Content uploaded by Yang Yang
Author content
All content in this area was uploaded by Yang Yang on May 26, 2020
Content may be subject to copyright.
SimAUD 2020 May 25-27, Online
© 2020 Society for Modeling & Simulation International (SCS)
An Adaptive Workflow to Generate Street Network and
Amenity Allocation for Walkable Neighborhood Design
Yang Yang1, Samitha Samaranayake2, Timur Dogan3
1Cornell University
Ithaca, NY, USA
yy848@cornell.edu
2 Cornell University
Ithaca, NY, USA
samitha@cornell.edu
3 Cornell University
Ithaca, NY, USA
tkdogan@cornell.edu
ABSTRACT
This paper proposes a novel generative workflow for
walkable neighborhood design. Key components of the
workflow include automating the process of parsing the
map data, building contextual models with population and
amenity data, conducting an integrated mobility simulation,
and generating a street network and amenity allocation plan
accordingly. The proposed framework is versatile and
adaptive by allowing designers to tune simulation
parameters and customize the decision-making process. The
applicability and effectiveness of the workflow are tested in
a pedestrian-oriented neighborhood design case study.
Three scenarios that adapt to different design goals and
boundary conditions are presented. This research equips
designers with capabilities to co-design of mobility
solutions and urban form early on in the design process.
Further, it can be leveraged by more stakeholders in sectors
such as real estate, public services, and public health to
make decisions as the urban built environment has a
fundamental impact on all these fields.
Author Keywords
generative urban design; walkability; mobility; simulation.
1 INTRODUCTION
Population growth, urbanization and ever-increasing
vehicle use in urban areas have a significant impact on the
quality of life and the environment. Increasing traffic-
related energy consumption, greenhouse gas emissions, air
and noise pollution, as well as lifestyle-related health issues
such as obesity and diabetes can be promoted by poor urban
design [1]. While these are worrisome circumstances, the
need for urban renewal and densification [2] also provides a
unique opportunity to rethink planning paradigms and
design approaches. Emerging design movements aim to
remedy the aforementioned mentioned issues. They [3]
promote high density, walkable neighborhoods as one
solution for these challenges. Studies have shown that
walkable neighborhoods can significantly reduce traffic-
related pollution and lower the risk for chronic diseases
[3,4], support local businesses, promote tourism, attract
investors, higher property values [6] and foster an increase
in social capital and political participation [7]. Walkable
amenities, one of the most important ingredients of a
walkable city, have also been associated with
socioeconomic growth [7,8] and quality of life [10].
Understanding the implications of urban design choices on
walkability while incorporating this understanding into
early stages of urban design process provides a unique
opportunity to address these issues. This is particularly
important because street grids hardly ever change once the
urban design is set [11].
One of the major challenges in designing walkable
neighborhoods is the lack of effective metrics and
workflows that can provide measurable and actionable
feedback to facilitate design decision making. To evaluate
the walkability of cities, researchers have proposed to rank
neighborhoods based on the distance and density analysis of
points of interest (POI) in the city. These walkability
ratings, commonly referred to as Walkscore [12], are
computed on a scale of 1-100 and include factors such as
accessibility to amenities like grocery stores, restaurants,
banks, parks, and schools. Generative design workflows for
walkable neighborhoods that leverage this metric have been
developed [13]. They usually first generate an urban layout
by spatial logics and then optimize the Walkscore through
an evolutionary process that places additional amenities
until a sufficiently high score is reached. However, the
main question regarding the use of Walkscore as a sole
metric in such workflow is its insensitivity to the interactive
relationship between key urban design parameters including
street network, amenity allocation, and population
distribution in the model. This can result in several
questionable design decisions: Firstly, generating a street
network without considering amenity placement or
population distribution can be problematic because the
latter two factors significantly influence street utilization
and therefore play important roles in urban morphology.
Secondly, placing or adding services and amenities only to
drive up Walkscore may not be feasible as those new
services may not be sustainable as demand is spread too
thinly. Thirdly, the amounts and categories of amenities to
which it is essential to have walking access differ by
population groups. Thus, designers should be able to
evaluate walkability with demographic-specific metrics. As
a result, it is imperative to propose a more integrated
195
generative urban design workflows that can take into
consideration all the mentioned factors.
