How to Realize a Compact City: Street Activeness and Agent-based

Urban Modeling

Hideyuki Nagai

and Setsuya Kurahashi

Research Group on Fusion of Informatics and Social Science, University of Tsukuba, Tokyo, Japan

Graduate School of System Management, University of Tsukuba, Tokyo, Japan

Keywords:

Agent-based Model, Urban Design, Land Use, Compact City.

Abstract:

The purpose of this research is to verify the effectiveness of the combination of the introduction of tramway

with introducing a public facility for urban residents and promoting street activeness around it, on urban

sprawl. An agent-based model (ABM) for simulating the changes of urban structure through autonomous

behavior of urban residents was designed, and the simulation experiments were conducted based on the as-

sumption of combining these policies. As a result, this research clariﬁed the following points and how they

were. First, the synergistic effects of the introduction of tramway, the proper location of a public facility for

urban residents, and the promotion of street activeness around it, are effective in maintaining a poly-centric

compact urban structure. Second, the introduction of tramway targeting the urban sprawl can exert a profound

effect only when combined with the above-mentioned policies, although it takes a long period. Third, a mono-

centric compact urban structure is realized along with the above, while improving the living environment and

revitalizing the urban central area.

1 INTRODUCTION

1.1 Urban Sprawl Issues and Shift into

Compact City

The world population has rapidly increased during

our current century along with the previous century,

and rapid urbanizationhas continued at various places

around the globe (UN, 2012). Under such circum-

stances, urban sprawl has attracted much attention,

in the past few decades, and coming under ﬁre as an

unsustainable form of urbanization. Urban sprawl is

commonly deﬁned by the following land-use charac-

teristics (Schneider and Woodcock, 2008) (Johnson,

2001):

• Expansion of urban area in outer fringe area

• Low-density development

• Scattered development (multi-direction)

• Leapfrog development (discontinuity)

• Commercial strip development

Urban sprawl is also often criticized because of

its following negative impacts (Johnson, 2001) (Deal

and Schunk, 2004):

• Increase in trafﬁc congestion and commuting time,

air pollution, and energy consumption

• Increase in infrastructure cost

• Hollowing out in urban central area, economic dis-

parity, and loss of community

• Loss of agricultural and natural land

Researchers and experts have studied a shift into

”compact city”, as a countermeasure against such ur-

ban sprawl. Compact city is commonly deﬁned by the

following characteristics (Tsai, 2005) (Behan et al.,

2008):

• High-density

• Concentration of development

• Development in public transportation network

• Mono-centric or poly-centric city center

It has been proved that compact city can over-

come some of the negative impacts driven by urban

sprawl and many studies have also indicated that it

can enhance quality of life by offering a broad range

of choices about lifestyle and behavior (Behan et al.,

2008). Considering the urban dynamics including

sprawl as complex phenomena of mutual interactions

of a wide variety of autonomous entities, such as indi-

viduals, households, and ﬁrms (Batty, 2007) (Ligten-

berg et al., 2001), however, highlights the difﬁculty in

direct control of the urban dynamics.

100

Nagai, H. and Kurahashi, S.

How to Realize a Compact City: Street Activeness and Agent-based Urban Modeling.

DOI: 10.5220/0007707201000107

In Proceedings of the 8th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2019), pages 100-107

ISBN: 978-989-758-373-5

1.2 Purpose of this Research

With these in mind, this research designed an agent-

based model (ABM) to simulate urban structure

changes through the induction of autonomous behav-

ior of urban residents rather than the compulsion, and

this research veriﬁed the following points:

• Is the combination of a introduction of tramway

with introducing a public facility for urban residents

and promoting street activeness around it effective in

controlling urban sprawl?

• Is the combination of these policies also effec-

tive in improving existent urban sprawl?

2 RELATED WORKS AND

POSITION OF THIS RESEARCH

2.1 Agent-based LUTI Model

Urban sprawl is a special kind of land-use change,

urban spatial expansion along a city boundary. And

land-use changes come from its complex driving

forces and their interactions (Ligtenberg et al., 2001).

