WSEAS TRANSACTIONS on SYSTEMS and CONTROL

DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

The General Principles of the Transportation Simulation Model

Development and Validation
NADEZDA ZENINA
YURI MERKURYEV
ANDREJS ROMANOVS
Department of Modelling and Simulation
Riga Technical University
1 Kalku Street, Riga
LATVIA
nadezda.zenina@rtu.lv

Abstract: - Transportation simulation model development allows simulating traveller’s decisions, evaluating
various transportation management strategies and complex solutions. The aim of the paper is to set the general
principles of the transportation simulation model development and validation. The paper contains the overview
of the transportation simulation models types with the examples from the conducted projects for the Riga city.
The basic steps of the simulation model development procedure: initial data preparation and analysis,
transportation model development and simulation, scenarios planning and evaluation, and simulation models
outcomes evaluation are considered. Simulation model verification, validation and calibration definitions are
given. The basic checks for the transportation macroscopic and microscopic simulation model validation are
listed. A summary of the transportation simulation model validation and calibration methods and parameters is
given.

Key-Words: - Simulation, transportation, development, validation, verification, microscopic, macroscopic

1 Introduction and trends in road use, the impact of congestions on

Nowadays, simulation has become one of the most the street network is essential. The paper contains
common research methods in the transportation the description of the main transportation simulation
planning and analysis to provide the sustainable models types – macroscopic, mesoscopic,
transportation systems, increase the accessibility of microscopic and hybrid. Development of the
the transportation infrastructure, improve the transportation simulation model is considered step
availability of means of transport, increase the by step, following the simulation model
mobility for the non-motorized transportation development procedure. The figures in the paper
system participants, reduce congestions and driving represent the examples of the various transportation
time etc. Transportation simulation is based on the models for projects conducted in the Riga city.
mathematical models that imitate drivers’ decisions Simulation model validation usually done during
and evaluate complex transportation management simulation model development in order to make
solutions and tasks [2]. Transportation simulation sure that in the particular stage of the model
models allow: development the simulation results qualify the initial
 Simulating systems, that very difficult or defined requirements [3, 4]. Simulation model
impossible to evaluate analytically. verification, validation and calibration definitions
 Checking new solution before are listed to represent that there is not one unified
implementation in the real environment. It is definition of these terms. And to highlight the
safer, cheaper and faster solution. importance of the fact that developers should agree
on the checking terms definition before the
 Determining “bottlenecks” in the
transportation simulation model development
transportation network.
process. Checking process have common and differ
 It helps to understand how a complex
checks for the macroscopic and microscopic
transportation system works.
simulation models, because of the different model
In order to measure the performance of
scale and level of details. A summarization of the
transportation system, a good understanding of the
mains checks for transportation macroscopic and
traffic flow dynamic at microscopic and
microscopic simulation model validation is
macroscopic level, drivers and traveler behaviors

E-ISSN: 2224-2856 81 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

presented. The examples of validation are presented period of time, while the definite vehicles behavior
for the macroscopic simulation model with GEH patterns (e.g., car following) are assigned to the
statistics and the microscopic simulation model with driver that crosses the road network. There are
root mean square error. several transportation simulation tools for the
microscopic model development, e.g. Syncro,
Aimsun, Transyt, Vissim, Paramix, Corsim etc.
2 Transportation simulation models Aimsun tool provide a transportation network
Simulation models belong to macroscopic (Fig.1) simulation in detail, divides vehicles and drivers
microscopic (Fig.2), mesoscopic (Fig.3) and hybrid into a categories, uses different road network
models. Transportation macroscopic simulation geometries, allows simulating road traffic accidents
models simulate the whole vehicles’ behavior, and conflicting turnings for the microscopic
focusing on global factors simulation (e.g., land use, simulation models.
socioeconomic, demographic etc.). To describe the
transportation intensities the following parameters
are used: flow, density and average speed, etc.
Macroscopic transportation simulation models are
classified into static and dynamic.
Static macroscopic simulation model is not time-
depending and does not include the congestion
impact on the transportation assignment. In turn,
dynamic macroscopic simulation model takes into
account congestion impact on the transportation
intensities distribution (for example, based on the
fundamental diagram). Fig.2. The example of the microscopic simulation
There are several commercial transportation model [38].
simulation tools for macroscopic model
development, e.g. Aimsun, Vissum, Emme, Cube Transportation mesoscopic simulation models
etc. combine the characteristics of the microscopic and
The Example of the macroscopic static simulation macroscopic simulation models. Mesoscopic
model assignment results of the central part of the simulation models describe the simplified view of
Riga city is shown Fig 1. Transportation assignment the traffic flow dynamic; individual vehicles are
was made with commercial transportation integrated simulated, but to determine vehicles behaviour,
simulation tool Aimsun (Advanced Interactive macroscopic simulation model relationships are
Microscopic for Urban and Non-urban networks). used (Fig.3).
Aimsun tool provide discrete, continuous and
combined simulation; the macroscopic functionality
include the four-step travel demand modeling [9,
30].