Additionally, common generative design workflows rely on
optimization solvers such as the Genetic Algorithm, which
searches for optimal solution evolutionarily based on each
iteration’s performances on certain metrics. Although such
a strategy is widely used in many generative urban design
studies [13,14], it remains questionable in terms of speed of
convergence and stability [15]. Although such a trial &
error approach is unavoidable for certain ill-defined design
problems, this paper proposes a simplified, efficient and
transparent approach using the Urbano toolkit [16]. In this
workflow, the first step is to build a mobility model with
urban data such as streets, buildings, amenities and
population densities. Then the potential pedestrian volumes
of streets and amenities are evaluated by a mobility
simulation. The simulation outcome can directly inform the
generative process of the street network and amenity
allocation. Throughout this process, designers can
customize the model and control the simulation by tuning
key parameters and changing variable constraints so that
different design conditions can be accommodated.
Overall, this paper introduces a novel generative urban
design workflow that is sensitive to street networks,
amenity allocation, and population distribution. The
workflow is implemented in a pedestrian-oriented
neighborhood design case study, and its adaptivity is tested
by accomplishing different design requirements.
2 METHODS
Urbano allows designers to build mobility models, run the
network and amenity analyses within the Rhinoceros CAD
platform and the visual scripting environment Grasshopper.
The automated modeling and simulation process is used to
drive the generative processes described in this paper.
2.1 Data-Driven Modeling
There are three layers of data that are necessary for the
mobility model: Street network, amenities (points of
interest), and buildings with building-level population
information. Knowing the location and boundary of the site,
Urbano can import streets, points of interest (POIs) and
buildings, along with their metadata, from sources such as
shapefiles (shp), OpenStreetMap (osm) [17] or Google
Places API [18]. Streets, buildings, and amenities are
represented by geometric primitives such as curves or
points. Metadata such as names, types, and addresses are
attached to the geometric data using serializable
dictionaries that can be modified and customized alongside
the geometric objects parametrically within Grasshopper or
through the CAD user interface in Rhino. If required
information, such as building-level population, is not
accessible from the sources, Urbano provides functions that
can infer data or can help to synthesize this information
using other data sources. For example, it can estimate
building-level population size using total building floor
area, customized area usage breakdown, and generalized
occupant densities [16,17].
2.2 Trip-Sending Simulation
The simulation framework is initially driven by the Activity
Demand Profile (ADP). The ADP describes pedestrian
activities over time and can be adapted to reflect activities
of specific demographics. One way to derive location-
specific ADP is to interpret the spatiotemporal distribution
of human activities in a local area by measuring the
activeness in urban amenities in this area [21]. The main
data source for this method is Google Places “Popular
Times” data. Table 1 shows a sample set of the integrated
ADP which presents the hourly percentage of population
that engages in particular activities in one day in the case
study area (Figure 5). Each column represents a one-hour
time slot in a day, which can be further synthesized into
time periods such as morning, noon and evening. Figure 1
is a graph for ADP data using a 24-hour timeline, which
depicts a more detailed activity distribution. The y-axis
refers to the overall amount of activities, which peaks
during the day and dips in the early morning. Each color
layer represents the demand pattern of an amenity. For
example, banks and post offices tend to stop service in the
early afternoon, while bars and pubs become dominant
activities at midnight. Nevertheless, ADP data can also be
customized by the designer to target an assumed
demographic group. In the simulation, one or multiple sets
of ADP can be utilized to represent different human activity
patterns coexisting in the area.
