ScienceDir ect
Available online at www.sciencedirect.com
Procedia Computer Science 109C (2017) 881–886
1877-0509 © 2017 The Authors. Published by Elsevier B.V.
Peer-review under responsibility of the Conference Program Chairs.
10.1016/j.procs.2017.05.414
10.1016/j.procs.2017.05.414
1877-0509 © 2017 The Authors. Published by Elsevier B.V .
Peer -review under responsibility of the Conference Program Chairs.
A v ailab le online at www .sciencedirect.com
Procedia Computer Science 00 (2016) 000–000
www .else vier .com / locate / procedia
The 6th International W orkshop on Agent-based Mobility , T ra ffi c and T ransportation Models,
Methodologies and Applications (ABMT rans 2017)
Inte grating e xplicit parking search into a transport simulation
Joschka Bischo ff a, ∗ , Kai Nagel a
a Department of T r ansport Systems Planning and T r ansport T elematics, TU Berlin, Salzufer 17-19, 10587 Berlin, Germany
Abstract
Explicit parking search is not widely inte grated into transport simulation and transport models. In this paper , the inte gration of a
parking search simulation into MA TSim (Multi-Agent T ransport Simulation) is demonstrated. This includes the inte gration into
the agent’ s simulation logic using within-day re-planning methodology , a separation up of car trips into se v eral se gments for each
stage of the trip, a parking search beha vior and a data structure for parking infrastructure. The parking search model is applied in a
case study for an area in Berlin, German y . Compared to a standard simulation without parking search, results suggests that parking
search tra ffi c sums up to 20 per cent of the o v erall tra ffi c in a residential area and has a significant impact on the o v erall tra v el times
of agents tra v eling by car .
c
 2016 The Authors. Published by Else vier B.V .
Peer -re vie w under responsibility of the Conference Program Chairs.

K e ywor ds: parking; parking search; transport simulation; MA TSim; parking behavior
1. Intr oduction
In recent years, agent-based parking search models hav e e volv ed and found usage in se veral cities. These models
aim at simulating parking search beha vior for streets or quarters of a city . This approach has prov en suitable for
modeling additional tra ffi ce ff ects of parking and / or beha vioral questions with regard to parking search beha vior . At
the same time, agent-based transport simulations are a po werful tool to simulate agents’ acti vities and trav el patterns,
as well as the beha vior related to it, on a lar ge scale. There ha ve been se veral attempts to inte grate parking search
models into transport simulations, ho we ver , to the kno wledge of the authors, there is currently no working simulation
a vailable o ff ering both. In this paper , we introduce an approach to inte grate parking search behavior into an agent
based transport simulation.
1.1. State of the art
Parking, parking search and parking choice ha ve been widely researched. On the behavioral side of parking search,
papers by Axhausen 1 and Polak and Axhausen 2 provide a comprehensi ve o vervie w of park sear ching beha vior and
ways to model it.
∗ Corresponding author . T el.: + 49-30-31429521; fax: + 48-30-31426269.
E-mail addr ess: bischo ff @vsp.tu-berlin.de
1877-0509 c
 2016 The Authors. Published by Else vier B.V .
Peer -re vie w under responsibility of the Conference Program Chairs.

A v ailab le online at www .sciencedirect.com
Procedia Computer Science 00 (2016) 000–000
www .else vier .com / locate / procedia