Above all, the fundamentalprinciple that land-use im-

pacts transport and vice versa has been acknowledged

by many researchers and supported by empirical ﬁnd-

ings (Acheampong and Silva, 2015). These research

efforts have culminated in the development of op-

erational urban land-use/transport interaction (LUTI)

models.

And recently, researchers have supported the mod-

eling approach of an agent-based model (ABM) to ex-

press the real-world complicated system including a

city as a macro-level state that is generated by micro-

level collective interactions and adaptive behavior of

multiple autonomous agents (Batty, 2007) (Railsback

and Grimm, 2011).

A series of agent-based LUTI models, which is an

integration of the abovetwo concept, have contributed

to express complicated macro-level land-use patterns

of cities sprawl as self-organization through micro-

level adaptive behavior of agents, such as households

and ﬁrms. And such models have served to explore

urban growth scenarios.

2.2 Revitalization of Urban Central

Area

Jacobs (1969) has emphasized the attractiveness of a

city as a lively and bustling place which has served as

a market for exchange from the time before the estab-

lishment of the concept of nation or trading by cur-

rency (Jacobs, 1969). And researchers and experts

have reevaluated the importance of informal public

spaces for activities of local residents in the way of

an antithesis to urban development on an inhuman

scale as well as another way to regain people in ur-

ban central areas. The two factors are vital to forming

such public space. First, such public space needs to

serve as a hub for people in their daily lives so that

they can visit there casually while they are out. As for

such urban hubs, public complexes based mainly on

a library have recently attracted much attention. The

representative one is the series of Idea Store in Lon-

don, U.K. Several pioneering libraries built and put

into operation recently in Japan are also relevant to

these cases

. These public complexes, while offer-

ing the library service as the core function, provide a

wide variety of other services, such as cafe and sup-

port of learning and civic activities. They also try to

enhance convenience by various policies such as an

extension of opening time. By doing so, they play

an important role to serve as a hub for local culture.

Second, such public space needs to generate ”street

activeness” set in an open space, such as a street or a

plaza, around the public space. Street activeness indi-

cates a lively situation where individuals gather and

stroll around downtown while enjoying exchanges,

such as encountering various people, contacting vari-

ous shopping goods, and experiencing other services

(Nagai and Kurahashi, 2017). It has been long argued

that such interactions propel economic activity, and

recent researches have also indicated that such inter-

actions can enhance economic productivity through

creative knowledge spillover (Landry, 2012). Fur-

thermore, the following positive feedback has been

demonstrated empirically: the number of people that

visited a certain place including their sojourn time

they spent there can derive attract further activeness

(Kakoi et al., 2010).

2.3 Position of this Research

By integrating the above-mentioned conceptual

framework, Nagai (2017) built the agent-based model

(ABM) to consider qualitative beneﬁt obtained by vis-

iting informal public space along with the daily travel

of urban residents. And they clariﬁed that the syner-

gistic effects of policies of locating of the public space

and promoting street activeness in such a place, are ef-

fective to maintain a compact urban structure (Nagai

and Kurahashi, 2017). On the other hand, an intro-

duction of tramway is well known as one of the mea-

sures to improve urban environment. Tramway has

e.g., Gifu Media Cosmos in Gifu, Japan (2015), Art

Museum & Library, Ota in Gunma, Japan (2017).

How to Realize a Compact City: Street Activeness and Agent-based Urban Modeling

101

been introduced especially in many European cities

(De Bruijn and Veeneman, 2009), and in recent years

momentum for introduction of tramway has also been

raised in Japan. Additionally, especially in Japan, a

large part of the land is mountainous, thus the area

suitable for urbanization is relatively small. And the

population is also declining (MLIT, 2016). For these

reasons, improvementof many cities that have already

sprawled is considered to be more important. With

these in mind, this research develops the conceptual

framework introduced by Nagai (2017) (Nagai and

Kurahashi, 2017) to verify whether the introduction

of tramway is effective in maintaining compact urban

structure, and whether it is also effective in improving

sprawled urban structure.