Figure 1. 24-hour timeline representing Activity Demand Profile.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
errands
0
0
0
0
0
0
0
0
0.4
1.1
1.3
1.5
1.6
1.8
1.8
1.8
1.6
1.1
0.3
0.2
0.2
0.2
0
0
restaurant
1.1
0.1
0
0
0
0.4
1.7
5.5
8.3
9.1
9.9
13.3
19.6
22.1
19.5
15.8
14.2
16.8
20.6
21.8
19.4
14
8.2
3.3
grocery
1.8
0.9
0.7
0.5
0.7
0.8
1.6
3.1
5.7
7.6
9.4
11.2
12.9
13.8
14.2
14.1
14.5
14.9
14.3
12.3
9.7
7.3
5
3.2
shopping
0
0
0
0
0
0
0
0
0.8
3.3
7.1
10.7
12.8
14.9
15.1
15
16.1
17.5
16.2
10.7
4.4
1
0.4
0
entertainment
6.2
4
3
1
0
0
0
0
0
0.3
0.4
2.7
5.3
5.7
7.2
7.9
11.6
17.2
24.8
31.6
29.1
25.8
21.8
17
Table 1. Sample of Activity Demand Profile data for the study area. (unit: %)
196
Figure 2. Diagram of the trip-sending simulation algorithm.
The following simulation is based on a trip-sending
process, and the concept of each trip is made up of multiple
information: the origin building, the destination amenity,
the route taken and the corresponding population. The
executed trip-sending algorithm is as follows (Figure 2):
1. Input model consisting of streets, amenities, and
buildings; one or multiple ADPs with same time metrics
2. For each time step of ADPs
3. Initialize empty list of trips
4. For each building in the model
5. Set the building as origin and get its population size
6. Divide the population into activities according to
the percentage data in its ADP
7. For each activity
8. Search for corresponding amenities as
destinations within walking distance (user-
defined) using a shortest-path algorithm
influenced by biased routing factors
9. Distribute the activity population to
destinations according to biased destination
factors
10. Generate trips with the distributed population
and add the trips to the list
11. Output a data tree of trips grouped by time steps
The biased factors are inputs allowing for more control by
designers. There are currently two types of them in the
presented algorithm. The first one is the biased destination
factor, defined as the weight of a destination which
determines the proportion of the population sent to them.
The higher the weight, the more proportion of the total
accessible population the amenity can receive. This factor
can be set according to the quantifiable quality parameters
of amenities such as the capacity, popularity or rating. The
other one is the biased routing factor, defined as the
coefficient of calculated street length. This coefficient
allows certain street segments to be “shortened” in the
simulation so that they can be more utilized in the shortest-
path routing process, or “lengthened” in the opposite way.
By modifying this coefficient, the simulation can count into
more factors influencing the route choice other than
distance, such as shade, landscape, facades, urban
environment, etc.
Output: Each time step has a distinct set of resulting trips.
Trips data can be post-processed into three complementary
metrics: Street Hits and Amenity Hits. Street Hits counts
how many people use a certain street segment on all trips.
Amenity Hits tallies up the total number of people that are
sent to a specific amenity on all trips. Moreover, a building-
level Walkscore can also be computed according to its
original method [12] using data of all trips that originate
from a single building. All these metrics will inform the
generative design process.
2.3 Generative Process
The primary setup for the generative method is to replace
the original design site with a dense grid mimicking a
virtual environment that lets people cross freely. This step
parallelly establishes a perpendicular coordinate grid
representing all potential locations for amenities (Figure 3).
Street Hits results from the simulation on this dense grid
can reveal people’s potential movement trails across the
site. Street segments with high Hits can be transformed into
new roads in the designed network. Amenity Hits results on
the coordinate grid can identify the most profitable
locations for amenities that grant most walking access to
the population in the model, which can inform the
placement of new amenities.
Figure 3. Using part of the case study’s site as an example (a), the
primary setup is to replace the original site with a dense grid (b)
mimicking a virtual environment that lets people cross freely. It
parallelly establishes a coordinate grid (c) representing all potential
locations for amenities
197
Figure 4. Two series of visualizations of Street Hits output. The first row (a) shows evolving results when increasing the user-defined walking
distance. The second row (b) presents changing results when placing new amenities on the design site.
