Agent-based Transportation Demand Management
Demand Effects of Reserved Parking Space and Priority Lanes in Comparison and
Combination
Markus C. Beutel
1
, Sebastian Addicks
2
, Barbara S. Zaunbrecher
2
, Simon Himmel
3
,
Karl-Heinz Krempels
1
and Martina Ziefle
3
1
RWTH Aachen University, Information Systems, Ahornstr. 55, 52074 Aachen, Germany
2
RWTH Aachen University, 52074 Aachen, Germany
3
RWTH Aachen University, HCIC, Communication Science, Campus-Boulevard 57, 52074 Aachen, Germany
Keywords:
Transportation Demand Management, Agent based Simulation, Transportation Demand Simulation, Multi-
method Approach.
Abstract:
Fostering the usage of alternative mobility modes, e.g., carsharing or carpooling becomes more and more
urgent in modern urban planning. Politicians and city planners have already recognized that putting targeted
incentives can influence people’s mobility behavior in an effective way. Agent-based simulations of transporta-
tion demand can be a valuable tool to support these planning processes. This work is based on a state-of-the-art
transportation demand simulation and shows modeling and simulation modifications related with agents under
the influence of incentives. These agents have been assessed in qualitative and quantitative studies prior to the
simulation. Results show that agent-based simulation of transportation demand is suitable to evaluate impacts
of transportation demand management measures. More specifically, all investigated measures show certain
impacts on mobility mode choice, at which an incentive combination is most effective.
1 INTRODUCTION
Traffic congestion, air pollution, and rising economic
costs are major problems that go hand in hand with
inefficiencies in the public transport sector. People
can already decide between various mobility options
that hold the potentials of efficient capacity utiliza-
tion. Unfortunately, their decisions are not always the
best regarding the overall (urban) mobility system, the
ecosystem, or public welfare. A possible solution for
this problem is transportation demand management
(TDM), which is defined as the art of modifying travel
behavior, usually to avoid more costly expansions of
the transportation system (Ferguson, 1990). Research
has investigated a large variety of TDM measures and
their effects for a long time, because of their crucial
importance for city planning. Agent-based demand
simulation can support these planning processes by
predicting the impact of different measures. This
work primarily investigates the hypothesis that agent-
based simulation (ABS) is a suitable tool to simulate
the effects of TDM measures. To do so, we collect
required data and modify an existing transportation
demand simulation to include TDM measures.
In this work, we followed a three step procedure.
Because mobility needs and demands, choice of mo-
bility mode and the willingness to change the pre-
ferred mode of transport are socio-psychological is-
sues, a qualitative study (Rennekamp and Nall, 2006)
was conducted in a first step. In an exploratory at-
tempt the current mobility usage, on the one hand,
and possible incentives to a more environmentally
friendly behavioral change, on the other hand, were
discussed within a group of students and research as-
sistants. In the second step, these qualitative results
were validated in an online questionnaire study to
gain quantitative data for the agent-based simulation,
led by computer scientists, as the third and main step
discussed in this paper.
The work is structured as follows. Section 2 gives
an overview of demand side management measures
stated in the literature. In addition, the field of agent
based transportation demand simulation is described.
Section 3 explains the related data basis, before Sec-
tion 4 dives more deeply into the used simulation ap-
proach and presents the results. Finally, Section 5 re-
flects the approach and discusses contributions from
interdisciplinary perspectives.
317
Beutel M., Addicks S., Zaunbrecher B., Himmel S., Krempels K. and Ziefle M..
Agent-based Transportation Demand Management - Demand Effects of Reserved Parking Space and Priority Lanes in Comparison and Combination.
DOI: 10.5220/0005411503170323
In Proceedings of the 4th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS-2015), pages 317-323
ISBN: 978-989-758-105-2
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
Table 1: Survey on TDM measures based on (Rodriguez and Murtha, 2009), (Seik, 2000), (Smith, 2008), (Hensher and
Pucket, 2007), (Schlag and Schade, 2004) and (Sonnenberger and Ruddat, 2013).
