Content uploaded by Faysal Kabir Shuvo
Author content
All content in this area was uploaded by Faysal Kabir Shuvo on Sep 02, 2020
Content may be subject to copyright.
MODELLING INFORMAL SETTLEMENTS USING A
HYBRID AUTOMATA APPROACH
Faysal Kabir SHUVO and Patrick JANSSEN
National University of Singapore, Singapore
fkshuvo@yahoo.com, patrick@janssen.name
Abstract. Automata based modeling of urban environments focuses on the
bottom up emergence of particular phenomena through interactions at the
disaggregated level. Two popular approaches are cellular automata (CA) for
modelling fixed automata and agent based modelling (ABM) for modelling
mobile automata. However, certain urban phenomena cannot easily be mod-
elled by either CA or ABM alone. This research focuses on one such
phenomenon, referred to as leapfrog development, which is a discontinuous
and dispersed type of urban growth. A hybrid automata model is proposed
for modelling such phenomena that combines both CA and ABM into a sin-
gle integrated model. The proposed model is demonstrated by applying it to
a case-study in Dhaka city for simulating the growth of informal settlements
associated with the readymade garment industry.
Keywords. Hybrid automata; GIS; agent based systems; cellular automata;
informal settlements.
1. Introduction
Automata-based modelling of urban environments is a dynamic, bottom up, and
disaggregated approach to modelling (Longley, 2005). Automata are the compo-
nents within the model that process information and change the state of the system
based on a set of transition rules (De Smith et al., 2007). Benenson and Torrens
(2004) has formulated the automata as
A= {S,T}; S= {S1, S2, S3, ..., Sn }; T: {St, It} →St+1, (1)
where T is the set of transition rules, St is the state of the automaton at t time, It is
the input information derived from neighboring automata, and St+1 is the state of
the automata process at time t+1.
Automata may be either fixed (such as a land parcel) or mobile (such as a
pedestrian) (Benenson and Torrens, 2004; Batty, 2007). Two popular approaches
R. Stouffs, P. Janssen, S. Roudavski, B. Tunçer (eds.), Open Systems: Proceedings of the 18th International
Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA 2013), 591–600. © 2013,
The Association for Computer-Aided Architectural Design Research in Asia (CAADRIA), Hong Kong, and
Center for Advanced Studies in Architecture (CASA), Department of Architecture-NUS, Singapore.
591
5B-205.qxd 4/29/2013 1:18 PM Page 591
to automata based modelling are cellular automata (CA) for fixed types of
automata and agent based modelling (ABM) for mobile types of automata.
This research focuses on the modelling of the growth of the informal settle-
ments within cities. Both CA and ABM models have been used for modelling this
type of growth. Sietchiping (2004) and Hill and Lindner (2010) have developed
CA models, while Barros (2004) and Vincent (2009) have developed ABM mod-
els. However, there is evidence that certain discontinuous and dispersed types of
urban development, called leapfrog development, play a significant role in infor-
mal settlement growth. The concept of leapfrog development emerged from
studies on the spatial evolution of urban areas (Benguigui et al., 2001). This type
of development is by definition influenced by both local and global dynamic fac-
tors. For example, industries providing particular types of job opportunities may
concentrate in certain areas of a city and influence the ‘leapfrog’ growth of infor-
mal settlement by pulling low-paid migrants worker away from outside of cities.
Through this ‘leapfrogging’, informal settlements develop at specific locations
within the city that are close to work locations and where there is favourable land
in combination with other informal mechanisms.
Modelling using either CA or ABM alone cannot easily capture the dynamics
of such leapfrog development since both fixed and mobile automata need to be
considered. From an urban modelling perspective, there is little research that
investigates both the local (at the settlement level) and global (on a city scale) pat-
tern of such leapfrog type informal settlement growth. For such types of growth,
it is argued that a hybrid model is required that combines CA and ABM.