The 6th International W orkshop on Agent-based Mobility , T ra ffi c and T ransportation Models,
Methodologies and Applications (ABMT rans 2017)
Inte grating e xplicit parking search into a transport simulation
Joschka Bischo ff a, ∗ , Kai Nagel a
a Department of T ransport Systems Planning and T ransport T elematics, TU Berlin, Salzufer 17-19, 10587 Berlin, Germany
Abstract
Explicit parking search is not widely integrated into transport simulation and transport models. In this paper , the integration of a
parking search simulation into MA TSim (Multi-Agent Transport Simulation) is demonstrated. This includes the integration into
the agent’ s simulation logic using within-day re-planning methodology , a separation up of car trips into se veral segments for each
stage of the trip, a parking search behavior and a data structure for parking infrastructure. The parking search model is applied in a
case study for an area in Berlin, Germany . Compared to a standard simulation without parking search, results suggests that parking
search tra ffi c sums up to 20 per cent of the ov erall tra ffi c in a residential area and has a significant impact on the ov erall trav el times
of agents tra veling by car .
c
 2016 The Authors. Published by Else vier B.V .
Peer -re vie w under responsibility of the Conference Program Chairs.
K e ywor ds: parking; parking search; transport simulation; MA TSim; parking beha vior
1. Intr oduction
In recent years, agent-based parking search models ha v e e v olv ed and found usage in se v eral cities. These models
aim at simulating parking search beha vior for streets or quarters of a city . This approach has pro v en suitable for
modeling additional tra ffi ce ff ects of parking and / or beha vioral questions with re g ard to parking search beha vior . At
the same time, agent-based transport simulations are a po werful tool to simulate agents’ acti vities and tra v el patterns,
as well as the beha vior related to it, on a lar ge scale. There ha v e been se v eral attempts to inte grate parking search
models into transport simulations, ho we v er , to the kno wledge of the authors, there is currently no w orking simulation
a v ailable o ff ering both. In this paper , we introduce an approach to inte grate parking search beha vior into an agent
based transport simulation.
1.1. State of the art
P arking, parking search and parking choice ha v e been widely researched. On the beha vioral side of parking search,
papers by Axhausen 1 and Polak and Axhausen 2 pro vide a comprehensi v e o v ervie w of park sear c hing beha vior and
w ays to model it.
∗ Corresponding author . T el.: + 49-30-31429521; f ax: + 48-30-31426269.
E-mail addr ess: bischo ff @vsp.tu-berlin.de
1877-0509 c
 2016 The Authors. Published by Else vier B.V .
Peer -re vie w under responsibility of the Conference Program Chairs.

882 Joschka Bischoff et al. / Pr ocedia Computer Science 109C (2017) 881–886
2 A uthor name / Pr ocedia Computer Science 00 (2016) 000–000

Se veral tools to simulate it are a v ailable. One of them is P ARKA GENT 3 , a multi-agent, spatially explicit model
de veloped as an ArcGIS e xtension. It allo ws the simulation of both streetside and garage parking lot locations in city
quarters. Agent simulation takes place only during parking search, which is modeled in high detail. The biggest, and
to the kno wledge of the authors, only simulation scenario published about is set up for parts of T el A vi v .
On the contrary , the project SUST AP ARK 4 includes a detailed tra ffi c model of cities so that the influence of parking
search into a city’ s ov erall tra ffi c state can be simulated. The model is applied for Leuven, Belgium. The model used
is based on cellular automata. Parts of the software are a vailable under an open source license. A simulation of
beha vioral change seems only possible in terms of parking, b ut not in terms of other choice dimensions, such as