Due to the nature of the trip-sending algorithm, the outcome
spontaneously concerns the interactivity between streets,
amenities and population distribution in the model. it is also
responsive to changes. Figure 4 shows the example of two
series of visualizations of Street Hits output from the same
simulation (Figure 3) but only with a specific parameter
modified. The first row shows when increasing the user-
defined walking distance, more trips are generated because
more amenities become accessible to all buildings within
that distance. The second row presents changing results
when placing new amenities on the design site. Besides
these, other parameters such as the biased routing and
destination factors, or ADPs can all impact the generative
process. A highly customizable framework like this enables
designers to tune the generative process for specific
conditions or goals in design practice as in the case study.
3 CASE STUDY
The site in Figure 5 is located in New Haven, Connecticut.
It has residential neighborhoods to the South, a high-density
commercial district to the North, institutions to the West,
and an industrial area to the East. A highway and a railway
adjacent to the East and the North form obstacles isolating
the site. The current street network does not connect urban
amenities well as Street Hits analysis on the original site
reveals that most activities do not take routes across the site
(Figure 6). The site is predominantly used as parking lots
and is considered as an empty area as the initial condition in
this study. However, the site has great potential as it can
connect the railway station to the city and fill the gap in
pedestrian mobility between different urban areas
surrounding it.
The initial site model consists of existing streets, amenities,
and buildings. Building types are categorized into
residential and non-residential so that population data can
be synthesized accordingly. The overall design objective is
to develop a new mixed-use and pedestrian-oriented
neighborhood that can alleviate some of the described
connectivity issues. All the existing streets and buildings on
the site are supposed to be overridden.
Figure 5. The site and the main components in the initial site model.
Figure 6. Street Hits analysis of the original site shows that most
activities do not take routes across the site.
198
Figure 7. Diagram of the generative workflow of Scenario One.
To test the proposed workflow, three scenarios are
generated under distinct design intentions and assumptions.
They follow the same primary generative methods but
differ in modeling and decision-making process.
3.1 Scenario One
This scenario aims to create a better linking zone. The new
network should contribute more efficient passing routes so
that increasing pedestrians can go across the site and
support new businesses and amenities.
In Figure 7, Step 1 is the primary setup. Step 2 presents the
Street Hits result using one normalized ADP data of Table
1. The progress in Step 2 is expanded below to show how
the main routes with high Street Hits gradually become
visualized during the simulation iteration of all buildings.
Using this result, Step 3 generates a new street network by
straightening the busiest streets, merging the minor links,
and converting the largest intersections into potential
plazas. This step is drawn manually at the current stage.
Based on this network, Step 4 analyzes Amenity Hits
distribution by setting all cells on the 20m*20m coordinate
grid as one amenity type. The results are visualized in heat
maps highlighting the recommended locations for each
amenity type (grocery, errands, library, entertainment,
restaurant, shopping). Since the population density is much
higher in the northern downtown, all new amenity locations
tend to concentrate at the north edge for maximized
potential patronage. However, heat maps still vary in color
uniformity due to the influences of existing nearby
amenities. For amenity allocation, designers can place
amenities on the best performing locations, conduct a new
simulation for the scenario and evaluate the Amenity Hits
of the newly placed amenities. Comparing the resulting Hits
with the Hits of other same-type amenities existing in the
model can help designers measure the balance between
supplies and demands of new amenities, and then make
decisions about their amounts and locations.
One thing to mention is this paper does not focus on the
parcellation in the lots or generation of building footprints.
However, to present that the previous results can be further
developed into actionable design scenarios, the workflow
integrates the last two steps. Step 5 computes the final
Walkscore based on the coordinate grid, which is used in
Step 6 to inform the distribution of development density
(FAR) on the generated lots. Since better walkability
indicates improved property values, the lots with a greater
Walkscore get a higher density. Figure 8 presents an
example of how the final masterplan could look.
Figure 8. Sample masterplan developed for Scenario One.
199
Figure 9. Diagram of the generative workflow of Scenario Two.
3.2 Scenario Two
This scenario intends to include several design conditions:
(1) a proposed pedestrian boulevard linking the railway
station and the downtown area through the new bridge; (2)
a predefined and specifically located high-rise cluster which
will hold a high density of population on design site; (3) a
prospective new school mostly serving the southern
residential neighborhood. The generative workflow is
adjusted to address these particular design issues (Figure 9).