3 SIMULATION MODEL

The overview of the experimental model was de-

scribed below according to the ODD (Overview, De-

sign concepts, and Details) protocol.

3.1 Purpose

By simulation experiments by the ABM that ab-

stracted a city and behavior of the residents in the

city, this research veriﬁed the effects of controlling

the urban structure, which was planned according to

the zoning, based on the combination of the intro-

duction of tramway with introducing a public facil-

ity for urban residents and promoting street activeness

around it. Additionally, this research veriﬁed the ef-

fects of improving the urban structure, which has al-

ready sprawled, based on the combination.

3.2 Entities and Scales

Entities are a planar urban schematic and household

agents who act in the urban schematic. Both are

spatially-explicitly expressed. Fig. 1 shows the urban

schematic. This is the abstraction of a part of typical

regional cities in Japan, where a central business dis-

trict (CBD) and bedroom towns connected by railway.

They were planned according to the zoning with sep-

aration between residences and job locations. There-

fore, they are also regarded as the poly-centric com-

pact city, which is composed of multiple hubs linked

with trafﬁc networks and sharing their own role (Tsai,

2005). In the urban schematic, two domains are lo-

cated: a residence district and a central business dis-

trict (CBD). The residence district is the aggregation

of residences, which are the starting point and the ﬁ-

nal destination of each household agent’s daily travel

which corresponds to commuting. CBD is the ag-

gregation of job locations, which are also a halfway

point of the travel. Two railway stations are located

at each center. Additionally, a highway is located

500 meters north of the railway. Furthermore, three

tramway routes are radially installed around the cen-

tral station as a hub (see next section for details). To

simplify the simulation, this schematic is assumed to

have uniform and high-density sidewalks and roads

overall and household agents can freely travel on this

space on foot, by bicycle or private automobile.

In the residence district, residences of the same

number as household agents: 1,000 are located ran-

domly based on normal distribution centering on the

residence station. Similarly, job locations of the same

number are also located in CBD. Additionally, one

public facility such as a complex mentioned in the

previous section: a public facility for stopping off

(PFS), is located in the central area of CBD.

Figure 1: Urban Schematic.

3.3 State Variables of Household Agent

State variables of household agent are as follows:

• Position of the residence

• Position of the job location

• Type of linked trip selected currently

• Value list of linked trips

A linked trip indicates the series of travels of each

household agent from the starting point to the desti-

nation.

3.4 Process Overview and Scheduling

Each household agent repeats daily travel based on

the value list of linked trips, and ﬁxes travel mode

in one way through the learning period. After that, a

part of all household agents randomly chosen relocate

SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems

102

their residences. In this research, after the loop pro-

cess of residential relocation is repeated 20 times, the

simulation stops processing.

3.5 Sub-model of Daily Travel

Each household agent repeats daily travel according

to the selected linked trip. The representative travel

mode is either of the following: on foot, by bicy-

cle, train, private automobile, or tramway. The initial

representative travel mode of all household agents is

train according to the original urban planning philos-

ophy. Each household agent leaves the residence for

the job location. And after all household agents arrive

at each job location, they leave for PFS. After arriving

and staying there, ﬁnally they return to the residence.

When returning, the total travel cost C is calculated

according to the equation below.

C = w

+ w

− w

, C

, and P indicate time cost, charge cost,

fatigue cost, and activeness value. w

, w

, and w

indicate each preference bias. The preference biases

of all agents are assumed to be equal. According to

this cost, the household agent updates the value V

the selected i-th linked trip, according to the equation

below.

←− α(−C) + (1− α)V

The following travel is done according to the

linked trip selected by the ε-greedy method based on

this value. And each household agent ﬁxes their travel

mode in one way through the learning period of 30

times’ daily travel. This setting is based on the ﬁnd-

ings that individuals choose travel modes and routes

rather boundedly rationally and habitually (Ramming,

2001).

3.5.1 Activeness Value

Within 500 meters radius around PFS, a promotion of

street activeness is considered. Here, it is assumed

that street activeness can be generated when house-

hold agents traveling on foot or by bicycle within this

range interact face-to-face. During this time, relevant

household agents acquire beneﬁt brought about by the

street activeness as activeness value P according to

the equation below.