Category Measure
Information (real-time) travel information provision, public information campaigns, context-sensitive
design, TDM marketing, multi-service smartcard pilot application, information via regis-
tration offices, online information database, station signs, sharing stations on city maps,
public authorities using carsharing, celebrity advertising
Infrastructure high-occupancy vehicle (HOV) lanes, parking policy, improved mode paths, park and
ride, pedestrianized zones, avoiding major new road infrastructure, bicycle parking, clus-
tered land use, shared parking, priority parking, carsharing integrated in housing com-
plexes, flexibility for stationary carsharing, combined car- and bikesharing stations, shar-
ing on peripheral public transport stations, parallel station and free floating carsharing
Economical product bundling, congestion pricing strategies, (electronic) road pricing, fuel tax, parking
charges, tradable permits, car tax, distance-based pricing, comprehensive market reforms,
least-cost planning, congestion and variable user charging, travel card for parking and bus
services, park and ride with integrated ticketing, tariff integration by monthly pass for all
public transport modes, regional mobility card
Regulatory parking controls, closure of city centers for individual car traffic, decreasing speed lim-
its, traffic management, school transit management, special event management, tourist
transport management, mode integration, vehicle use restrictions, smart growth planning,
access managment, reducing parking space, new mobility concepts, rural carsharing net-
works, programs to encourage ridesharing
Institutional guaranteed ride home, shuttle services, campus transport, additional capacity, mode im-
provements, alternative working patterns, opening organizational car pool
2 RELATED WORK
This section is divided into two parts. The first gives
an overview of transportation demand management
measures stated in literature. The second subsection
describes the field of agent-based transportation de-
mand simulation.
2.1 Transportation Demand
Management Measures
In the literature, similar terms, e.g., Travel Demand
Management (Rodriguez and Murtha, 2009), Trans-
port Demand Management (Ison and Rye, 2008),
or Transportation Demand Management (Ferguson,
1990), describe the issue of influencing people’s mo-
bility behavior in an efficient way. Table 1 summa-
rizes prominent TDM measures stated in current liter-
ature. Basically, measures can be divided into incen-
tives, on the one hand, and constraints, e.g., variable
costs and pricing measures, on the other hand. Some
measures like parking provision or high-occupancy
vehicle (HOV) lanes are stated by most references and
have been investigated for years (Jacobs et al., 1982).
More recent measures, e.g., product bundling, com-
plement the survey and give a comprehensive impres-
sion of possible measure options.
Since the 1970s, the topics carpooling and us-
age incentives have been investigated by researchers
worldwide (Teal, 1987), (Ben-Akiva and Atherton,
1977). But 40 years later, car pooling still plays a mi-
nor role in the urban mobility mixture. In this paper,
we especially focus on the effects of incentives on the
modal split of younger adults in western societies.
2.2 Agent-based Transportation
Demand Simulation
The described work is based on agent-based simula-
tion (ABS). ABS is a microscopic approach for de-
scribing complex and dynamic systems (Macal and
North, 2013; Magg, 2012). Because of the broad field
of application, ABS is popular for investigations of
manifold research problems (Macal and North, 2009).
Hence, it is an interdisciplinary approach which of-
ten touches various research fields (Axelrod, 2006).
If ABS is used for modeling complex economic sys-
tems, e.g., consumer or business model analysis (Zut-
shi et al., 2013), it is also called agent-based compu-
tational economics (ACE) (Tesfatsion, 2006). Espe-
cially because of its basis on individual decision mak-
ing, ABS is also used in the area of transportation de-
mand simulation (Magg, 2012).
There are different microscopic demand simula-
SMARTGREENS2015-4thInternationalConferenceonSmartCitiesandGreenICTSystems
318
tion approaches, e.g., mobiTopp, TAPAS and ORI-
ENT (Magg, 2012). We build up our investigations
on the valuable developments in this area and use
the tool MATSim
1
for demand simulation. It enables
agent-based urban transportation simulation for ex-
traordinary large areas and has been developed and
improved over several years (Horni et al., 2011). Due
to its modular component structure, MATSim offers
suitable room for adjustments and extensions (Magg,
2012).
