Content uploaded by Gonzalo Bustos-Turu
Author content
All content in this area was uploaded by Gonzalo Bustos-Turu on Apr 24, 2015
Content may be subject to copyright.
BOTH Agent-based simulation as an urban design tool
– Iterative evaluation of a smart city masterplan
Koen H. van Dam
1
, Dietmar Koering
2
, Gonzalo Bustos-Turu
1
and Henry Jones
3
1
Imperial College London, Dept of Chem Eng, London SW7 2AZ, UK, {k.van-dam, gonzalo.bustos-turu12}@ic.ac.uk
2
Technical University Berlin, IFA, Strasse des 17. Juni 152, 10623 Berlin, Germany, dietmar.koering@tu-berlin.de
3
CHORA, 24a Bartholomew Villas, London NW5 2LL, UK, hjones@chora.org
ABSTRACT
Urban planners designing masterplans for smart neighbourhoods
and smart cities face many challenges dealing with complexity
and requiring expertise in various domains at the same time.
Simulation models could be used to offer decision support and a
tool to evaluate different scenarios and designs. This way the
masterplans move from being a static design to an interactive and
dynamic test bed. Agent-based modelling has been used to create
a synthetic population for a new urban area which can be linked
directly with GIS based designs. In two workshops this idea has
been tested through the iterative evaluation of masterplans created
by architecture students for the Old Oak Common site in London.
Here cases on mobility and land values are shown as illustrative
examples. The final discussion focuses on city data and the impact
of simulation on the work of architects, as well as collaborative
decision making and consensus building using simulation tools.
Categories and Subject Descriptors
I.6.3 [Simulation and Modeling]: Applications
General Terms
Design, Experimentation, Human Factors
Keywords
Smart city, agent-based modelling, urban planning, learning from
feedback, data, behavioural models, dynamic masterplanning
1. INTRODUCTION AND CONTEXT
Urban environments are designed by the integration of multiple
levels and the computer is more and more imbedded into this
process not only as a design tool but also to evaluate the design
with local circumstances, such as wind, shadows, energy or
airflows. The current urban design debate suffers from allowing
unexpected incidents and unplanned developments to be accepted
as part of a process-driven urban production. Smartness in
architecture is nowadays achieved by data which oscillates
between human and machine. This leads to a new intrinsic and
immersive world of complexity in the field of contemporary
architecture [1]. Therefore, inspiring projects and concepts are
demanded which can be highly speculative and provide a planning
methodology, which can act and react to any unforeseen events
and human failure. This is also a challenge for educating a new
generation of architects and urban planners.
The key in contemporary urban design is handling the complexity
of a city whilst maintaining an understanding of running systems
and related data. Cities are becoming more sensory driven where
the generated data feeds into databases which could power a new
planning device: “Can a technological or abstract understanding
of these devices and their construction influence and redefine the
potential for architecture and spatial thinking?” [2].
Agent -based modelling can provide the operational mechanism to
run such a planning device through simulation of socio-technical
systems [3,4] and such models are deployed in spatial planning
and architecture already [5].Simulation is used to learn how cities
work and “most of these models have been focused on
understanding as a prelude to their use to inform the planning and
design process.” [6].
Much of this computational feedback is based on programs where
architects have to trust the output without being able to validate
the data or model. This problem can be tackled by the creation of
multidisciplinary design teams including architects, chemistry,
computer and social sciences, etc. But how do we overcome the
complexity of the different views? Data simulation and design can
be used as decision support tool and inform the participatory and
collaborative design process [3]. Such output is not final, but it is
an iterative design process, like a “fly-by-wire” system [7] for
architects where the designer is supported and guided by digital
tools.
In this paper we show how an agent-based simulation model
based on the spatial description of an urban layout is used to
evaluate a masterplan and provide decision support to the urban
planner for smart city projects. With this tool the analysis of city
operation and evolution in a context of participatory design and
planning context of future smart cities can be done. Bringing
together different stakeholders (urban planners, architects,
engineers, developers, etc.), different scenarios are created, tested
and re-designed in an iterative process, using the results from this
simulation tool as a feedback for urban design. This approach is
illustrated with the outcomes of two smart city modeling
workshops giving students hands-on experience with this process.