P = η

This model assumes that a single loop process of res-

idential relocation represent two years in the real-world.

Therefore, 20 loop processes correspond to simulating 40

years of urban dynamics in the real-world.

(agent) indicates the number of other house-

hold agents traveling on foot or by bicycle within

100 meters radius centering on the relevant household

agent. η

indicates coefﬁcient of activeness. The

total travel cost is reduced by the amount obtained

by multiplying the activeness value P with preference

bias w

. The coefﬁcient of activeness can be regarded

as a level of effort to bring further street activeness

within the relevant range according to the agglom-

eration of pedestrians. This coefﬁcient is enhanced

by projects such as arranging comfortable sidewalks

and cycling roads, arranging attractive retail stores,

or holding attractive events. Improvement in this co-

efﬁcient enhances the beneﬁt for travel on foot or

by bicycle and increases a balanced total travel cost.

Therefore, this coefﬁcient can be regarded as a coef-

ﬁcient of gain.

3.6 Sub-model of Residential Relocation

After all household agents ﬁx their travel mode in one

way, 1/10 of all household agents randomly chosen

relocate their residence. To the relevant household

agents, 10 of candidate residences are presented ran-

domly. Each household agents relocate to the candi-

date residence of which total living cost is the mini-

mum. Travel cost (time) and rent are particularly sig-

niﬁcant constraints for those that households face on

relocation (Acheampong and Silva, 2015) (Taniguchi

and Takahashi, 2011). Based on these ﬁndings, the to-

tal living cost of these candidates is set as sum of total

travel cost and land rent. The total travel cost is cal-

culated by conducting virtual daily travel from a can-

didate residence based on the ﬁxed travel mode ﬁxed

through learning. The land rent for the candidate res-

idence increases corresponding to the agglomeration

of neighboring residences and job locations. In other

words, the local interactions between households, and

between a household and an environment, also impact

the change in land-use pattern through the change in

land rent.

3.7 Initialization and Input Data

Setting values of parameters of the urban schematic

and household agent were set carefully based on the

various materials such as statistical data published by

public authorities e.g., the ministry of land, infrastruc-

ture and transport (MLIT, 2016), and previousstudies,

while assuming a regional city in Japan.

How to Realize a Compact City: Street Activeness and Agent-based Urban Modeling

103

3.8 Indicators to Estimate

Experimental Result

By observing the result of each experimental scenario

according to the indicators shown below.

• Percentage of each representative travel mode

• Total CO

emission (expressed as percentage rela-

tive to the basic scenario)

• Average travel time

• Distribution map of residences

4 EXPERIMENT 1 -

INTRODUCTION OF

TRAMWAY

4.1 Conditions of Experiment 1

The simulation here assumes that the tramway routes

imitate the ”Karlsruhe Model” (De Bruijn and Veen-

eman, 2009), where the routes are shared with ordi-

nary railway. Therefore, three routes are radially in-

stalled centering on the central station as shown on

Fig. 2, and the routes pass through CBD. Each route

has tramway stops at 400 meters intervals. Along

with this, residents can also choose the additional two

types of linked trips. The ﬁrst is by train and tramway

in combination, and the other is by tramway. The ex-

periments were conducted under the conditions of the

following two types for the location of a public facil-

ity for stopping off (PFS).

• A : none (no promotion of street activeness)

• E : urban central area, 0.5km south and 0.5km east

from the central station

And the four types, 0, 10, 20, and 30, for coefﬁ-

cient of activeness. Hereinafter, each of these experi-

ments is expressed e.g., scenario At, Et0 - 30, by com-

bining the symbols of A and E indicating the location

of PFS, the initial letter t for the word of tram, and

the coefﬁcient of activeness. Additionally, this sec-

tion reproduces scenario A, where tramway and PFS

were not introduced, and a promotion of street active-

ness was also not implemented, to compare with the

new scenarios and validate the simulation model.