3 DATA BASIS
As a reference for the simulation, insights into users
views on possible incentives were gathered using a
focus group study and an online survey.
3.1 Qualitative Data
A focus group study is an explorative method to
gather broad and detailed data on a certain topic from
a limited number of participants. It is used in early
stages of research to identify aspects of the research
question that are relevant to users. It does not provide
statistically relevant data due to its qualitative nature
but serves as a pre-study for the quantitative follow-up
study.
Figure 1: Focus Group on Mobility Demands.
Methodology
For the focus group study, 8 participants in their mid-
twenties to mid-thirties were invited to take part in a
discussion. 6 of them were male, 2 female, and all of
them were either students or had already completed
their university degree courses. The participants had
mixed mobility profiles that ranged from passionate
1
http://www.matsim.org
motorbike riders and car-owners to participants us-
ing only their bike and car-sharing. One participant
reported not to hold a drivers licence. Participants
did not receive any gratification for their participation.
Half of the participants took part in the discussion as
a university course requirement.
First, the group was introduced to the topic of ur-
ban mobility by being asked to name all types of mo-
bility modes they knew. In a next step, advantages and
disadvantages of different means of transport were
discussed. Participants were then asked which incen-
tives would be needed for them to change their current
preferred mode of transport and how the change to
more environmentally friendly modes could be pro-
moted. The discussion was led by two experienced
moderators and lasted approximately 2 hours. Audio-
recording and note-taking by two assistants and the
moderators were used for data collection.
Results
The productive group discussion showed manifold re-
sults and creative outputs. At this point, most relevant
findings are described.
Participants predominantly showed most interest
in provided parking space and ticket service bundling
as incentives. In particular, an encompassing ticket,
covering the usage of public transport modes, com-
mercial and private carsharing was discussed in de-
tail. Participants discussed the idea of using a ticket
with unlimited subscription to public transport modes
and addionionally including a flexible component for
alternative mobility modes, e.g., carsharing or bike-
sharing. There was a broad interest in this concept
which indicates that this product bundling has the po-
tential to lower barriers towards alternative mobility
modes.
Moreover, diverse ticket alternatives were dis-
cussed. For this purpose, every participant was asked
to specify a certain mobility budget. They had prob-
lems concerning this issue which leads to the conclu-
sion that modeling of agent’s budget is going to be
problematic. Therefore, we excluded related pricing
measures from this investigation.
Moreover, other incentives, e.g., priority lanes,
were discussed. Remarkable was the fact that every
group participant had individual requirements con-
cerning different incentives. This fact will be con-
sidered during agent modeling Section 4.
3.2 Quantitative Data
The qualitative input by the focus group built the
foundation for a constitutive online survey.
Agent-basedTransportationDemandManagement-DemandEffectsofReservedParkingSpaceandPriorityLanesin
ComparisonandCombination
319
Methodology
This survey served as a quantitative data input for
the agent modeling and for the simulation validation.
Corresponding to these aims, the survey was con-
ducted to deliver insights about participants estima-
tions of selected incentives as well as related mode
choice. Fifty applicable datasets were used for sim-
ulation. Corresponding to the focus group, most par-
ticipants were students in the mid thirties. 36 percent
of them were female and 64 percent were male.
After introducing the basics needed for fundamen-
tal understanding, e.g., mobility mode definitions,
subjects had to specify which modes they are cur-
rently using. Thereupon, a distribution of not used
modes was modeled into the agent database. Next,
the incentives ”reserved parking” and ”priority lanes”
had to be evaluated concerning personal importance.
Finally, subjects had to state their mode choice with-
out and within a specific incentive scenario.
Results
The quantitative data was used for two purposes: on
the one hand, the requirements for specific incen-
tives was evaluated. Thereby, average results showed
that parking space was more important to the partic-
ipants than priority lanes. Hence, the distribution of
individual requirements was modelled into the agent
database. On the other hand, we used the survey data
to validate the model and simulation outputs. The re-
spective survey results concerning the mode choices
are depicted in Figure 3.