In the field of urban or city modelling, the ‘hybrid’ term has been used at two
levels. At one level, there have been attempts to integrate micro/meso scale mod-
els with macro-level models, where micro/meso scale models are provided as
modular components for simulating various urban local dynamics (Torrens, 2001).
At another level, the idea of ‘hybrid’ models have been seen as integrating fixed
and mobile units of cities (Torrens and Nara, 2007; Wu and Silva, 2009).
However, these hybrid approaches appeared to be insufficient as they do not fully
integrate CA and ABM approaches into a single model.
The aim of this paper is to propose a hybrid automata model where both CA
and ABM approaches are tightly coupled in order to be able to model leapfrog
type informal settlement growth. The proposed hybrid model is applied to a case
study in Dhaka, the capital city of Bangladesh.
2. Case Study
Rural to urban migration has been the dominant component of urban population
growth (Islam 2006a; Alam and Rabbani 2007). The readymade garment (RMG)
industry in Dhaka makes a significant contribution to the country’s economic
592 F. K. SHUVO AND P. JANSSEN
5B-205.qxd 4/29/2013 1:18 PM Page 592
development, and the factories that have proliferated throughout the city have cre-
ated a strong ‘pull factor’ for internal migration (Hoque et al., 2007). An estimated
300,000–400,000 migrants, mostly poor, arrive at the city annually (Rashid, 2009;
World Bank, 2007). The workers in the RMG factories mostly come from this pool
of migrants (Islam, 2006b; Mottaleb and Sonobe, 2011).
Despite the socio-economic importance of the RMG industry, the settlement
issues of the factory workers have not been resolved, and as a result most work-
ers will end up living in informal settlements. These settlements consist mainly of
residential units made of fragile (kutcha) structure with one floor.
Unlike in other cities where informal settlements mainly develop at the urban
periphery (Barros, 2004; Sietchiping, 2004), in Dhaka informal settlements have
typically emerged within the city itself (Masum, 2009). Topological relations exist
between informal settlement clusters and factory clusters. Figure 1 shows scat-
tered pattern of informal settlements and factories.
The locations of the RMG factories strongly influence the location and pattern
of the informal settlements. A reverse relationship can also be identified, whereby
MODELLING INFORMAL SETTLEMENTS 593
Figure 1. Scattered pattern of IS and RMG factories (black circle is industrial zone).
5B-205.qxd 4/29/2013 1:18 PM Page 593
the location choice of factories is influenced by the locations of the informal set-
tlements. In addition, Hoque et al. (2007) have found that most of the factories
have been established besides main arterial roads and in buildings designed for
other uses such as residential or commercial buildings.
2.1. HYPOTHESIS AND PROPOSED MODEL
Whenever there is a clustering of RMG factories, a clustering of informal settle-
ments will also emerge. This clustering confluence leads to the development of
non-homogenous patches of informal settlements around the city.
Based on analysis of existing land use data of the study area, it is hypothesized
that a development pattern exists, whereby informal settlements at a particular
location will grow up to a certain size limited by the constraints of the site and then
leapfrog to a new location, where growth will again take place.
In order to test the validity of the leapfrog development hypothesis, a hybrid
automata based model is proposed. In the proposed model, the factories and the
workers are modelled as mobile automata, while the land units are modelled as
fixed automata. The CA and the ABM are tightly coupled in a cyclical loop. The
local suitability of parcels of land is modelled using CA while the location
choice behaviour of the factory and worker agents is modelled using ABM. In
each iteration of the loop, the CA is used to assign suitability scores for differ-
ent activities to each land parcel and the agents in the ABM then make location
choices based on these suitability scores. The conceptual framework is shown in
Figure 2.
In order to reduce the complexity of the modelling task, this paper will focus
only on the informal settlement dynamics (see the right hand side of Figure 2),
594 F. K. SHUVO AND P. JANSSEN
Figure 2. Conceptual framework of the hybrid automata model.2
5B-205.qxd 4/29/2013 1:18 PM Page 594
including a CA for assigning suitability scores to parcels of land, and an ABM for
modelling the worker location choices. RMG factories will therefore not be mod-
elled as automata, but will instead be defined as static entities based on the
time-series GIS data.