2. LINKING GIS AND AGENT-BASED
SIMULATION MODELS
In our approach we connect the spatial description of the
masterplan, which is annotated in GIS to add semantics to the
lines in the drawing, with an agent-based simulation of a synthetic
population of this area. Figure 1 shows the iterative process of
design-simulation-redesign as part of a collaborative activity.
First, the different urban designs with their corresponding land use
distribution, city layout, road network, etc. are created in GIS
software. Next, the agent-based simulation model (based on [8])
uses this map as input to generate a synthetic population following
certain socio-demographic parameters. This population of agents
uses behavioural rules related with their activities (e.g. going to
work or going to the shops) leading to transport, energy and
resource consumption for the given urban layout. The ability to
add complex behaviour is what sets this approach apart: by
describing the actions of individuals the system behaviour can
emerge, making it possible to experiment with different responses
to interventions and smart city services. Thus different urban
designs can be simulated and evaluated in terms of energy,
transport and economic performance, etc. The results of the
simulation are visualized using the same GIS software for further
analysis and redesign, after which any changes could be re-
evaluated. The agent-based simulation is implemented in Repast
Simphony while the spatial design and visualization uses QGIS.
Both tools are free and open-source software.
3. CASE STUDY AND WORKSHOPS
In 2014 two collaborative workshops were organized between TU
Berlin, London Metropolitan University and Imperial College
London in which students could experience this cycle in a multi-
disciplinary team and test various key performance indicators
(KPIs) such as land values, energy use and movement of people
for their masterplan. The workshop was integrated in the “Smart
City London” studio and the students could then inform their
designs with simulated data. All groups used the same ABM
simulation to evaluate their plans, but had different objectives so
each produced new inputs using the GIS software.
3.1 Case: Old Oak Common
Old Oak Common (OOC) is an area in North-West London,
largely a brownfield site and especially after a new proposed high
speed rail link and transport hub one of Britain’s best-connected
places [9]. While it is not certain this project will go ahead, this
plan can be seen as a case project of new infrastructure as a
catalyst for urban innovation and the creation of a smart
neighbourhood. The task for the students was to design a smart
district and prototype architecture for the OOC redevelopment.
3.2 Example domain: transport
One domain that was analysed during the workshop and would be
the core component for one masterplan was transport. This being
a brownfield site surrounded by train tracks and main roads there
is a need to re-think the way people can move to, from and
through the new neighbourhood to access jobs, amenities and
transport nodes. The students proposed an analysis of transport
demand with possibility of free movement using a grid-like
infrastructure to determine which paths are most-travelled given
the proposed land use. Figure 2 shows a screenshot of a
simulation run and Figure 3 then shows a visualisation produced
using the number of people who travelled on each link in the grid
as a KPI. The output inspired a new public transport network as
well as the location for nodes such as bus stops, and influence the
location of key services within the smart neighbourhood.
3.3 Example domain: land value
To discover new emerging financial patterns for the growth of a
smart “on demand” city, KPIs were selected to generate values
from. These values were assigned to the existing QGIS model to
see how values of the different masterplans for OOC change over
the course of one day, based on the agent activities. In the model
financial transactions were simulated as an effect of activities and
building use. The data process determined individual values of
income and expenditure per hour for each building and informed
how much projected money they will then gain or lose over the
course of an hour. This data fed the software to observe
relationships of change over longer periods of time.
The data simulation model of OOC (See Figures 4 and 5)
visualizes the change in value over the course of a year to draw
direct conclusions from. In conclusion, only 2 out of the 4 main
typologies studied and analyzed show any real change in value
over the course of 1 year. These are significant though and allow
for a new architectural playground of change to be created.
Developers can use these tests to input money into areas which
produce the most return but also add another socially valuable
layer to the existing cityscape so that the local businesses and
employees do not suffer but in fact benefit from the new wealth of
property investment.
4. DISCUSSION
We provided an overview of challenges with planning smart city
projects, dealing with complexity and requiring expertise in
various domains at the same time. Simulation models can offer
decision support and provide a tool with which different scenarios
and designs can be compared and evaluated in an iterative and
collaborative fashion with a team of experts from different
disciplines to address the socio-technical nature of cities.
There are key challenges in building and connecting various
specialist models, e.g. on transport, healthcare and economics to
give a multi-KPI evaluation of a masterplan in a coherent fashion.