4 AGENT-BASED
TRANSPORTATION DEMAND
SIMULATION
After the collection of empirical data, the simulation
on incentives and their effects on choice of transport
mode was performed.
4.1 Simulation Methodology
The executed simulation approach is based on the
open source tool MATSim. This toolbox for agent-
based transportation demand simulation was devel-
oped and improved during the last decade by several
entities and therefore established as a suitable instru-
ment for research. To examine the TDM effects, we
modified the simulation approach in different areas.
Data Input
Connection
Empirical Data
Collection
Mat-Sim Modeling / Configuration
Agent Population
Agent 1
Agent 2
Agent n
Population Network
SeSAm Activities OSM Network
config.xml
Scoring Function
Iteration Loop
Output
Analysis
Running
Scoring
Replanning
Network
Focus Group
Online Survey
1
2
4
3
5
Data Input
Data Input
Connection
Connection
Legend
.
.
.
Figure 2: Simulation Architecture based on MATSim.
Figure 2 depicts the overall simulation process. At
the beginning, essential input data needs to be trans-
formed into a data format (1) which can be used by
MATSim. In the past years, researchers identified ac-
tivity chains of agents as an important part of pub-
lic transportation simulation. The required data is of-
ten gathered by surveys. In this work, we used an-
other agent-based simulation of individual daily rou-
tines (Freudenstein, 2003), which is based on the tool
SeSAm
2
, as a data input. Additionally, the network
data of the observed German city region of Aachen
was converted.
After adjusting the population and network data of
the German region of Aachen, elemental procedures
and agent strategies were configured (2). Then, ev-
ery agent which was part of the overall population
database was modeled, depending on the quantitative
survey results. The specific sensitivity of every person
towards different incentives was evaluated (3), and,
as a consequence, modeled in the agent database (4).
This was realized by adding specific agent attributes.
During the simulation, normally agents estimate
their proceeding via a scoring function. The default
scoring function by MATSim does not estimate the
transport mode explicitly, and the new designed agent
2
http://130.243.124.21/sesam/index.php/
SMARTGREENS2015-4thInternationalConferenceonSmartCitiesandGreenICTSystems
320
Figure 3: Simulation results: Effects of different incentives on mode choice
attributes are not incorporated into this process either.
Hence, the scoring function was extended to incor-
porate incentive estimation, which is modeled as an
individual agent attribute, into the overall agent scor-
ing. Our modifications were based on the LegScor-
ing. In case of incentives, agents estimated the plan
depending on their requirements towards the certain
incentive.
Running the simulation consisted of loading rele-
vant network and population data. Afterwards, agent
plans were routed in the network. Thereby, depending
on events, e.g. traffic congestion, agent plans mutu-
ally influenced each other. Then, agent plans were es-
timated via the extended scoring function. After scor-
ing, agent plans were modified, depending on given
strategies during the configuration phase (2). This
process (5) describes one iteration, in which lower
scored plans are dropped in favor of better estimated
plans. This iteration loop serves as a growing deci-
sion basis for agents to optimize their behavior. The
simulation was performed in up to fifty iterations.
Finally, the results were compared with the survey
data to validate the model parameters.
4.2 Simulation Results
After the validation process, simulation results (Fig-
ure 3 depicted the demand effects corresponding to
the quantitative survey data. All investigated incen-
tives showed relevant effects: the agents choice of
carpooling increased, while the use of ”own car” de-
creased. These were also the only two means of trans-
port for which the shares changed dramatically when
the incentives were introduced, the usage of bike,
public transport and walking hardly changed at all.
Therefore, the latter will not be discussed here in de-
tail. The simulation further showed that when only
one incentive was used, priority parking had the least
effect compared to the scenario with no incentives.
Concerning the combination of incentives, it can be
seen that the more incentives are introduced at the
same time, the greater the share of car pooling be-
comes (at the expense of individual car usage).