2.2. MODEL IMPLEMENTATION
For automata based models, time series data is required. From the detailed land-
use data for the study area (e.g. construction year), the relevant data are sliced
into three time series, for the years 1985, 1995, and 2005. The base year is
selected as 1985 as this is considered to be the year when the RMG started to
grow significantly.
The main aim is to demonstrate the validity of the leapfrog development pat-
tern using a hybrid automata model. A hybrid model will be developed, and a set
of transitional rules will be created for both the fixed and mobile automata. The
temporal resolution or time step is considered one year. Starting from the 1985
dataset, the model will be run for 20 time steps in order to get a pattern of infor-
mal settlements for the years 1995 and 2005. The output generated by the model
will then be compared to the actual datasets for 1995 and 2005.
The time-series data is created using ArcGIS, and the automata model is imple-
mented in NetLogo, an integrated automata modelling environment. The shape
file for the base year is first created in ArcGIS and then imported into NetLogo.
The area of study is represented using a square cellular grid with a spatial resolu-
tion of two meters, where each grid cell is described as a patch. Each patch is
assigned certain attributes that describe the patch. For example, a patch may be an
RMG factory, an informal settlement, a water body, and so forth.
For informal settlements, patches are defined as being either habitable or non-
habitable. For the 1985 data set, all patches in areas within the city that are vacant
are assigned as being habitable. A habitable patch can be in one of three states:
empty habitable patches have zero residents, occupied habitable patches have one
resident, and full habitable patches have two residents.
The transition rules for the CA and the ABM are then defined based on these
patches. The transition rules for the CA assign a suitability score to each habitable
patch, and the transition rules for the ABM define how agents choose which hab-
itable patch to settle on.
2.2.1. CA transition rules
The CA transition rules assign a suitability score to each habitable patch based on
the local factors that are assumed to have influence on the workers’ locational
MODELLING INFORMAL SETTLEMENTS 595
5B-205.qxd 4/29/2013 1:18 PM Page 595
preference. These factors include proximity to RMG factories, proximity to exist-
ing informal settlements, presence of vacant public land and so forth. These
factors are selected based on the analysis of the time series GIS data, field obser-
vation, and expert opinion.
The transition rules use two methods for describing the local conditions of
a particular patch: the neighbourhood method and the distance method. The
neighbourhood method examines the patches that are immediately adjacent
to the current patch. In this case, the von Neumann neighbourhood is used,
comprising the four patches orthogonally surrounding a central patch. The dis-
tance method examines patches further away, and uses a distance rule. The
scoring system uses a scale from 10 to 1, where 10 indicates that the patch is
highly suitable for informal settlement. The CA transition rules are shown in
Table 1.
The above rules only apply to habitable patches. However, over time, patches
may change their land use due to a variety of reasons. Such changes are land use
type can be handled by special transition rules, referred to as transformation
rules. In this version of the model, one set of transformation rules have been
implemented to account for the phenomena whereby workers will create informal
settlements directly adjacent to or over water bodies. Patches adjacent to or over
water bodies can become habitable if all other habitable patches in the vicinity
are full. Under these conditions, the pressure is deemed sufficiently high to result
in informal settlements encroaching onto the water bodies. In future, transforma-
tion rules for other types of phenomena will be implemented, such as the
596 F. K. SHUVO AND P. JANSSEN
Table 1. CA Transition rules for habitable patches.
Rule Condition: Rule Condition: Rule Condition: Rule Result:
RMG factories Informal settlement Waterbody Score
Distance < 100 meter Within neighborhood Within neighborhood 10
Distance < 100 meter Within neighborhood 9
Distance < 500 meter Within neighborhood Within neighborhood 8
Distance < 100 meter Within neighborhood 7
Distance < 500 meter Within neighborhood 6
Distance < 500 meter Within neighborhood 5
Distance > 500 meter Within neighborhood Within neighborhood 4
Distance > 500 meter Within neighborhood 3
Distance > 500 meter Within neighborhood 2
Within neighborhood 1
5B-205.qxd 4/29/2013 1:18 PM Page 596
transformation of habitable patches to non-habitable due to the infrastructure
construction projects such as road systems.