Fig.3. The example of the mesoscopic simulation

model.
Fig.1. The example of the macroscopic simulation
Mesoscopic transportation simulation models can
model assignment results of the central part of the
dramatically differ from one another based on the
Riga city [36].
selected transportation simulation methodology.
Some mesoscopic simulation models are closer to
Transport microscopic simulation model imitates
the behavior of a single vehicle during the stated

E-ISSN: 2224-2856 82 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

macroscopic simulation models, others to Fig.4. The basic steps of the simulation procedure
microscopic model. for the transportation model development.
Transportation hybrid simulation models are the
combination of the macroscopic and microscopic
simulation model, and one or another simulation 3.1 Initial data preparation and analysis
model condition can be used depending on the Initial data preparation and analysis is the first step
transportation system objectives and tasks. For of the simulation procedure for the transportation
example, the microscopic simulation model features model development. This step includes initial data
can be used to simulate the vehicles behaviour at the collection and preprocessing (geometry data,
single intersections, and macroscopic features can intensity data etc.), identification of the possible
be used to simulate the vehicle’s behaviour at street solutions and restrictions for the transportation
sections between the intersections [8]. system, such as analysis time, size of the research
zone, number of the analyzed intersections etc.
In this step assessment methods are selected,
3 Development of the transport analysis methods are determined, and decision is
simulation model based on simulation made about the simulation model development. The
type of the developed simulation model is
procedure determined – macroscopic, microscopic or
The development of the transportation simulation mesoscopic. In most cases the type of the simulation
model generally is done based on the basic steps of model will depend on the tasks. For example, if it
the simulation procedure. The microscopic necessary to analyse the possible locations of the
simulation model is developed based on the Park and Ride parking around and inside the Riga
combined, discrete and continuous simulation city, evaluate the impact of the parking on the
model. The simulation procedure basic steps include transportation infrastructure, estimate the changes of
the problem formulation and tasks definition, the transportation intensities over a long period of
conceptual model development, input data time, take into account Riga city development plans
preparation, simulation model drawing and with new streets, bridges and others large-scale
programming, verification, validation, planning and infrastructure objects, the macroscopic simulation
design of experiments, and results evaluation and model can be the acceptable choice. The example of
observation (Fig.4). the macroscopic simulation model transportation
assignment results for Park&Ride project is shown
in Fig.5. Assignment results show the streets
capacity in the central part of the Riga city in 2025
year with Park&Ride parking.

Period: 08:00 - 09:00

Street capacity, %

Fig.5. Transportation assignment results for

Park&Ride project, 2025 year [36].

E-ISSN: 2224-2856 83 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

In case if necessary to determine the number of Fig.7. Blue color represents the transport lanes and
generated trips by the some small object, to arrows show the intensities at the intersections.
distribute the new traffic and assign it to the
transportation network; to evaluate and recommend
traffic organization changes (lanes, signal timing
etc.) at accesses, intersections and sections in the
definite research area, then the acceptable choice is
microscopic simulation model. The examples of the
microscopic simulation model animation for the
transport impact analysis project in the Riga city is
shown in Fig.6. The aim of the project was to make
changes in the research area to reduce the new
Fig.7. Working mode for geometry data input in the
object impact on the transportation network and
Synchro transportation simulation tool [39].
improve the routes capacity.
Another example, the working mode for the
intersection geometry drawing in Aimsun tool is
shown in Fig.8.

Fig.6. The example of the microscopic simulation

model animation for the transport impact analysis
project [37]. Fig.8. Working mode for geometry data input in the
Aimsun transportation simulation tool [38].
If the simulation would be part of an analysis, then
conceptual model is developed. The conceptual In most cases, the simulation tool selection is
model should be simple, clear, and understandable limited to the tool that already is available to the
and follows SMART (Specific, Measurable, developer. The tool selection also can depend on the
Attainable, Realistic, Timely) principle [32]. tasks that should be done. For example, if task is to
do the intersection signal timing optimization, the
acceptable choice is the Synchro or Transyt
3.2 Transportation model development and transportation simulation tool. If the task is to do
simulation transportation intensities assignment for the whole
Simulation model development includes the city, the modeling some small areas, then acceptable
transportation network drawing, programming, choice could be Aimsun or Vissim/Vissum
signal planning, transportation demand data (cars, transportation simulation tool.
trucks pedestrian, cyclists’ intensities) input into the After the geometry of the transportation network
simulation model, model verification, validation and and other input data are added, programming is
calibration [14]. performed, the next step is to do simulation model
Transportation simulation drawing including lanes, checking (chapter 4 for details).
reserved lanes, turnings, nodes etc. depends on the
selected simulation tool. Each transportation 3.3 Scenarios planning and evaluation
simulation tool has its own characteristics. It is The next step of the simulation procedure is to
important to draw the intersections and geometry of design and plan scenarios, and estimate and
the street network to provide that the developed evaluates the transportation simulation model
transportation system is as close as possible to the outcomes.
observed ones. In this step is highly recommended to pay attention
The example of the working mode for microscopic to the number of the scenarios. Too many scenarios
simulation model for the Synchro tool is shown in can lead to the situation, when only developers will

