the integrated approach using a
driving simulator and traffic simulation offers a valuable complement to the analysis of real
traffic and that some challenges associated with using real world data could be overcome by
using simulation.
The effects of selected evaluated cooperative functions in urban and inter-urban contexts in
traffic simulations as well as real traffic regarding the performance indicators specified
uniformly for all test environments (traffic simulation, field operational test) are described.
1. Introduction
Intelligent Transportation Systems (ITS) based on cooperative technology are considered a
valuable means to contribute to the improvement in traffic efficiency, traffic safety as well as
the reduction of environmental impacts. Various applications that focus on efficiency and
safety were developed and tested in the Field Operational Test (FOT) and the simulation
laboratory of the sim
TD
project (see section 2).
In addition to conducting additional tests in a FOT, traffic simulation offers a practical means
to simulate and compare different well-defined (cooperative traffic) scenarios to assess the
effects under 100% controlled constraints. Particularly, by comparing the different results
under ceteris paribus conditions, the effects of the various applications on overall traffic can
be estimated much easier than in real traffic (see section 6).
The integrated approach by combining different test environments: a FOT in real traffic, its
interactions with a traffic simulation environment and the usage of data from other sources like
a driving simulator presented in [1] is used for further analysis. It delivers a solution for
evaluating the sim
TD
applications with regard to traffic efficiency and traffic safety by using
traffic simulations based on behavioral data of a driving simulator as well as real world
parameters for calibration purposes.
In the following section, the sim
TD
project is presented. The concept of the simulation lab,
which consists of a driving simulator and traffic simulations, is shown in section 3. The sim
TD
traffic simulation environment and the detailed scenario setup is illustrated in section 4 and
section 5, supplemented by the results of the impact evaluation of selected applications in
section 8. The impact assessment is complemented by explaining the usage of statistical tests
and the challenges in getting valuable results from the FOT in section 6 and section 7. A short
conclusion as well as an outlook is given in the end (see section 9).
2. The sim
TD
project
1
In the German research project sim
TD
various in-vehicle and central applications based on
cooperative technology were developed and tested in reality and simulation [2]. The field
operational test took place from July 2012 to December 2012 around the Hessian city of
Frankfurt am Main and was complemented by tests in the simulation laboratory (see
section 3).
The FOT testbed consisted of routes on motorways, rural highways and urban streets where
a fleet of up to 120 vehicles was driven in pre-planned and (mostly) controlled tests in real
traffic situations. The target was to evaluate the different applications in real (traffic) situations
with regard to their technical functionality and to assess the impact on driving and traffic
efficiency as well as driving and traffic safety.
Within the sim
TD
project, the planning of the tests for traffic efficiency/safety analysis, cloning
the behavior of the applications in traffic simulation as well as assessing the results with
respect to traffic impacts in both test environments was carried out.
3. The sim
TD
simulation laboratory
As a complement to sim
TD
’s real world testing sites, the field operational test was accompanied
by a simulation laboratory consisting of
• a physical driving simulator as well as
• a traffic simulation environment.
The driving simulator setup and respective results can be found in [3]. The two simulation
facilities maintained an integrated design approach regarding common scenario and
evaluation schemes as well as input data exchange interfaces. Scenarios in both
environments were designed conjointly and according to the real-world scenarios for the sim
TD
1
http://www.simtd.de
applications. Driving behavior trajectory records from several single real drivers in the virtual
driving environment could then be aggregated at a scenario level. This way, cumulative speed
distributions lead to desired speed input generating the basis for the traffic simulation’s driving
behavior. More details on this integrated design approach can be found in [1], [4] and [5].
4. The sim
TD
traffic simulation environment
The aforementioned extrapolation of single driver’s behavior allows for large-scale traffic
related evaluations, which would not be possible in the real world test or in driving simulator
studies.
The sim
TD
traffic simulation is based on a specific model framework that consists of
components such as
• a specific implementation of each sim
TD
application,
• driver behavior based on driving simulation studies,
• optional roadside equipment,
• virtual traffic control modules, e.g. adaptive traffic signal control as required,
• a virtual representation of the underlying road network,
• traffic input and assignment according to traffic measurements on respective real roads
and
• a message transmission module (VCOM) which is described in [6].