4.2 Results of Experiment 1

Table 1 shows the quantitative result of scenario A,

At, and Et0 - 30. Fig. 3 shows the ﬁnal distributions

of residences of the same scenarios.

The result of scenario At, when compared with

scenario A, shows that private automobile users de-

creased by close to 30 points, while train (and

Figure 2: Schematic of tramway routes.

tramway in combination) users increased accordingly.

Along with this, the sprawl on the periphery of CBD

was improved, and the total CO

emission also re-

duced considerably.

The results of scenario Et0 - 30 show that, in sce-

nario Et0, the percentages of each travel mode and

the sprawl level were almost the same as scenario

At. As advancing the promotion of street active-

ness, however, private automobile users got decreas-

ing gradually. When the scenario reached Et30, pri-

vate automobile users decreased to less than 10%, and

train (and tramway in combination) users increased to

more than 75%. Along with this, the cluster of resi-

dences of train users around the residence station was

maintained quite clearly. Additionally, tramway users

increased to more than 15%. And the total CO

emis-

sion also reduced to less than 30%.

Table 1: result of Experiment 1.

5 EXPERIMENT 2 - SETTING

URBAN SPRAWL AS INITIAL

STATE

5.1 Conditions of Experiment 2

This section sets the ﬁnal state of scenario A as the ex-

perimental initial state. Most of the residences were

distributed on the periphery of CBD as sprawl, and

private automobile users reached close to 90%. This

shows the state after 20 loop processes of residential

relocation (corresponding to 40 years) from the zon-

ing between residences and job locations. The edge

SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems

104

Figure 3: Residences’ ﬁnal distribution of Experiment 1.

routes of tramway also pass through the suburb area

with sprawled residences.

The experiments were conducted under the condi-

tions of the two types for the location of PFS and the

four types for coefﬁcient of activeness, like the previ-

ous section. Hereinafter, each of these experiments is

expressed e.g., scenario SAt, SEt0 - 30, by combining

the initial letter S for the word of sprawl, the symbols

of A and E indicating the location of PFS, the initial

letter t for the word of tram, and the coefﬁcient of ac-

tiveness. Additionally, scenario SEt30+, which was

run for twice as long as SEt30, was executed.

5.2 Results of Experiment 2

Table 2 shows the quantitative result of scenario SAt,

SEt0 - 30, and SEt30+. Fig. 4 shows the ﬁnal distri-

butions of residences of the same scenarios.

The result of scenario SAt shows that private auto-

mobile users increased further, and the sprawl of their

residences on the periphery of CBD also advanced

further, unlike scenario At.

The results of scenario SE0 - 30 also show that

both the decrease in the private automobile users and

the cluster of residences of train users around the res-

idence station were not observed, unlike the series of

Et. Particularly in scenario SEt0 - 20, private auto-

mobile users increased further, and the sprawl also

advanced further, like scenario SAt. In advancing

the promotion of street activeness, however, private

automobile users decreased, and tramway (and train

in combination) users increased and reached close to

50% in total in scenario SEt30.

Furthermore, the results of scenario SEt30+,

where scenario SEt30 was run further, shows that

tramway (and train in combination) users reached

close to 90% in total. Along with this, the total CO

emission also reduced considerably, and the follow-

ing two clusters of residences were emerged. One

is the cluster by residents commuting by train and

tramway in combination (about 20%), on centering

the residence station. The other is the cluster by res-

idents commuting by tramway alone (about 70%),

along tramway routes from the center to the periph-

ery of CBD.

Table 2: Result of Experiment 2.

6 DISCUSSION

6.1 Estimation of Experimental Results

By combining the introduction of tramway with in-

troducing a public facility for stopping off (PFS) and

promoting street activeness around it, private automo-

bile users decreased, when compared with the cases

when no policies were implemented. And train (and

tramway in combination) users increased accordingly.

Along with this, the total CO

emission reduced, and

the compact urban structure, which was formed ac-

cording to the zoning, was maintained. These suggest

that the synergistic effects of the above-mentioned

policies could impact positively on a both static and

dynamic urban environment. This also seems to be

because the promotion of street activeness, which is

How to Realize a Compact City: Street Activeness and Agent-based Urban Modeling

105

Figure 4: Residences’ ﬁnal distribution of Experiment 2.

incentive to stroll about downtown, was effective to

increase tramway users.