E-ISSN: 2224-2856 84 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

understand the difference between scenarios, and The estimation step of the simulation procedure
practically impossible will be clear define the includes results evaluation in the real environment,
benefits of the each developed scenario to the e.g. when intersection had been reconstructed
customer and select acceptable one. according to the proposed improvements, does the
The example of the microscopic simulation model specified number of lanes for each road section
animation with too many scenarios is shown in correspond to the observed data, any unexpected
Fig.9. It is the central part of the Liepaya city in the events was occurred etc. All these finding should be
Latvia. The aim of the project was to analyse used to evaluate results of simulation and transfer
different traffic organization changes in the whole knowledge.
research area [38]. The changes included: change
some streets directions, restrict vehicles movements
at some streets, and change some intersections 4 Simulation model verification,
traffic organization (e.g., provide left turn or right
turn pockets, prohibited some left turns), offset tram
validation and calibration definitions
The purpose of the transportation simulation model
lanes, develop signalizing at some intersections,
verification, validation and calibration is to provide
change some intersection from unsignalized to
that simulation outputs do not exceed the expected
signalized or maybe to roundabout, add public
results and an abnormal tendencies do not occurred
transport lanes etc.
(e.g., the vehicles cross the intersection at red light
signal, pedestrians cross the road with vehicles in
one time, vehicles run through one another etc.).
But, if this happens, the reasons for such behaviour
might be input data errors, expectations errors or
animation errors [33]. In such cases, input data,
parameters, geometry of the street network, and
observed driver behaviour data on the road should
be rechecked. Sometimes simulated outcomes do
not correspond to the observed ones, and error exists
in the observations, not in the model. Sometimes
errors are associated with the representation of the
objects, improperly drawn cars trajectories,
incorrect road sign placement, cars spin around
themselves (Fig.10), human errors etc. As the result
simulation model does not match the observed one
Fig.9. The example of the microscopic and cannot be used for further analysis. In order to
transportation simulation model animation in minimize such situations, the transportation
Aimsun tool [38]. simulation model needs to be thoroughly checked.

And all these changes were simulated and checked

firstly one by one and then all together. But in time
the number of scenarios increases, because all the
time changes in the requirements: what if left
changes from A to D, but E and D changes removed
from scenario. And now remove A and B changes
and add E change. In the result more than sixty
scenarios were developed, in which developer was
able to somehow navigate, but mostly it was
impossible to describe the benefits and
disadvantages to others. And it returns to the
simulation procedure basic steps – plan the number
of the scenarios and to the first step of procedure -
define the restriction.
Fig.10. Animation error example: cars spin around
themselves.
3.4 Simulation models results evaluation