The VCOM message transmission module proved to be a perfect fit in terms of emulating
vehicular message transmission appropriately. Current work on comparing the model’s base
functions to real-world message transmission measurements shows promising results towards
a comprehensive validation of this communication model. Integrated scenario setup, as
described in section 5, proved to be helpful in gathering appropriate base data for the traffic
simulation’s driving behavior input from preceding driving simulator study results. Extensive
simulation studies based on this input encompassed about 1600 hours of simulation time and
produced around 1000 GB of raw trajectory and traffic measurement data.
After the simulation studies were finished and reported, results could be processed in a final
step. Traffic related estimations on performance indicators (see [7] for details) such as
simulated vehicle’s average travel times, mean speeds or time gaps build the basis for a
broader economic assessment.
5. Scenario setup for comparisons
Traffic Simulation
The integrated scenario setup approach described in [1] and [5] proved to be valuable for the
simulation and FOT evaluation. The sim
TD
scenario and evaluation design was developed with
the goal of generating comparable and valid test results from different simulation environments
as well as avoiding redundancy. Constraints associated with each environment are important
to consider in the design.
Scenarios in traffic simulation were adapted to the respective requirements of each system
under test. This refers to
• the road network,
• special situations, e.g. road construction sites or traffic jams and
• specific traffic control.
The distinct road network models were selected and modeled in detail, based on the following
requirements. Motorway routes were modeled for the analysis for safety applications, a large
network (consisting of motorways and rural highways) with route alternatives for routing
applications and urban scenarios for traffic-adapted signal control and the traffic light phase
assistant.
Variations of these scenarios proved to allow for estimations in various aspects such as
• road network selection,
• variation of the vehicle equipment rates,
• variation of the communication medium (UMTS, ITS-G5) and ITS road side station
equipment rates or
• traffic flow variation.
In traffic simulation, all scenario constraints (such as the road type or the traffic situation) can
be setup exactly. An example of different scenarios can be seen in the following table:
Scenario
ID
Road Traffic
characteristics
Vehicle
equipment rate
Roadside
equipment
V2X
Message
transmission
Specific condition
1 A661 Dense traffic None None ITS-G5
Limited queue visibility
2 A661 Dense traffic 20% None ITS-G5
Limited queue visibility
3 A661 Dense traffic 20% 3 Stations ITS-G5
Limited queue visibility
… … … … … …
Table 1: Example of scenario definitions in traffic simulation
The equipment rate was used as the primary independent variable for the evaluation.
Simulations were always conducted for a baseline scenario (vehicle equipment rate 0%) and
for vehicle equipment rates of 20%, 50% and 80%.
Field Operational Test
In contrast to the simulation environment, a FOT does not allow for the control of the
constraints of the tests completely. Even though the tests leading to traffic related evaluation
were mainly scripted to get the best results regarding the evaluation of each application, some
challenges occurred (see section 6). For example, the routes of some vehicles varied
according to the driver’s behavior, or the weather and traffic situations turned out to be different
than expected. This often led to little reproducibility and a small sample size that had to be
taken into account for the evaluation. Hence, the scenario definition in reality (compared to
simulation) is more of a reaction to the situation (ex-post analysis) than a predefined setup.
6. Challenges of FOT analysis
As mentioned in the previous section, the analysis and evaluation of the FOT imposed
numerous tasks. There was, of course, a limited number of test runs that could be evaluated.
Additionally, to get valuable results the data had to be pre-processed by various filtering
algorithms. The filtering process started at a very basic level, namely the test control itself,
where it was ensured that a test vehicle was initialized properly for each test run. This
addresses mainly communication of each vehicle with the test control center. In case of
communication failures or faulty setup of the scenarios, the test vehicle had to be removed
from the evaluation procedure. In a next step, the data delivered by the sim
TD
-proxy to the
CAN-Bus system had to be examined for corrupt values or discontinuous logging. In some
cases, GPS signals were not logged frequently. Some vehicles also did not log values such
as ‘relative speed’ or ‘distance to the front vehicle’ due to missing sensors. Another not
negligible issue was the CPU load of both the Vehicle Application Unit (VAU) and the Car
Communication Unit (CCU) which sometimes stayed at 100% for a long time and led to very
high latency of information flow. Thus, sometimes in-time presentation of information for the
driver via the Human Machine Interface (HMI) in the vehicle could not be guaranteed. Finally,
a rather complicated a posteriori analysis of HMI activity had to be implemented in order to be
able to distinguish between the equipped and the non-equipped groups of drivers.