Figure 10. Sample masterplan developed for Scenario Two.
Step 1 models the boulevard and high-rise cluster based on
the primary setup. The boulevard is modeled by setting the
segments on the grid along the route with a biased routing
factor of 0.5. The high-rise cluster is modeled by setting
500 population to each of five high-rise locations on the
coordinate grid. Step 2 visualizes heat maps of Amenity
Hits. These heat maps differ from Scenario One because the
“shortened” boulevard in the simulation attracts more trips
crossing the site through this route, and the high-rises also
bring more population as consumers. Among all amenities,
the analysis of Hits for school differs from the others
because it only considers pedestrians coming from the
southern residential area. The school’s heat map result also
reveals this adjustment as it is best located on the corners
that are closest to the neighborhood side. With all amenity
allocation decided, Step 3 visualizes Street Hits on the grid,
and Step 4 generates the new street network accordingly.
The final two steps of Walkscore analysis and lot-level
FAR distribution remain the same as Scenario One. Figure
10 presents an example of how the final masterplan looks.
3.3 Scenario Three
Instead of using one normalized ADP data in the first two
scenarios, this scenario considers the temporal difference in
street and amenity utilization. It aims to create a 24-hour
active neighborhood by overlapping the generative results
of time steps. However, this scenario only generates street
network while the amenity allocation is an input.
In Figure 12, Step 1 specifies an input of the amenity
allocation scenario. Step 2 visualizes the Street Hits for
three time-periods: morning, noon and evening. Step 3
generates the network by filtering the most vibrant streets at
these times and overlapping them together. Step 4 and Step
5 remain the same for Walkscore analysis and lot-level
FAR distribution. An additional Step 6 demonstrates
people’s dynamic movement on the network over time.
Figure 11. Sample masterplan developed for Scenario Three.
200
Figure 12. Diagram of the generative workflow of Scenario Three.
3.4 Comparison
To be adaptive to different design goals and conditions, the
proposed workflow varies in three scenarios. Firstly, the
sequences of decision-making are different. Scenario One
generates streets first because it aims to create better links
between the surrounding built environment. Scenario Two
has more specific requirements for program allocation.
Thus, the street network is generated later in the workflow
so as to address the new design conditions. Secondly, the
generative parameters change. The same one normalized
ADP is used in the first two scenarios while the third one
uses the ADP of multiple time periods. Also, the biased
routing factors and the population distribution are modeled
differently in Scenario Two. More parameters such as
biased destination factors or walking distance limits have
not yet been modified among three scenarios. They are able
to allow more precise controls by designers.
4 LIMITATION AND PROSPECT
The limitation of modeling is data quality. For example,
high-quality GIS data is only provided in major
metropolitan areas. Also, some open data source such as
OpenStreetMap has significantly fewer POI entries
compared with other sources such as Google data.
Consequently, a model that uses data where only a few
POIs have been recorded, will yield misleading results. The
workflow proposed in this paper will be able to benefit
from the ongoing efforts to improve urban data systems.
As for the simulation, there are difficulties in thorough
validation, because there is no openly available reference
data with which to compare the results. The current ADP
data is derived using user-generated data such as Google
Places “Popular Times” data, which mostly relies on GPS.
To provide a basic check of the simulation results, five
randomly selected samples of cafes and restaurants in the
simulation model are used for comparative study. Figure 13
plots their Amenity Hits and their real profiles in Google
“Popular Times” data (both normalized and scaled to 1.0)
together. There is a certain level of consistency, but
exceptions also exist. In the future, more detailed data such
as opening hours can be leveraged to improve consistency
further. Though “Popular Times” data is indeed an input of
deriving ADP, such comparison can still verify the
interpretation process of the framework along with other
synthesized parameters such as population distribution.
Figure 13. Comparison of five random samples’ Amenity Hits
results and their real profiles in Google Popular Times data.