E-ISSN: 2224-2856 85 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

There is no single unified validation, verification and calibration definition, see Table 1.
Table 1. Simulation model verification, validation and calibration definitions
Validation Verification Calibration Source
The process of determining whether the logic of the The process that determines how the The process in which the user [6]
simulation model is correctly represented in a computer developed transportation simulation model determines set of the input
program. Validation states that simulation results represents the actual transportation situation. parameters to represent local
correspond to the model logic. Validation does not transportation regulations.
check simulation model consistently with proposed
model logic or theory on which the logic was based.
The process of checking whether simulation model is The reviewing process of the simulation The process of the parameters [9]
representative, based on calibration data and the whole model development to make sure that model adjustment to model the real
transportation system analysis. In practice, model is working as expected. transportation system.
validation is calibrated, which means the model output
comparison with input data that were not included in
the calibration process.
The process in which the user checks the total forecast The process to avoid the problems in the The process in which user selects [5]
capabilities of the transportation network in terms of calibration process. Simulation model parameters of the simulation model
the traffic flows, travel times, average speeds and verification is not considered as a separate that realistically reflect the
average delays. step in the validation process. researched transportation system.
The testing process that determines the simulation Verification is considered as part of The process of the adjusting the [3]
model viability and usefulness in comparison with the validation. parameters values and others user-
transportation model under the study. It also requires controlled variables to match the
model calibration, model parameters adjustment until modelled transportation flow
the model output will be close to expected. intensities with observed.
In simulation model validation is necessary to check
that input data and specific parameters are correctly
Validation and calibration steps are used in any added into transportation simulation model, and no
transportation simulation model checking process. errors in the numerical parameters or basic settings.
Simulation model verification is used by the Validation process have common and differ checks
individual authors and in the validation guidelines for the macroscopic and microscopic simulation
as the third additional check for the validation and models, because of the different models scale and
calibration. In other cases, validation includes the level of details.
validation and calibration steps and simulation
model verification occurs during the validation
process. Sometimes verification is used as a 4.1 Transportation macroscopic simulation
synonym for validation, and in some cases is called model validation
evaluation and testing [3]. In most cases macroscopic transportation simulation
It is significant to ensure that simulation model models are developed basing on the four-step
development team agrees on definitions terms and transport travel demand model (trip generation, trip
meanings before the simulation model checking distribution, and transportation mode choice and trip
process. assignment). The first step of macroscopic
The author the simulation model validation defines simulation model validation is to ensure that the
as process in which it is checked, how well each step of the four-step transport travel demand
transportation simulation model fits the observed model is properly validated. [35]. Next step is to
transportation system by comparing simulated and check simulated model traffic data (intensities on
observed transportation system logic and structure, the key intersections, flow structure etc.) and
by adjusted calibration parameters in the way that infrastructure data (number of lanes, restricted turns,
they are consistent with observed transportation reserved lanes for public transport, timing,
system, and by checking that simulation model calculated section capacity etc.). The following
effectiveness measures adequately describe the checks should be carried out:
observed transportation situation. Simulation model  Load at connectors. Simulation model can
validation includes validation and calibration steps have multiple connections points
where simulation model verification occurs during (connectors) that are connected to
the validation process. transportation network. Check that capacity
distribution at these connectors match the
observed ones. Check that connectors are not
4 Simulation model validation connected to the intersections, but are
connected to streets with the low category.

E-ISSN: 2224-2856 86 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

 Street and route categorization. Check that GEH < 5 5 < GEH < 10 GEH > 10
street and route categories correspond to the
26% 56% 18%
normative acts and documents accepted for
the research area.
 Section capacity. Check the entered section The GEH value is between 5 and 10 score for the
capacity, each street and route category is several road sections, transportation simulation
assigned to own volume delay function. model validation is unacceptable and necessary to
 Transport intensity. Check that transportation carry out additional simulation model validation and
intensities on the individuals sections and check the input data, load at connectors, street
turns correspond to observe ones. categories, added street capacity, allowed speeds
To determine the correctness of data checks and transport intensity. The error cause should be
recommended conducting the sensitivity tests. finding in the parameters of the transportation
Transportation assignment should be done by simulation model or in input data.
decreasing free flow speed by 10% then by 20%;
Period: 08:00 - 09:00
decreasing public transport usage by some percent;
increasing traffic flow intensity by some percent etc.
All these actions help to determine the atypical and
odd patterns in the transportation network and
improve the simulation model correctness.
Measured intensities values at intersections and
travel time between main origin and destination
points can be used to do the macroscopic
transportation simulation models calibration.
Calibration of the simulation model transportation
intensities made based on the GEH statistics (1) by
comparing observed and simulated transportation
intensities.
Traffic flow intensity per hour

2(𝑠−𝑜)2
𝐺𝐸𝐻 = √ 𝑠+𝑜
(1)
Fig.14. The example of the EMME macroscopic
where s – observed transportation intensity.
simulation model for the Riga city [36].
o − simulated transportation intensity.

Validation with GEH statistics is considered

acceptable, if values of the GEH are between 0 and 4.2 Transportation microscopic simulation
5 (good score). If values of the GEH are greater than model validation
10, validation is unacceptable; the observed and The following checks should be done for the
simulated transportation intensities differ transportation microscopic simulation model:
significantly. The values of the GEH in the range of  Input data. Check that specified parameters
5 to 10 require an additional in-depth analysis [34]. e.g., intensities, number of lanes, signs,
According to the [34] at least 85% of the road permitted speed, traffic lights, intensities
sections must be GEH value less than 5 score and at match the observed.
least 90% of the road sections must be GEH value in  Basic settings. Check the basic parameters of
the range of 5 to 10 score. the simulation tool. Number and type of
The example of the GEH statistics results for the parameters will depend on the selected
macroscopic simulation model for the Riga city simulation tools.
(EMME model, morning scenario, maximum hour)  Geometry data. Check that number of lanes,
is shown in the Table 2. The view of macroscopic lane widths, allowed directions and speeds,
simulation model for the Riga city is shown in the turns corresponds to the observed.
Fig. 14 (colours represent the transportation  Driver behaviour parameters. Check the lane
intensities; dark red colour - high intensities). changing and queue discharge parameters,
overtaking and car-following models etc.
Table 2. Example of the GEH statistics for the  Transportation flow intensities (cars, trucks,
macroscopic simulation model pedestrian, cyclists, public transport, etc.).
Check intensities for typing errors, traffic