Furthermore, it was necessary to take into account some additional circumstances for the
computation of performance indicators on the grounds of validated data. As already
mentioned, it is virtually impossible to guarantee a perfect test design (regarding traffic state)
in a FOT. The prevalent traffic state had to be determined ex post from the mean speed
profiles, thus allowing for clustering of different scenarios that could then be compared.
Another aspect to consider was the driving behavior of test drivers included in the evaluation.
For meaningful results, it is necessary that the way of driving represents the habits of the
majority off all drivers, which means not too aggressive nor too defensive. This was tested by
the indicators ‘Time-to-Collision’ and the speed profile. The results show that the control of
constraints is very important for the correct evaluation of such systems.
7. Statistics
After the computation of the Performance Indicators (PI) addressed in a specific use case, the
crucial step is testing the results for statistical significance, or in other words to determine
significant changes or shifts of a PI for two or more test groups (distributions of values of the
PI). This was done for the simulation as well as the FOT data.
The procedure of testing was similar for both two sample tests and multiple sample tests. After
testing for normal distribution in every sample and homogeneousness of variance between all
samples, a test was chosen to determine significant differences between the groups. The null
hypothesis stated that there are no differences between them. With a p-value less than 0.05
the null hypothesis was rejected and significant differences were assumed. To evaluate driving
efficiency/safety, two sample tests, the t-test and Welch test, were performed for the
comparison of the two groups (equipped versus not equipped). To evaluate the traffic
efficiency and safety, a comparison of several groups (for example scenarios with different
equipment rates) was performed using the ANOVA test (analysis of variance) or Kruskal-
Wallis test.
It is important to mention, that in the latter case, a post hoc procedure needs to be applied to
do pairwise testing between all single groups in case there are significant differences detected
by the test in all of them on the whole.
8. Results
The following section shows some of the main results of the impact analyses within the FOT
and the traffic simulation for selected applications and different scenarios.
Warning Applications
The sim
TD
Obstacle Warning informs and warns the driver before an obstacle, such as lost
cargo, poses a risk. The timing of the warning is calculated depending on the distance to the
obstacle and the speed of the vehicle approaching. By using only Vehicle-to-Vehicle (V2V)
communication via ITS-G5, no significant impacts on the surrounding traffic can be measured
within the traffic simulation for equipment rates lower than 20%. With higher equipment rates
positive impacts can be shown. All vehicles pass the critical situation with lower speed and
change the lane in advance.
The sim
TD
Congestion Warning provides information to the driver about dangerous tail ends
of traffic jams, such as congestions behind a curve or below a hilltop. Using only V2V
communication via ITS-G5 and for equipment rates of 20% and lower, significant benefits for
all vehicles (also the ones without sim
TD
system) within this critical situation can be shown in
the traffic simulation. Larger headways between the vehicles, lower deceleration rates and
lower velocities show a situation adequate behavior for all vehicles. The reaction of the
equipped vehicles influences the whole traffic positively. The higher the equipment rate, the
higher are the respective impacts. The analysis of the FOT data shows the same tendencies.
Vehicles using the sim
TD
system react earlier to the situation and drive with significant lower
speeds in front of the congestion back-end (see Figure 1).
Figure 1: Comparison of speed profiles for sim
TD
Congestion Warning (traffic simulation , FOT)
The sim
TD
Electronic Brake Light gives an early warning to the driver as soon as preceding
vehicles initiate a hard or emergency braking. This allows the driver to quickly adapt to
potentially dangerous situations even when the view is restricted by traffic ahead. Significant,
positive effects can be shown for equipment rates of 50% (via traffic simulation) using V2V
communication via ITS-G5 only. The deceleration rates are significantly higher than in the
baseline scenario.
The sim
TD
Road Works Warning System informs and warns a driver when approaching a
construction site. Using Vehicle-to-X communication with ITS road side stations (IRS),
significantly positive impacts can be found starting from the analyzed equipment rate of 20%
in traffic simulation. All vehicles reduce their speeds and change lanes earlier in front of the
hazardous location. The equipped vehicles influence the unequipped ones positively by
reducing their speeds as well. The analysis of the FOT data shows that the equipped vehicles
have to decelerate less strongly than the non-equipped ones.
The sim
TD
Traffic Sign Assistant informs the driver about currently valid traffic signs,
including variable message signs such as speed limits and road work restrictions. Data on the
location and content of the traffic signs originate from the ITS Central Station (ICS). The traffic
simulation shows that with an increasing equipment rate the vehicles drive with speeds much
closer to the speed limit.