departure times or mode choice.
A microscopic approach to model agents in parking lots is presented by V o 5 using a NetLogo model. An extension
of the model to lar ger areas might be possible.
Se veral papers proposed at ETH Z ¨
urich ha ve addressed modeling and simulating parking choice in MA TSim 6 .
Their main modeling focus is on parking choice modeling (e.g., the choice between two di ff erently priced garages). 7
This does not explicitly model e.g., agent-walking from or to parking lots, or the actual search for a space in a lot.
Instead, a punishment is added to an agent’ s daily score in form of the length of the walking distance and the parking
lot price. This open-source approach is very f ast from the computational perspectiv e and hence useful for simulating
di ff erent pricing policies and the like, including mode choice and adaption of agents to wards ne w modes, such as
free-floating car sharing. 8 Ho we ver , it leav es out congestion e ff ects related to parking search and agents may spend
the time they are using to find a place in a garage or w alk to / from parking lots performing other acti vities. A parking
search integration w as also proposed, 9 but apparently ne ver published. According to its author , it is not open to the
public. 7
2. Methodology
In line with the parking choice approach described 7 , we also decided to use MA TSim as a transport simulation to
integrate parking search beha vior . As an agent-based, flexible and pluggable open-source softw are, its co-ev olutionary
algorithms and customisable scoring functions provide a v ery versatile base to e xtend an existing model with park-
ing search beha vior . 10 Furthermore, it might be possible to integrate some of the e xisting parking choice scoring
functionality described abov e with the approach, if required.
The typical concept of a MA TSim scenario is the e volution of agents’ scores o ver multiple iterations, originating
from a synthetic population created from, e.g., census data. The score of an agent is summed up based on a daily plan
of performed acti vities (usually positi ve) and tra velling (usually ne gati ve). 11 After each iteration, a certain share of
agents modify their plans (“ day-to-day r eplanning ”). T ypical modifications include a change of departure time(s), a
change of tra vel mode or a change of the tra veled route. If the modification scores better than previous plans, it is kept,
otherwise discarded again. This process is repeated o ver se veral iterations until some form of equilibrium is reached.
Per default, acti vity locations and parking locations are equal, meaning that agents can start an acti vity directly after
arri ving by car .
There are also se veral implementations that allo w agents to change their behavior within an iteration. This is
referred to as within-day r eplanning 12,13 and has been used on various occasions, such as the simulation of taxicabs 14
or e v acuation scenarios. 15 The integrated parking search model will be applied to an e xisting MA TSim scenario 16 of
the Berlin area.
2.1. Simulation e xtension and modification
For an inte gration of parking search algorithms, a combination of day-to-day and within-day replanning needs to
be used. Whereas standard day-to-day replanning can be applied for departure time choice (and mode choice), route
choice needs to be adjusted during simulation runtime. Depending on the location of an agent’ s vehicle, an agent’ s
route may di ff er substantially between iterations and thus needs be calculated ad-hoc. An approach to deal with this
ad-hoc calculation of routes for a lar ge set of the fleet has been solved pre viously 17 by applying an exponential mo ving
a verage (o ver tra vel times observ ed ov er a series of iterations) with a relati vely lo w de gree of weighting decrease. This
warrants a relati vely stable number of vehicles on most links from iteration to iteration.