2.2.2. ABM transition rules
An ABM has been implemented to model the location choice of worker agents.
For this demonstration, a very simple model is developed. In future, it is planned
to elaborate this model in more detail.
The main source of settlement growth is due to new immigrants arriving in the
city. Therefore, at each time step, a fixed number of worker agents are added to
the city, and randomly assigned to work in the RMG factories in existence at that
point in time. Though the annual immigrants are about 100,000, in this model the
number of worker agents being added to the city is set to 2000 per year.
Hypothetically, each working agent is considered to be equivalent to 5 persons in
reality due to the system limitations of NetLogo.
Once the worker argents are assigned to factories, they then need to choose
where they are going to settle. To facilitate this choice, the vacant and occupied
habitable patches within a certain radius of the factory are first sorted according
to their suitability score (as assigned by the CA). Each worker agent then selects
the highest scoring available patch. Once they settle on a patch, the number of res-
idents residing on that patch will be incremented by one. (Vacant patches become
occupied, and occupied patches become full).
2.3. MODEL RESULTS
The hybrid model was run for a period of 20 years in yearly time steps, using the
1985 data set as a starting point. At five yearly intervals, visual maps of the sim-
ulation results were generated that show the extents of informal settlement growth.
These simulated maps allow the gradual growth of the settlement to be better
understood. The maps are shown in Figure 3.
For the final simulated 2005 map, the results are visually compared to the 2005
map showing the actual observed condition. It is found that there is significant
similarity between the observed map and the predicted map.
Analysing the time series data in more detail reveals that older informal settle-
ments tend to grow larger before new informal settlements appear at other
locations. The simulated model produces similar patterns of growth. However, as
a result of the input parameters, important differences between simulated and
actual conditions can be seen. These differences are mainly due to the parameters
for the ABM, such as the number of new workers per year entering the city. Future
MODELLING INFORMAL SETTLEMENTS 597
5B-205.qxd 4/29/2013 1:18 PM Page 597
research will analyse historical and current data in order to define appropriate
settings for these parameters.
3. Future Work
A hybrid model for leapfrog type urban growth has been proposed that tightly
couples CA and ABM, thereby allowing both fixed and mobile automata
598 F. K. SHUVO AND P. JANSSEN
Figure 3. Comparison between the simulated and observed (last image) informal settlement
patterns.
5B-205.qxd 4/29/2013 1:18 PM Page 598
approaches to be used. In the hybrid mode, the area of investigation is first split
into a series of small land patches. A CA is used to model suitability of patches for
certain activities, and an ABM is used to model location choices for agents engag-
ing in these activities. The CA and ABM models are executed within a cyclical
loop and affect one another through the data associated with the patches. The suit-
ability scores generated by the CA will be used by the ABM, while the location
choices generated by the ABM will be used by the CA.
The proposed hybrid model has been demonstrated by applying it to a case-
study in Dhaka city for simulating the growth of informal settlements associated
with RMG industries. Visual comparison of the actual and predicted maps showed
significant similarity. These initial results suggest that the proposed hybrid model
may be of value for simulating such leapfrog type urban growth.
The proposed hybrid model is in its early stages of development. Various types
of sensitivity tests and validation processes are yet to be performed. Athree phased
validation process is proposed for this model, borrowed from Hill and Lindner
(2010).
Acknowledgements
The authors are greatly thankful to NetLogo users community especially James F. Steiner, who vol-
untarily help on coding for the development of simulation. Also the authors would like to thank
RAJUK (capital improvement authority of Dhaka) for providing GIS data of the Detailed Area Plan
(DAP).
References
Alam, M. and Rabbani, M.: 2007, Vulnerabilities and responses to climate change for Dhaka.
Environment and Urbanization, 19, 81–90.