E-ISSN: 2224-2856 87 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

lights and cycle length input errors, vehicles according to observed situation. The result of the
interfere with each other's in the initial simulation model validation for the developed
transportation network. intersection is shown in Fig. 16.
 Road user cooperation. Check public There are several methods to determine the
transport priority on transit lanes, correctness of data checks. Some of them include
cooperation between cars and pedestrian at the simulation model run with the moderate load or
intersections. the maximum load, and review of how the system
The simple simulation model for the single behaves (e.g., vehicles behave atypically, major
intersection in the Riga city, in the Pulkveza Brieza congestions or queues do not detect) and how many
street area was developed to demonstrate the initial vehicles can drive through the system under such
steps for the transportation microscopic simulation condition.
model validation (Fig. 15).

Fig.16. The results of the initial simulation model

validation for the single intersection.

Other methods include the observation of the

Fig.15. The initial developed simulation model. individual travel routes with the highest
transportation intensity, of the potential conflicting
The input data for the transportation simulation situation that may involve multiple road users [33].
model included intersection geometry (number of All these actions allow reviewing transportation
lanes, allowed turns, pedestrian crossings, stop lines, simulation model for input errors. As the result of
signs, public transport stops), signal timing data these actions, transportation simulation model is
(signal groups, signal phases, control type, cycle ready for further model validation and calibration.
length) and transportation intensity data (cars,
trucks, pedestrians, cyclists; schedule and frequency
data were added for public transport). 4.3 Validation methods
The initial step of the simulation model validation is Validation methods divide into quality methods and
to check correctness of the entered data and quantitative methods. Quality validation methods
parameters. For the developed simulation model include the graphical results comparison (e.g.,
(Fig. 15) the checking was done by answering the animation, graphs and histograms) with others
following questions: simulation model outputs, and model behaviour
 Intersection geometry. Whether the specified research.
number of lanes, turns, stop lines and signs The example of the quality validation method usage
corresponded to the observed? (average approach delay for the evening maximum
 Signal timing. Whether the specified signal hour) for developed single intersection is shown in
controls corresponded to the observed? the Fig. 17.
 Transportation intensity. Whether the Quantitative validation methods include the model
specified intensities for the cars, trucks, outputs analysis with statistical methods, e.g.,
pedestrians corresponded to the observed? percent error, root mean square error (2), mean
If there were some “No” answers, then percentage error, GEH statistics or Theil’s
transportation simulation model was changed uncertainly coefficient [3, 10, 11, 12, 22].

E-ISSN: 2224-2856 88 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

By changing the trajectories, removing the stop lines

and pointing out the road signs, the transportation
intensities rate doubled for all lanes. And by
changing at least one parameter in the simulation
model can improve transportation flow and reduce
congestion. It is recommended to analyse the each
road section individually and change the parameters
based on its characteristics.

4.3 Calibration parameters

Fig.17. The average approach delay for the The selection of the validation and calibration
developed single intersection. parameters and transportation system performance
indicators depend on the task complexity, initial
data quantity and quality. Several parameters are
1 n
used to evaluate the performance of transportation
RMSEi   (wij  vij ) 2
n j 1
, (2) system model and to obtain the acceptable and
reliable results. Standard model parameters are
where n − number of records; iteratively adjusted until a satisfactory level between
wij − modelled transport flow modelled and observed data is achieved. It is tedious
intensity; and time-consuming process when it’s done
vij − observed transport flow. manually, even if some engineering techniques are
used to reduce the number of attempts. Automated
The example of the root mean square error simulation model calibration is based on the
calculation for developed single intersection in the systematic approach. This approach views the
Riga city (Fig. 15, Fig. 16) is shown in Table 3. model calibration as an optimization problem that
Three groups with varied changes in the traffic looks for the parameters value combinations that
organization were developed during the initial best describe the target function, e.g., performs
validation stage of the transportation simulation microscopic simulation model calibration with the
model. The first group was without any change in genetic algorithms to calibrate the global and local
the traffic organization. The second group provided model capacity parameters. The function objective
changes for signs and connections trajectories. And is formulated as a black box model and solution is
the third group provided changes for the stop lines. searched using heuristics. The computation
Root mean square error was calculated for the four complexity is exponential. Optimization procedures
access road sections for the each group to observe require a large number of attempts, which in turns
any changes in the road sections and impact on the are time-consuming process and depends on the
analysed transportation system. transportation network and modelled transportation
conditions. Thus, a subset of parameters with
Table 3. Results of the RMSE calculation engineering solution is selected, or sensitivity or
RSME values dispersion analysis is used to reduce the complexity
Changes in the
transportation 1. 2. 3. 4. of the optimization [13]. The wide range of the
system sectio sectio sectio sectio optimization algorithms exists, started from genetic
n n n n algorithms to an enhanced simultaneous
No changes (Fig.
perturbation stochastic approximation (SPSA) [15].
160 611 288 42 For example, the author [16] uses the simultaneous
15)
perturbation stochastic approximation to calibrate
Signs and trajectory the simulation model parameters and demand
203 356 280 30
changes templates by using different sections. The authors
Stop lines changes advise to use an enhanced SPSA algorithm with the
with signs and 203 356 280 36 Bayes sampling technique. Several authors [17] use
trajectory (Fig. 16) SPSA to calibrate supply and demand parameters.
The driver behaviour parameters are pre-calibrated
RMSE calculation results showed that the relevant based on lane sensor data. For example, the author
road signs and connection trajectories have the most [18] used a simplex algorithm to perform the
significant impact on the transportation intensities. calibration of one parameter subset - the