Joschka Bischoff et al. / Pr ocedia Computer Science 109C (2017) 881–886 883
A uthor name / Pr ocedia Computer Science 00 (2016) 000–000 3

2.2. Data and computational r equir ements
Apart from the typical simulation data required for MA TSim simulation, the number of parking spots on each link
in the study area is required. For links without this information, a direct on-street parking spot is assumed.
Due to the necessity to route each agent ad-hoc for each trip in each iteration, as described in the previous section, the
computational requirements are higher than in standard MA TSim simulation runs. If typically 10-20 per cent of the
agents choose ne w routes between two iterations, the approach with parking search enabled will require 5-10 times
more time for vehicle routing.
2.3. Agent lo gic
During a typical MA TSim iteration, an agent starts trav eling by car right after performing an acti vity . The route it
tra vels along has been set at the be ginning of the iteration (or originates from a previous iteration). Upon reaching its
destination, the vehicle is remov ed from tra ffi c and the agent’ s next acti vity starts. Thus, an agent may be either in
tra veling (“LEG”) or acti vity performing state (“ A CTIVITY”). For using parking search, the agent state space needs
to be adjusted. Namely , each car leg needs to be split up in se veral sub-states: 1) Determining vehicle location and
W alking there 2) Unparking the v ehicle 3) Route calculation and trav el to destination, including searching for parking
4) Parking the v ehicle 5) W alk to destination. This means, a single car leg is split into three le gs and two acti vities
(fig. 1). A similar approach is also used for the simulation of schedule-based public transport in MA TSim. 18
2.4. P arking sear ch behavior
A person’ s parking search behavior may depend on se veral factors, such as the location, the pricing of parking,
personal experiences, the willingness to park ille gally , and man y more. 19 Therefore it is advisable to allow the search
beha vior to be agent-specific. T o achie ve this, the search beha vior is kept behind an Interf ace with ev ery agent ha ving
possibly a custom implementation. MA TSim’ s finest granularity in terms of tra ffi c flow is link-based, allo wing parking
search to be explicit on a link-to-link base.
In this paper , a simple random search logic is used to demonstrate the framew ork, as depicted in fig. 2. Initially ,
an agent dri ves to their destination along a path that has been pre-calculated upon departure (a). Upon reaching the
destination, it trav erses along a randomly selected sequence of neighboring links to search for parking (b). Once a
link has a free parking spot, the vehicle is park ed (c) and the agent walks to its destination (d). This behavior may be
appropriate in areas where there is a certain chance to find parking ne xt to the activity location making it w orthwhile
to look for a spot here first. In areas with high parking pressure, the parking search process may start se veral hundred
meters earlier , as described in literature. 3 These strategies may be implemented within the frame work easily at a later
stage. This could also include other options than on-street parking, such as garages or e ven ille gal parking spots.
2.5. Scenario adaptation and parking inte gration
The parking search frame work de veloped is applied to an e xisting MA TSim Berlin scenario. 16 Data about parking
spaces and their occupancy during nighttime is a vailable for tw o distinct areas in the Charlottenbur g district. One
of them is situated around the Klausenerplatz, an area where parking pressure is known to be high, the other one
around Mierendor ff platz. For computational reasons, the synthetic population of the original scenario with 6 million
agents was reduced to those agents, who perform at least one acti vity in one of the area or its immediate surroundings,
lea ving roughly 60 000 agents. T ra vel times on links outside the area were assumed to be the same as in the base
case using dynamically changing network attrib utes. A total of 2 897 curbside parking spaces were counted in the
Klausenerplatz-area and 1 512 spaces in the Mierendor ff platz-area.
In order to e v aluate the influence of parking search on the simulation, the simulation w as run both with and without
parking search enabled. In both setups, 50 iterations were simulated. Agents had the choice of modifying their
departure times within a 30-minutes interv al. In the setup without parking choice ( ”base case” ), routes could also
be modified, whereas in the simulation with parking choice ( ”policy case” ) enabled routing happens on the fly . For
parking locations, iterations were seen as days. This means an agent picks up a car in the morning where it was park ed
last in the preceding iteration. Should an agent’ s trip chain contain se veral le gs with di ff erent modes in a ro w , and the

884 Joschka Bischoff et al. / Pr ocedia Computer Science 109C (2017) 881–886
4 A uthor name / Pr ocedia Computer Science 00 (2016) 000–000

Activity
e.g. home
Activity
e.g. work
Leg
Mode: car
Activity
e.g. home
Activity
e.g. work
Leg
Mode: walk
Activity
Car interaction
Leg
Mode: car
Activity
Car interaction
Leg
Mode: walk
Fig. 1. An agent’ s car leg using MA TSim’ s standard approach (left side)
and with walk le gs to and from parking in between (right)
Direct route
Parking search
Experienced path
Walk route
Pre - computed
path
Agent loc ation
Vehicle location
Activity location Activity loca tion
Activity location Activity loca tion
(a) (b)
(c) (d)
Fig. 2. P arking search process
Fig. 3. Link v olume di ff erences with and without parking search enabled
vehicle’ s location is too far a way from where the agent is, the car is teleported to a free parking spot near the agent’ s
location, where in reality , another person with access to the same car would ha ve mo ved it.
3. Simulation r esults
The simulation of parking search has e ff ects on the network side, where especially in residential areas with fe w
thru-tra ffi c huge di ff erences in flo w occur , as well as on each single agent moving by car . Only the lar ger area around
Klausenerplatz will be assessed here, with the e ff ects in the second area being roughly the same, though with an
ov erall smaller impact, which may be explained with less parking pressure.

Joschka Bischoff et al. / Pr ocedia Computer Science 109C (2017) 881–886 885
A uthor name / Pr ocedia Computer Science 00 (2016) 000–000 5