Such models require teamwork and collaboration between experts,
as the students in these workshops could experience. This also
requires data collection in cities with the added issues of
uncertainty, data quality, trust, privacy, etc. Thus there is a need to
standardise the simulation and data-collection processes, so that
the results of the urban planning tests can be verified. Incubation
sites are necessary to create these standards.
As the computation power increases, we can conclude that in the
near future the iterative evaluation of a masterplan, through the
use of agent based modelling, could happen in real-time as part of
the normal workflow. This is where the real power lies, as
complex planning and design decisions can be tested and verified
instantly by simulating the real impacts that a new urban proposal
will have on a city, creating added efficiency to the planning
process. We will need simulation performance spaces to make use
of these new tools, such as the BrainBox urban lab that is being
developed by CHORA and TU Berlin [10][11], getting closer to
the idea of “city planning as a valuable creative activity in which
many members of a community can have the opportunity of
participating” [3].
From the perspective of the urban planner and architect, this
simulation-based approach has the potential to cause a major shift
in the planning process. As cities become 'smarter' on the ground,
so too do the planning and design development processes that lead
up to construction. There will also be a shift in professional
practice, as an increasing breadth of professions will become
closer and more collaborative. Urban planners and architects must
step up to this, by adapting their traditional role as the overall lead
of a project; they must become knowledgeable of the experts
fields associated with simulation techniques to be able to engage
with them as a design tool.
The workshops were only a few days each so time constraints
made it difficult to completely redesign the proposals, but the
final project presentations showed how the design was influenced
by the possibility of simulation and dynamic evaluation.
Transferring the hard-coded software to an easy-to-use interface
for designers to use will be an invaluable step, and a potential
business model for commercialisation. Future work includes
providing an automated link between QGIS and Repast so
masterplans can be more easily evaluated without the need for
modelling experts and programmers to be present. The aim is to
further test multi-disciplinary data and simulation-driven smart
city case studies in research as well as design projects.
5. ACKNOWLEDGEMENTS
The authors would like to thank the students from London
Metropolitan University and TU Berlin who participated in these
workshops for their input, the additional tutors Tomaz Pipan, Dirk
Lellau and Holger Prang and last but not least Profs Raoul
Bunschoten and Nilay Shah for their support.
6. REFERENCES
[1] Koering, D., Schulz, J., Hanses, K. & Siegemund, J. (Eds.),
Smart City Concepts: Konzepte für den energetischen
Stadtumbau, Avedition, 2013
[2] Lim, C.J. Devices, Architectural Press, 2006
[3] Simon, H.A. The Sciences of the Artificial, third edition ,
MIT press, 1996
[4] van Dam, K.H., Nikolic, I. & Lukszo, Z. (Eds.) Agent-Based
Modelling of Socio-Technical Systems, Springer, 2013
[5] Heppenstall, A.J., Crooks, A.T., See, L.M. and Batty, M.
(Eds.) Agent-Based Models of Geographical Systems,
Springer, 2012
[6] M. Batty et al, Smart Cities of the Future , UCL Working
Paper Series Paper 188 –Oct 2012, ISSN 1467-1298
[7] Philippe, C. Digital Fly-by-Wire Technology, in Samad, T.
and Annaswamy, A.M. (eds.) The Impact of Control
Technology, IEEE, 2011
[8] Bustos-Turu, G., van Dam, K.H., Acha, S. & Shah, N. A
systems approach to understanding the impact of electric
vehicle charging infrastructure on grid services. Energy
Systems Conference. London, UK, June 2014.
[9] Greater London Authority, Old Oak - A Vision for the Future
URL:http://www.london.gov.uk/sites/default/files/Old_Oak_
Part_1_0.pdf, June 2013
[10] TU Berlin Urban Lab. URL: www.smartcity.tu-berlin.de/tu-
berlin-smart-city-platform
[11] Brainbox – CHORA website URL: http://chora.org/?p=123
Figure 1. Iterative process of design-simulation for smart
cities: an initial design is evaluated using a simulation to give
insight in various key performance indicators, which can lead
to a re-design and if required further analysis
Figure 2. Screenshot of the agent-based simulation run to
analyse need for potential transport connections using a grid.
Stars represent agents moving around the area and the blocks
are building zones (e.g. residential or commercial)
Figure 3. Visualisation of most travelled paths used as input
for the design phase
Figure 4. Screenshot of QGIS showing value ranges for
various building types in the masterplan, after simulation
Figure 5. New design influenced by land value changes from
the agent-based simulation