5 DISCUSSION
Our stated approach investigated people’s predispo-
sition towards selected incentives, which allowed to
simulate a measure comparison and combination. Re-
Agent-basedTransportationDemandManagement-DemandEffectsofReservedParkingSpaceandPriorityLanesin
ComparisonandCombination
321
sults indicated that the incentives chosen for this sim-
ulation only had an effect for car users in the way
that they would be more willing to participate in car-
pooling. The fact that bike usage and walking did
not change substantially when incentives were intro-
duced could be due to the nature of these two modes:
they can be used flexibly, do not cost anything and
no ticket is needed for them, so the motivation to use
them might be different from e.g. car and public trans-
port.
Principally, this work substantiates the hypothesis
that agent-based demand simulation is a suitable in-
strument to examine the effects of TDM measures and
incentives in urban transportation in comparison and
combination. Compared to established TDM evalua-
tions, agent-based demand simulation investigates the
interactions of large numbers of entities in complex
transportation networks. Furthermore, daily routines
and situational effects like traffic congestion can be
incorporated into investigations. This adds an addi-
tional value for planning processes.
Nevertheless, there are some drawbacks. First, the
explanatory power of ABS approaches is substantially
dependent on the underlying data quality. To validate
an ABS in this scientific field, a widespread investi-
gation under realistic conditions (Jacobs et al., 1982)
is more suitable than an online survey.
In addition, demand simulation becomes more
complex in the area of monetary incentives or regu-
lations. Even though it is basically possible to equip
agents with budgets concerning agent modeling, our
focus groups showed that participants have problems
to specify a certain mobility budget. Moreover, focus
groups indicated the individual requirements concern-
ing specific measures.
In fact, findings like these underline the valuable
contribution of focus group studies already in the be-
ginning of modeling and simulation processes of hu-
man behavior and interactions.
5.1 Limitations and Future Work
Concerning the qualitative and quantitative surveys
on mobility choices, the data has limited significance
for very young or aged population groups, because
the study concentrated on young students. Besides
this, the results were not analyzed for specific user
groups, but only on an aggregated basis for the sample
as a whole. Future research should take into account
that user groups might differ in regard to the most ac-
cepted and most efficient incentive, which could then
also help to strategically tailor incentives to certain
groups of people.
This simulation worked with a simplified model-
ing of car pooling. There are some research efforts
to improve joint trip modeling further (Dubernet and
Axhausen, 2012), which would also be valuable to
implement.
Furthermore, scientific contributions show that
ABS, and especially MATSim, is suitable for
economic investigations, e.g. pricing measures
(Kickh
¨
ofer, 2009). Concerning this background, it
would be interesting to evaluate the demand effects
of mobility service bundling.
ACKNOWLEDGMENTS
This work was funded by the Excellence Initiative of
the German State and Federal Government (Project
Urban Future Outline). Thanks to Lino Kolb for re-
search assistance.
REFERENCES
Axelrod, R. (2006). Agent-Based Modeling as a Bridge
between Disciplines. In Tesfastian, L. and Judd,
K. L., editors, Handbook of Computational Eco-
nomics, pages 1565–1638. Elsevier, Amsterdam.
Ben-Akiva, M. and Atherton, T. J. (1977). Methodology
for short-range travel demand predictions: analysis
of carpooling incentives. Journal of Transport Eco-
nomics and Policy, pages 224–261.
Dubernet, T. and Axhausen, K. W. (2012). Including joint
trips in a multi-agent transport simulation. 12th Swiss
Transport Research Conference.
Ferguson, E. (1990). Transportation Demand Management.
Planning, Development and Implementation. Journal
of American Planning Association, 56(4):442–456.
Freudenstein, J. (2003). Agentenbasierte Simulation indi-
vidueller Tagesabl
¨
aufe (Agent Based Simulation of
Individual Daily Routines). Master’s thesis, RWTH
Aachen.
Hensher, D. A. and Pucket, S. M. (2007). Congestion
and variable user charging as an effective travel de-
mand management instrument. Transportation Re-
search Part A: Policy and Practice, 41(7):615–626.