Ahmed, S. J., Bramley, G. and Dewan, A. M.: 2012, Exploratory growth analysis of a megacity
through different spatial metrics: A case study on Dhaka, Bangladesh (1960-2005), URISA,
24(1), 9–24.
Barros, J.: 2004, Urban Growth in Latin American Cities: Exploring urban dynamics through agent-
based simulation, PhD Dissertation, Bartlett School of Graduate Studies of Architecture and
Planning.
Batty, M.: 2007, Cities and complexity: understanding cities with cellular automata, agent-based
models, and fractals, The MIT press.
Benenson, I. and Torrens, P. M.: 2004, Geosimulation: Automata-based modeling of urban phenom-
ena, Wiley.
Benguigui, L., Czamanski, D. and Marinov, M.: 2001, City growth as a leap-frogging process: an
application to the Tel-Aviv metropolis, Urban Studies, 38(10), 1819–1839.
De Smith, M. J., Goodchild, M. F. and Longley, P.: 2007, Geospatial analysis: a comprehensive
guide to principles, techniques and software tools, Troubador Publishing.
Hill, A. and Lindner, C.: 2010, Modelling informal urban growth under rapid urbanization: A CA
based land-use simulation model for the city of Dar es Salam, Tanzania, PhD Dissertation, TU
Dortmund.
MODELLING INFORMAL SETTLEMENTS 599
5B-205.qxd 4/29/2013 1:18 PM Page 599
Hoque, D. M. S., Debnath, A. K. and Mahmud, S. M. S.: 2007, Impact of Garment Industries on
Road Safety in Metropolitan Dhaka, Proceedings of the Eastern Asia Society for Transportation
Studies, 2007.
Islam, N.: 2006a, Bangladesh, in B. Roberts and T. Kanaley (eds.), Urbanization and sustainability
in Asia: Case studies of good practice, Asian Development Bank.
Islam, N.: 2006b, Slums of urban Bangladesh: mapping and census, 2005, United States Agency for
International Development report.
Longley, P.: 2005, Geographic information systems and science, John Wiley.
Masum, F.: 2009, Urban Fringe Management and Role of Good Governance: Integrating
Stakeholders in Land Management Process, 7th FIG Regional Conference, Hanoi, Vietnam.
Mottaleb, K. A. and Sonobe, T.: 2011, An Inquiry into the Rapid Growth of the Garment Industry in
Bangladesh, Economic Development and Cultural Change, 60(1), 67–89.
Rashid, S. F.: 2009, Strategies to reduce exclusion among populations living in urban slum settle-
ments in Bangladesh, Journal of health, population, and nutrition, 27(4), 574–586.
Sietchiping, R.: 2004, A geographic information systems and cellular automata-based model of
informal settlement growth, PhD Dissertation, The University of Melbourne.
Torrens, P. M.: 2001, A hybrid geocomputation model for operational land-use and transport simu-
lation, Annual Association of American Geographers Meeting, New York.
Torrens, P. M. and Nara, A.: 2007, Modeling gentrification dynamics: A hybrid approach,
Computers, environment and urban systems, 31(3), 337–361.
Vincent, O. O.: 2009, Exploring spatial growth pattern of informal settlements through agent based
simulation. MSc thesis, Utretch University; Delft University of Technology; Wageningen
University; International Information for Geo-Information Science and Earth Observation.
World Bank: 2007, “Bangladesh-Dhaka: Improving Living Conditions for the Urban Poor”.
Available from: <https://openknowledge.worldbank.org/handle/10986/7686> (accessed on 6
December 2012).
Wu, N. and Silva, E. A.: 2009, “Integration of genetic agents and cellular automata for dynamic
urban growth modelling”. Available from: <http://www.biodiverse.unsw.edu.au/geocomputa-
tion/proceedings/PDF/Wu_and_Silva.pdf> (accessed on 6 December 2012).
600 F. K. SHUVO AND P. JANSSEN
5B-205.qxd 4/29/2013 1:18 PM Page 600