E-ISSN: 2224-2856 89 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

acceleration, deceleration, and lane-changing acceptable neural network configuration that will
behaviour of followers. Others authors [19] model provide acceptable answers. The study [23]
calibration based on the OptQuest/Multistart proposed to use the Memetic algorithm (MA) for
algorithm (OQMS), the author from a literature microscopic simulation model calibration.
source [20] proposed hybrid calibration - calibration A summary of the simulation model validation and
of microscopic models with macroscopic data. The calibration methods and parameters is given in
variability of the drivers in the speed-flow ratio has Table 4. Data mining methods are actively used in
to be taken into account when performing the transport management tasks, e.g., to simulate and
simulation model calibration. optimize intelligent traffic signalization plans [24],
Authors from the literature source [21] used neural to analyse the traffic accidents [25] and to pre-
network to perform the model calibration for the process the observed data to analyse the driver
roundabouts. The main unanswered questions are behaviour on the roads [26], to identify the black
the optimal size of the simulation model database to zones on the road [27], and to define the future
train the neural network, and a way to find the transportation demand [28].

Table 4. The Summary of the simulation model validation parameters and methods
Calibration Simulation
Validation parameters Validation methods Calibration parameters
methods tool
Mean square
Trial and error
Transportation flow intensity percentage error, Global, local parameters Aimsun
method
GEH statistics
Minimum headway, speed,
Average driving speed Not described transportation flow saturation Genetic algorithm Vissim
rate
Functional indicators Desired speed, safety distance,
Average driving time, maximum
comparison at average headway, Neural network Vissim
queue, vehicles stops
intersection inputs minimal/maximal look ahead
Transportation flow intensity Average drivers reaction time,
Mean absolute error SPSA algorithm Paramics
and density average arrival interval
Transportation counts such as
Average drivers reaction time,
flow rate, speed, queue length Mean absolute error Genetic algorithm Paramics
average arrival interval
etc.
Speed, transportation flow Average drivers reaction time,
Not described Not described Paramics
intensity average arrival interval
Transportation flow intensity, Average gap, average drivers
Not described Genetic algorithm Paramics
occupancy reaction time
Car-following factors, drivers’
Transportation flow intensity Mean absolute error Simplex algorithm Corsim
aggressiveness factors
More than 20 parameters, e.g.,
GEH statistics, Mean lane change frequency, Memetic
Transportation flow intensity Corsim
error maximum turning speed, algorithms
pedestrian delay, headway etc.
Transportation counts such as Mean square error,
traffic flow intensity, speed, mean square Drivers behaviour parameters Genetic algorithm MITSIMLab
queue length etc. percentage error
Transportation flow intensity Multistep
Not described Drivers behaviour parameters MITSIMLab
and speed sensitivity analysis

As shown by the study of comparing genetic and between the observed and simulated data over a
clustering algorithms [29], both genetic and given period of time on a typical day, or using
clustering algorithms provide practical solutions for separate transport system performance indicators or
the parameters values calibration, however, genetic the demand templates is already pre-calibrated.
algorithms require more time to find a solution.
The summary of the validation methods and
parameters suggests that the authors perform model 5 Conclusion
validation and calibration based on a subset of In the paper the general principles of the
model parameters, examining the difference transportation simulation model development and