Alternative Route Management
The sim
TD
Alternative Route Management redirects vehicles to alternative routes in case of
traffic jams or disruptions of traffic flow. The application uses existing intelligent detection and
management facilities of the ICS as well as dynamic information from the test fleet. The
objective is to offer comprehensive and reliable alternative route information to all drivers. The
effects depend strongly on local conditions and the availability of alternative routes. Within the
analyzed scenarios in the traffic simulation, very positive effects on travel times for all vehicles
could be proved. The benefits increase with increasing equipment rates from 20 over 50 up to
80%. Comparing the travel times for all routes a reduction of up to 7% (comparison of 80%
equipment rate with the baseline scenario) was realized. Especially the vehicles on the main
route benefit, because the rerouting of approaching vehicles decreases congestion (see
Figure 2).
Figure 2: Speed comparison sim
TD
Alternative Route Management
Traffic Control
The sim
TD
Local Traffic Adapted Signal Control application provides relevant information
from approaching vehicles about vehicle speeds and positions via IRS to the traffic light
controller. The traffic light controller uses this data to calculate traffic models for optimized
signal sequences, depending on how busy the roads are at that time. The effect of this
application depends on the quality of the traffic control for the baseline scenario and on the
traffic demand and its distribution at the crossing. Within the analyzed scenarios in the traffic
simulation positive effects on the overall traffic could be achieved with equipment rates already
beginning from 20%. In addition, travel times and the number of stops can be reduced
significantly. Travel times could be reduced up to 9% and the number of stops up to 16 %
(80% equipped vehicles compared to the baseline scenario, see Figure 3).
Figure 3: Number of stops for sim
TD
Local Traffic Adapted Signal Control
9. Summary and Outlook
As shown, the combination of a simulation environment (traffic simulation based on driver
behavior in a driving simulator) and a real field operational test delivers valuable results for
the evaluation of the systems developed in the sim
TD
project. The significantly positive effects
with regard to traffic efficiency and traffic safety can be seen according to the results in
section 8.
The FOT itself mainly serves for proving the technical feasibility and user acceptance. For
evaluating the traffic-related effects it raises some challenges, including small sample sizes
and non-controllable constraints which make it difficult to derive significant results as opposed
to only tendencies. However, by using such data as a supplement, it can support the validity
of the simulation results.
Further research that should obviously be based on the sim
TD
data may address the
examination of additional applications, the effect of additional equipment rates as well as the
traffic-related effects of combining the applications in traffic simulation.
Acknowledgement
This work was funded within the project sim
TD
by the German Federal Ministries of Economics
and Technology as well as the German Federal Ministry of Education and Research and
supported by the Federal Ministry of Transport, Building and Urban Development, the State of
Hesse and the German Automobile Industry Association.
References
[1]
M. Baur, F. Schimandl, S. Gabloner, M. Margreiter and S.
Hoffmann, "An integrated
approach for the traffic related evaluation of cooperative systems in FOTs and traffic
simulations," 19th ITS World Congress, Vienna, October 11-16, 2012.
[2]
S. Assenmacher and sim
TD
, "Field Operational Test for determining eff
ectiveness of
cooperative systems," 16th ITS World Congress, Stockholm, 2009.
[3]
sim
TD
, "D43.1 - Bericht über die Ergebnisse von Simulation und Feldversuch (German)",
2013.
[4]
sim
TD
, "Deliverable D41.1 Versuchsplan 1.0 (Version 3.0) (German)", 2010.
[5]
R. Braun, M. Baur, J. Müller, F. Schimandl, S. Hoffmann, M. Fullerton and F. Busch,
"Strategy to determine the effects of cooperative systems on traffic safety and efficiency,"
8th ITS European Congress, Lyon, June 6-9, 2011.
[6]
M. Killat, F. Schmidt-
Eisenlohr, H. Hartenstein, C. Rössel, P. Vortisch, S. Assenmacher
and F. Busch, "Enabling efficient and accurate large-
scale simulations of VANETS for
vehicular traffic management," In Proceedings of the fourth ACM International Workshop
on Vehicular Ad Hoc Networks (VANET), pp. 29-
38, Montreal, Quebec, Canada,
September 2007.
[7]
sim
TD
, "W41.2 - Auswertungskonzept (German)", 2012.