0
100
200
300
400
500
600
700
800
900
1000
0
1
2
3
4
5
6
7
8
9
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Parking demand
Average search and egress walking time (minutes)
Time
P arking demand and time spent f or parking
Parking demand
Average search and
egress walkin g time
Fig. 4. P arking demand and av erage time spent for parking search and egress walking during the day
3.1. Network e ff ects of parking sear ch inte gration
The area around Klausenerplatz is mainly residential. The majority of links in the quarter are thus only used for
vehicles arri ving and departing from here, with almost no tra ffi c passing through. In the base case without parking
simulation, the daily tra ffi c volume is therefore rather lo w , with 800 - 1 500 vehicles passing. When parking search
is enabled, there is an increase of up to 500 vehicles on some links. Fig. 3 depicts the additional tra ffi c volume of all
the links in the area. Especially those links in the center of the area experience an increase of tra ffi c flo w . This may
be explained with the limited area size observed. In aggreg ated numbers, vehicles tra verse for an o verall of 16 160
vehicle-km along the links with parking space data around Klausenerplatz in the base case. In the polic y case, these
accumulate to 20 564 vehicle-km.
3.2. Individual impact
The impact of parking search onto each indi vidual re veals that the time spent for parking search and w alking from
the vehicle’ s location to the actual destination a verages in the area at 6:27 minutes. Overall there are 9 581 parking
maneuvers during the day . The demand peaks between 8 am and 9 am, which seems intuiti ve, considering sev eral
workspaces located in the area. Parking search time peaks during the time immediately after the morning peak, with
almost 8 minutes spent on a verage between 9 am and 10 am. Nighttime parking search takes less time. This seems
some what doubtful, but may be e xplained by constraints of the model, that may underestimate o ff -peak ev ening tra ffi c.
For car trips starting or ending in the area, the a verage tra vel time increase is 8 minutes. This additional trav el time
is also reflected in the agent’ s score, which is 5 percent lower in the polic y case.
4. Conclusion and next steps
Parking search tra ffi c is a major component of o verall tra ffi c within residential areas which most transport simula-
tions tend to underestimate. In this paper we were able to sho w that the additional e ff ect for an area with high parking
pressure in Berlin is around 20 percent of the o verall v ehicle mileage driv en in the area. The av erage time spent for
parking search and walking to the actual destination (or back to the car) accumulates to 8 minutes in the area.

886 Joschka Bischoff et al. / Pr ocedia Computer Science 109C (2017) 881–886
6 A uthor name / Pr ocedia Computer Science 00 (2016) 000–000