E-ISSN: 2224-2856 90 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

validation were considered. The overview of the [4] Transportation Impact Analyses for Site
transportation simulation models type is given. The Development: An ITE Proposed Recommended
steps of the transportation simulation model Practice, translation: SIA Solvers and SIA
development procedure are specified. The Open World, 2007.
considered examples of simulation models are taken [5] S. Detering, L. Schnieder, E. Schniede, Two-
from the traffic organization and impact projects level validation and data acquisition for
conducted in the Riga city. microscopic traffic simulation models,
The geometric representation of transportation Advances in Systems and Measurements, vol. 3,
network is the only the first step in the simulation 2010, pp. 48-56.
model developing process. The next important step [6] J. Han, M. Kamber, J. Pei, Data Mining:
is to ensure that developed model corresponds to Concepts and Techniques, 3rd ed., USA:
observe one, made the transportation simulation Elsevier Inc., 2012, pp. 678.
checking. The varied checks are presented for the [7] D. H. Lee, X. Yang, P. Chandrasekar,
transportation macroscopic and microscopic Parameter calibration for Paramics using
simulation model validation. There is no unified genetic algorithm, Washington: TRB, 2001.
definition of the transportation simulation model [8] Department of Transportation Federal Highway
validation, verification and calibration. Validation Administration, https://www.fhwa.dot.gov
and calibration steps are used in any transportation [9] D. Ni, J. D. Leonard II, A. Guin, B. M.
simulation model checking process. And Williams, Systematic approach for validating
verification is used as the third additional check; traffic simulation models, Transportation
sometimes it is included into the validation step. Research Board, vol. 1876, 2004, pp. 20-32.
The summary of the parameters and methods for the [10] M. Maher, Inferences on trip matrices from
transportation simulation model validation and observations on link volumes: a Bayesian
calibration is given. The parameters and methods statistical approach, Transportation Research
selection in most cases depend on the simulation Part B: Methodological, vol. 17 (6), 1983, pp.
tool. Transportation flow intensities are the most 435-447.
common used parameter for transportation [11] Nelson \ Nygaard, Adjusting Site-Level
simulation model validation based on the summary Vehicle Trip Generation Using Urbemis, San
(Table 2). And drivers’ behaviors parameters are Francisko: Nelson\Nygaard Consulting
the most used for transportation simulation model Associates, 2005.
calibration. The topic for the further research is to [12] G. G. Schultz, L. R. Rilett, Analysis of
evaluate the transport travel demand microscopic distribution and calibration of car-following
simulation model calibration procedures for sensitivity parameters in microscopic traffic
transport impact analysis. The described principles simulation models, Transportation Research
of the transportation simulation model development Board, vol. 1876, 2004, pp. 41-51.
and validation can be used to develop the acceptable [13] D.Chiopu, Applying two-step cluster qnalysis
and stable transportation simulation model. for identifying bank customers' profile,
Buletinul Universit Petrol, Gaze din Ploie?ti,
vol. LXII, 2010, pp. 66-75.
References: [14] N. Zeņina, J. Merkurjevs, The Basis for the
[1] B. R. Hellinga, Requirements for the Transportation Microscopic Simulation Model
calibration of traffic simulation models, Validation. 60th International Scientific
Canadian Society for Civil Engineering, vol. Conference on Information Technology and
IVB, 1998, pp. 211-222. Management Science of Riga Technical
[2] T. J. Fratar, A. M. Voorhees, M. S. Raff, University (ITMS 2019), Proceedings, Latvia,
Forecasting distribution of interzonal vehicular Riga, 10-11 October, 2019.
trips by successive approximations/with [15] M. Ben-Akiva, L. Lu, Y. Xu, A. Constantinos,
discussion, Highway Research Board, An enhanced SPSA algorithm for the
Washington: Transportation Research board, calibration of dynamic traffic assignment
2008, pp. 53-65. models, Transportation Research Part C, vol.
[3] J. Bowman, M. Ben-Akiva, Activity-based 51, 2015, pp. 149-166.
disaggregates travel demand model system with [16] I. I. Otkovi, T. Tollazzi, M. Raml, Calibration
activity schedules, Transportation Research A, of microsimulation traffic model using neural
2000, pp. 1-28. network, Expert Systems with Applications, vol.
40, 2013, pp. 5965-5974.