Horni, A., Axhausen, K. W., and Nagel, K. (2011). High-
Resolution Destination Choice in Agent-Based De-
mand Models High-Resolution Destination Choice in
Agent-Based Demand Models. Technical report, Eid-
gen
¨
ossische Technische Hochschule Z
¨
urich, IVT, In-
stitut f
¨
ur Verkehrsplanung und Transportsysteme.
Ison, S. and Rye, T. (2008). TDM Measures and their Im-
plementation. In Ison, S. and Rye, T., editors, The Im-
plementation and Effectiveness of Transport Demand
Management Measures, pages 1–12. Ashgate Publish-
ing Limited, Hampshire, England.
Jacobs, H. E., Fairbanks, D., Poche, C. E., and Bailey, J. S.
(1982). Multiple Incentives in Encouraging Car Pool
SMARTGREENS2015-4thInternationalConferenceonSmartCitiesandGreenICTSystems
322
Formation on a University Campus. Journal of Ap-
plied Behavior Analysis, 15(1):141–149.
Kickh
¨
ofer, B. (2009). Die Methodik der
¨
okonomischen Be-
wertung von Verkehrsmanahmen in Multiagentensim-
ulationen (Methods of Economic Evaluation for Traf-
fic Measures in Multi Agent Simulations). Master’s
thesis, TU Berlin.
Macal, C. and North, M. (2013). Introductory tutorial:
Agent-based modeling and simulation. In Winter Sim-
ulation Conference (WSC), pages 362–376.
Macal, C. and North, M. J. (2009). Agent-based model-
ing and simulation. In Winter Simulation Conference
(WSC), pages 86–98.
Magg, C. (2012). Agentenbasierte Verkehrsnachfragemod-
ellierung f
¨
ur die Region Stuttgart (Agent Based Mod-
elling of Traffic Demand for the Region of Stuttgart).
Master’s thesis, Uni Stuttgart.
Rennekamp, R. A. and Nall, M. A. (2006). Using Fo-
cus Groups in Program Development and Evaluation.
Technical report, University of Kentucky.
Rodriguez, J. and Murtha, C.-a. T. (2009). Travel Demand
Management.
Schlag, B. and Schade, J. (2004). Public acceptability of
travel demand management. In Huguenin, D. and
Rothengatter, T., editors, Traffic and Transport Psy-
chology. Theory and Application, volume 41, pages
493–500. Elsevier Science Publ., Oxford.
Seik, F. T. (2000). An advanced demand management in-
strument in urban transport Electronic road pricing in
Singapore. Cities, 17(1):33–45.
Smith, R. a. (2008). Enabling technologies for demand
management: Transport. Energy Policy, 36(12):4444–
4448.
Sonnenberger, M. and Ruddat, M. (2013). Was tun?
Strategien zur F
¨
orderung des kollektiven Individu-
alverkehrs. (What to do? Strategies to Foster Collec-
tive Individual Transport.). In Sonnenberger, M., Gal-
lego Carrera, D., and Ruddat, M., editors, Teilen satt
Besitzen, pages 165–191. Europ
¨
aischer Hochschuld-
verlag, Bremen, Germany, 1 edition.
Teal, R. F. (1987). Carpooling: Who, how and why. Trans-
portation Research Part A: General, 21(3):203–214.
Tesfatsion, L. (2006). Agent-Based Computational Eco-
nomics: A Constructive Approach to Economic The-
ory. In Tesfatsion, L. and Judd, K. L., editors, Hand-
book of Computational Economics, chapter 16, pages
831–878. Elsevier, Amsterdam.
Zutshi, A., Grilo, A., and Jardim-Concalves (2013). DY-
NAMOD: A Modelling Framework for Digital Busi-
nesses based on Agent Based Modeling. In IEEE In-
ternational Conference on Industrial Engineering and
Engineering Management (IEEM), pages 1372–1376.
Agent-basedTransportationDemandManagement-DemandEffectsofReservedParkingSpaceandPriorityLanesin
ComparisonandCombination
323