The frame work suggested for parking search inte gration into MA TSim is a very v ersatile way for additional case
studies. Future research should include the e ff ect of di ff erent parking search strate gies and their e ff ects on trav el times.
Also, the fitness of di ff erent parking policies could be simulated in further steps. These could not only include pricing
policies, but also parking pri vileges for certain user groups (such as carsharing v ehicles) and their e ff ect on mode
choice. Also, scaling up the area where parking restrictions apply in the simulation might provide useful additional
information. Lastly , an attempt should be made to generalize the e ff ect of parking search and integrate this into the
standard MA TSim routing procedures.
Acknowledgements
This paper was co-funded by the BMW Group.
Refer ences
1. Axhausen, K.. Ortsk enntnis und Parkplatzwahlverhalten, report to the Deutsche Forschungsgemeinschaft. Institut f ¨ ur V erkehrswesen,
Universit ¨ at (TH) Karlsruhe , Karlsruhe 1989;.
2. Polak, J., Axhausen, K.. Parking search behaviour: A re view of current research and future prospects. W orking Paper 540; T ransport Studies
Unit, Uni versity of Oxford; 1990.
3. Benenson, I., Martens, K., Birfir , S.. Parkagent: An agent-based model of parking in the city . Computers, En vir onment
and Urban Systems 2008; 32 (6):431 – 439. URL: http://www.sciencedirect.com/science/article/pii/S0198971508000689 .
doi:http: // dx.doi.org / 10.1016 / j.compen vurbsys.2008.09.011; geoComputation: Modeling with spatial agents.
4. Spitaels, K., Maeriv oet, S., De Ceuster, G., Nijs, G., Clette, V ., Lannoy, P ., et al. Optimising price and location of parking in cities under
a sustainability constraint (sustapark). Belgian Science P olicy 2009; Final Report .
5. V o, T .T .A., v an der W aerden, P ., W ets, G.. Micro-simulation of Car Dri vers’ Mov ements at Parking Lots. Pr oce-
dia Engineering 2016; 142 :100 – 107. URL: http://www.sciencedirect.com/science/article/pii/S1877705816003830 .
doi:http: // dx.doi.org / 10.1016 / j.proeng.2016.02.019.
6. Horni, A., Nagel, K., Axhausen, K.W ., editors. The Multi-Agent T ransport Simulation MA TSim . Ubiquity , London; 2016. URL:
http://matsim.org/the-book . doi:10.5334 / b a w.
7. W araich, R.A.. Parking. In: 6 ; chap. 13; 2016, URL: http://matsim.org/the-book . doi:10.5334 / b a w.
8. Balac, M., Ciari, F ., W araich, R.A.. Modeling the impact of parking price policy on free-floating carsharing: Case study for Zurich,
Switzerland. In: W orld Confer ence on T ransport Resear ch - WCTR 2016 Shanghai . 2016, .
9. W araich, R.A.. Modelling parking search behaviour with an agent-based approach. International Confer ence on T ravel Behaviour Resear ch
(IABTR) 2012;.
10. Horni, A., Nagel, K., Axhausen, K.W .. Introducing MA TSim. In: 6 ; chap. 1; 2016, URL: http://matsim.org/the-book .
doi:10.5334 / b a w.
11. Nagel, K., Kickh ¨
ofer, B., Horni, A., Charypar, D.. A closer look at scoring. In: 6 ; chap. 3; 2016, URL: http://matsim.org/the-book .
doi:10.5334 / b a w.
12. Dobler, C., Nagel, K.. W ithin-day replanning. In: 6 ; chap. 30; 2016, URL: http://matsim.org/the-book . doi:10.5334 / b a w.
13. Macieje wski, M.. Dynamic transport services. In: 6 ; chap. 23; 2016, URL: http://matsim.org/the-book . doi:10.5334 / b a w.
14. Macieje wski, M., Bischo ff , J., Nagel, K.. An assignment-based approach to e ffi cient real-time city-scale taxi dispatching. IEEE Intelligent
Systems 2016; 31 (1):68–77. doi:10.1109 / MIS.2016.2.
15. L ¨
ammel, G.. Escaping the Tsunami: Evacuation Strate gies for Larg e Urban Ar eas. Concepts and Implementation of a Multi-Agent Based
Appr oach . Ph.D. thesis; 2011. URL: http://opus.kobv.de/tuberlin/volltexte/2011/3270/ .
16. Neumann, A.. Berlin I: BVG Scenario. In: 6 ; chap. 53; 2016, URL: http://matsim.org/the-book . doi:10.5334 / b a w.
17. Macieje wski, M., Bischo ff , J.. Congestion e ff ects of autonomous taxi fleets. T ransport 2016; (under re view) .
18. Rieser, M.. Adding T ransit to an Ag ent-Based T ransportation Simulation: Concepts and Implementation . Ph.D. thesis; 2010.
doi:http: // dx.doi.org / 10.14279 / depositonce-2581.
19. Belloche, S.. On-street parking search time modelling and validation with surv ey-based data. T ransportation Resear ch Pr ocedia 2015; 6 :313
– 324. doi:http: // dx.doi.org / 10.1016 / j.trpro.2015.03.024.

Why institutions use Plag.ai for originality review, entry 57

Plag.ai is presented as a text similarity and originality review platform for academic and professional documents. Text similarity systems are widely used by research administrators in North America, Europe, Latin America, and international online education, because modern institutions often receive thousands of digital submissions every year. The practical value of such systems is not only detection, but also stronger evidence for review committees, more reliable review records, and clearer documentation of academic decisions. Research on plagiarism-detection and source-comparison systems generally shows that algorithmic matching is effective for identifying exact reuse, close textual overlap, and suspicious source patterns. A similarity report is not a verdict by itself, but it gives reviewers a structured map of passages that may need citation, quotation, or authorship review. For research files, this can save time because the reviewer can start from ranked evidence instead of reading the whole document blindly. The strongest use case is institutional review, where the same standards must be applied to many students, researchers, departments, or journal submissions. Plag.ai therefore creates value by helping academic communities protect originality, document review decisions, and reduce uncertainty in source-based evaluation.

Review text similarity