E-ISSN: 2224-2856 91 Volume 15, 2020

 WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2020.15.10 Nadezda Zenina, Yuri Merkuryev, Andrejs Romanovs

[17] R. Balakrishna, C. Antoniou, M. Ben-Akiva, H. [30] International provider of software and services
N. Koutsopoulos, Y. Wen, Calibration of for traffic planning, simulation and prediction,
microscopic traffic simulation models: method https://www.aimsun.com
and application, Transportation Research [31] Aimsun next modelling guidelines, Nevada
Board, vol. 1999, 2007, 198-207. department of Transportation, June 2018.
[18] A. Khalik, M. A.-T. Amal, Prediction analysis [32] SMART Goals: A How to Guide, University of
of trip production using cross-classification California, 2017.
technique, Al-Rafidain Engineering, vol. 14 (4), [33] Методические рекомендации по разработки
2006, pp. 51-63. и реализации мероприятий по организации
[19] A. Chen, P. Chootinan, W. W. Recker, дорожного движения (in russian), Moscow,
Examining the quality of synthetic origin- 2017.
destination trip table estimated by path flow [34] UK Highways, Design Manual for Roads and
estimator, Transportation engineering, 2005, Bridges, UK Highways Agency. 1992.
pp. 506-513. [35] J. Schlaich, U. Heidl, P.Mohl, Multimodal
[20] UK Highways, Design Manual for Roads and macroscopic transport modelling: State of the
Bridges, UK Highways Agency, 1992. Art with a focus on validation & approval, 17th
[21] J. D. Ortuzar, L. G. Willumsen, Modelling IRF World Meeting & Exhibition, 2013.
transport, 4th ed., John Wiley & Sons ltd., [36] E.Daniševska birojs and Solvers Thinking
2011, p. 586. Ahead, Rīgas domes autonovietņu politikas un
[22] N. Zenina, A. Romanovs, J. Merkuryev, attīstības koncepcijas izstrāde Stāvparku
TRIP-based transport travel demand model for sistēmas sadaļas attīstības plans (In Latvian),
intelligent transport system measure evaluation 2013.
based on micro simulation. International Journal [37] E.Daniševska birojs and Solvers Thinking
of Simulation and Process Modelling, 2017, Ahead, Transporta plūsmu izpētes un
Volume 12, Issue 3-4, pp.207-220. prognozes projekts ēku kompleksam teritorijai
[23] D. Pavlyuk, V. Gromule, Discrete choice starp Rūpniecības, Pētersalas un Katrīnas ielām
model for a prefered transportation mode, Proc. (In Latvian), 2008.
of the 10th International Conference Reliability [38] E.Daniševska birojs and Solvers Thinking
and statistics in transportation and Ahead, Satiksmes plūsmu detalizēta izpēte
communication, Riga: Transport and Liepājas pilsētas centra teritorijai starp Krišjāņa
Telecommunication Institute, 2010, pp. 143- Valdemāra ielu, Jūrmalas ielu, Siena ielu,
152. Kuršu laukumu, Malkas ielu, Kungu ielu,
[24] J. G. Wardrop, Some theoretical aspects of road Bāriņu ielu, Kārļa Zāles laukumu un Jūras ielu.
traffic research, Institute of Civil Engineers, (In Latvian), 2009.
vol. 1 (3), 1952, pp. 352-362. [39] E.Daniševska birojs and Solvers Thinking
[25] V. C. Tatineni, M. J. Demetsky, Supply Chain Ahead, Satiksmes plūsmu izpētes un
Models for Freight Transportation Planning, modelēšanas projekts Brīvības gatvei no
University of Virginia, 2005. Gustava Zemgales gatves līdz Krustabaznīcas
[26] G. List, L Konieczny., C. Durnford, A best ielai (In Latvian), 2011.
practice truck flow estimation model for the [40] H. Largo, P. J. G. Ribeiro, G. Dias, C. P. B.
new, Transportation Research Record, Trujillo, Public Transportation Demand Model
Washington: Transportation research board, for Low Density Territories, WSEAS
2002, pp. 97-103. Transactions on Environment and
[27] J. Rzeszotko, S. H. Nguyen, Machine learning Development, 2019, vol. 15, pp. 395-407.
for traffic prediction, Pu?tusk: International [41] A. Carteni, A Cost-Benefit Analysis Based on
Workshop on Concurrency, Specification and the Carbon Footprint Derived from plug-in
Programming, 2011, pp. 442-453. Hybrid Electric Buses for Urban Public
[28] E., Ruiter M. Ben-Akiva, Disaggregate travel Transport Services, WSEAS Transactions on
demand models for the San Francisco bay area, Environment and Development, vol. 14, 2018,
Transportation Research Record, 1978, pp. pp. 125-135.
121-128. [42] M. Botte, B. Montella, L. Dacierno, A
[29] K. Daehyon, Neural networks for trip Geometrical Approach to Reduce Calculation
generation model, Eastern Asia Society for Times in the Definition of Rail Convoy Speed
Transportation studies, Vol. 2 (4), 2001, pp. Profiles, WSEAS Transactions on Environment
201-207. and Development, vol. 16, 2020, pp. 98-104.

E-ISSN: 2224-2856 92 Volume 15, 2020