On the other hand, where the initial state was

sprawl mainly with use of private automobiles, the

introduction of tramway could not serve to control

further sprawl and use of private automobiles. When

combined with the introduction of the public facility

and the promotion of street activeness, however, most

of private automobile users switched to tramway use,

although it took a long period. This suggests that once

residents established the lifestyle of low-density res-

idence in suburb and commuting by private automo-

bile, that becomes robust, irreversible, and very dif-

ﬁcult to be upset. As for the residence distribution

in the same scenario, most of the residences that are

distributed along the tramway routes deviated from

the initial poly-centric compact city. This, however,

can gain the following positive evaluations of a mono-

centric compact city. First, the residents can live

where residences and job locations are nearby based

mainly on use of public transportations, resulting in

being free from trafﬁc congestion and air pollution.

Second, mixed land-use provides the residents with a

broad range of social activities, while revitalizing the

central urban area.

Simply put, the promotion of street activeness is

a policy to lead people to walk by giving them in-

centives. On the other hand, many successful cases

of introducing tramway in the real world are charac-

terized with combining the introduction of tramway

with other policies which serve as a beneﬁt for peo-

ple traveling on foot. That is, this experiment clar-

iﬁed that the introduction of tramway can exert a

profound effect only when combined with policies,

which lead tramway users’ stroll before and after they

use a tramway, and how it can offer great beneﬁts.

6.2 Validation of the Simulation Model

Because of the property of emergence in complex

self-organizing systems, ABMs should be validated

in terms of whether it can capture the basic features

of the system in the real-world (Wu, 2002). Pattern-

Oriented Modeling (POM) procedure is an effective

validation procedure. In POM procedure, after iden-

tifying the observed patterns in the real-world charac-

terizing the system to be modeled, an ABM is evalu-

ated by whether the observed patterns are reproduced

(Railsback and Grimm, 2011).

First, the history of urban area in Japan for 40

years based on 1970 and the history of residential area

in scenario A for the ﬁrst experiment are very similar,

so this scenario can be regarded as reproducing the

fact of the constant expansion of urban area in many

cities in Japan (Eaton and Eckstein, 1997). Addition-

ally in this scenario, the residence distribution signif-

icantly changed from separation between residences

and job locations to sprawl on the periphery of CBD.

This can be regarded as the reproductionof the growth

process of a concentric low-density suburb based on

the mono-centric urban model proposed by Alonso

(1964) (Alonso, 1964), and subsequently supported

by many related researches. Next, the travel mode

used by most of the household agents has switched

from train to private automobiles. This can be also

regarded as the reproduction of the fact that the main

travel mode in commuting has switched from train to

private automobiles (MLIT, 2016).

Our highly abstracted urban dynamics model

based on a small number of elements and simple

rules reproduced the above multiple social phenom-

ena which were not directly incorporated into the

model. These reproductionsdemonstrate that the sim-

SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems

106

ulation model can explain the real society to a certain

level, and the experimental results of this research are

valid.

7 CONCLUSION

The purpose of this research was to verify the ef-

fectiveness of the combination of the introduction of

tramway with introducing a public facility for urban

residents and promoting street activeness around it,

on urban sprawl. An agent-based model (ABM) for

simulating the changes of urban structure through au-

tonomous behavior of urban residents was designed.

Then the simulations were conducted based on the as-

sumption of zoning between residences and job loca-

tions as the initial state and combining these policies.

These were followed by other simulations based on

the assumption of setting urban sprawl as the initial

state. As a result, this research clariﬁed the follow-

ing points and how they were. First, the synergistic

effects of the introduction of tramway, the proper lo-

cation of a public facility, and the promotion of street

activeness around it, are effective in maintaining a

poly-centric compact city in accordance with the ini-

tial plan. Second, the introduction of tramway tar-

geting the urban sprawl can exert a profound effect

only when combined with the above-mentioned poli-

cies, which lead tramway users’ stroll before and af-

ter riding the tramway, although it takes a long period.

Third, a mono-centric compact city is realized along

with the above-mentioned point, while improving the

living environment for the residents and revitalizing

the urban central area.

REFERENCES

MLIT (2016). White paper on land, infrastructure and trans-

port in 2016 - the ministry of land, infrastructure and

transport. http://www.mlit.go.jp/hakusyo/mlit/h28/

index.html.

Acheampong, R. A. and Silva, E. (2015). Land use–

transport interaction modeling: A review of the litera-

ture and future research directions. Journal of Trans-

port and Land Use, 8(3).

Alonso, W. (1964). Location and land use. toward a general

theory of land rent.

Batty, M. (2007). Cities and complexity: understand-

ing cities with cellular automata, agent-based models,

and fractals. The MIT press.

Behan, K., Maoh, H., and Kanaroglou, P. (2008). Smart

growth strategies, transportation and urban sprawl:

simulated futures for hamilton, ontario. The Canadian

Geographer/Le G´eographe canadien, 52(3):291–308.

De Bruijn, H. and Veeneman, W. (2009). Decision-making

for light rail. Transportation Research Part A: Policy

and Practice, 43(4):349–359.

Deal, B. and Schunk, D. (2004). Spatial dynamic model-

ing and urban land use transformation: a simulation

approach to assessing the costs of urban sprawl. Eco-

logical Economics, 51(1):79–95.

Eaton, J. and Eckstein, Z. (1997). Cities and growth: The-

ory and evidence from france and japan. Regional sci-

ence and urban Economics, 27(4-5):443–474.

Jacobs, J. (1969). The economy of cities. Vintage Books.

Johnson, M. P. (2001). Environmental impacts of ur-

ban sprawl: a survey of the literature and pro-

posed research agenda. Environment and planning A,

33(4):717–735.

Kakoi, M., Nakamura, R., and Saito, S. (2010). Causal

structure modeling of consumer’s decision-making on

selection of commercial facilities. Fukuoka University

economics review, 54(3-4):241–256.

Landry, C. (2012). The creative city: A toolkit for urban

innovators. Earthscan.

Ligtenberg, A., Bregt, A. K., and Van Lammeren, R. (2001).

Multi-actor-based land use modelling: spatial plan-

ning using agents. Landscape and urban planning,

56(1):21–33.

Nagai, H. and Kurahashi, S. (2017). Bustle changes the city

- facility for stopping off and modeling urban dynam-

ics. Transactions of the Japanese Society for Artiﬁcial

Intelligence, 32(1):D–G26

United Nations Department of Economic and Social Affairs

(2012). World urbanization prospects, the 2011 revi-

sion. In Final Report with Annex Tables. United Na-

tions, New York.

Railsback, S. F. and Grimm, V. (2011). Agent-based and

individual-based modeling: a practical introduction.

Princeton university press.

Ramming, M. S. (2001). Network knowledge and route

choice. Unpublished Ph. D. Thesis, Massachusetts In-

stitute of Technology.

Schneider, A. and Woodcock, C. E. (2008). Compact, dis-

persed, fragmented, extensive? a comparison of ur-

ban growth in twenty-ﬁve global cities using remotely

sensed data, pattern metrics and census information.

Urban Studies, 45(3):659–692.

Taniguchi, T. and Takahashi, Y. (2011). Multi-agent simu-

lation about urban dynamics based on a hypothetical

relationship between individuals’ travel behavior and

residential choice behavior. Transactions of the Soci-

ety of Instrument and Control Engineers, 47(11):571–

580.

Tsai, Y.-H. (2005). Quantifying urban form: compactness

versus’ sprawl’. Urban studies, 42(1):141–161.

Wu, F. (2002). Calibration of stochastic cellular automata:

the application to rural-urban land conversions. In-

ternational Journal of Geographical Information Sci-

ence, 16(8):795–818.

How to Realize a Compact City: Street Activeness and Agent-based Urban Modeling

107