In the generative workflow, caution is needed when taking
advantage of its adaptivity. Some customized parameters,
such as biased routing and destination factors, provide
designers with the power to control the generative process,
but they also open to the risk of being arbitrary or biased.
More sophisticated metrics defining these factors are in
demand.
201
5 CONCLUSION
Design decisions such as zoning, density, program
allocation, and the layout of public spaces and streets can
have a fundamental impact on the performance of mobility
systems. Employing urban planning to mitigate traffic-
related problems is widely recognized as an effective
strategy. It is expected that this research can equip
designers with capabilities that enable the co-design of
mobility solutions and urban form, thus motivating the
early discovery of cost-effective solutions.
This paper proposes a novel workflow of automating the
process of parsing the map data, building contextual models
with population and amenity data, conducting integrated
mobility simulation, and generating street network and
amenity allocation for urban design. The effectiveness and
adaptivity of the workflow are tested in a pedestrian-
oriented neighborhood design case study by generating
three scenarios for different design goals and conditions.
This versatile framework can contribute to the design
profession and education in terms of increasing awareness
and responsiveness to mobility-related urban factors.
Moreover, as mobility metrics also have economic and
environmental implications, the proposed framework can
become more valuable by including other stakeholders in
urban development. Practitioners in sectors such as real
estate, public services, and public health can leverage the
analysis result to make decisions, and designers can benefit
from involving a broader range of data and metrics from
these fields into the design solution-seeking process.
ACKNOWLEDGMENTS
The authors would like to thank Cornell CTECH for
funding this research and Nikhil Saraf for assisting with
software development of Urbano.
REFERENCES
[1] A. Loukaitou-Sideris, “Addressing the Challenges of Urban
Landscapes: Normative Goals for Urban Design,” Journal of
Urban Design, vol. 17, no. 4, pp. 467–484, Nov. 2012, doi:
10.1080/13574809.2012.706601.
[2] UNEP, “Sustainable Buildings and Climate Initiative,” 2015.
Accessed: Aug. 05, 2015. [Online]. Available:
http://www.unep.org/sbci/AboutSBCI/Background.asp.
[3] T. Haas, Ed., Sustainable urbanism and beyond: rethinking
cities for the future. New York: Rizzoli, 2012.
[4] L. D. Frank, J. F. Sallis, T. L. Conway, J. E. Chapman, B. E.
Saelens, and W. Bachman, “Many Pathways from Land Use
to Health: Associations between Neighborhood Walkability
and Active Transportation, Body Mass Index, and Air
Quality,” Journal of the American Planning Association, vol.
72, no. 1, pp. 75–87, Mar. 2006, doi:
10.1080/01944360608976725.
[5] I.-M. Lee and D. M. Buchner, “The Importance of Walking to
Public Health:,” Medicine & Science in Sports & Exercise,
vol. 40, no. Supplement, pp. S512–S518, Jul. 2008, doi:
10.1249/MSS.0b013e31817c65d0.
[6] S. Claris and C. Luebkeman, “Cities Alive: Towards a
walking world,” ARUP Publications, 2016. [Online].
Available: https://www.arup.com/-
/media/arup/files/publications/c/cities-alive_towards-a-
walking-world_lowres.pdf.
[7] K. M. Leyden, “Social Capital and the Built Environment:
The Importance of Walkable Neighborhoods,” Am J Public
Health, vol. 93, no. 9, pp. 1546–1551, Sep. 2003, doi:
10.2105/AJPH.93.9.1546.
[8] T. N. Clark, R. Lloyd, K. K. Wong, and P. Jain, “Amenities
Drive Urban Growth,” Journal of Urban Affairs, vol. 24, no.
5, pp. 493–515, Dec. 2002, doi: 10.1111/1467-9906.00134.
[9] A. Zandiatashbar and S. Hamidi, “Impacts of transit and
walking amenities on robust local knowledge economy,”
Cities, vol. 81, pp. 161–171, Nov. 2018, doi:
10.1016/j.cities.2018.04.005.
[10] G. F. Mulligan and J. I. Carruthers, “Amenities, Quality of
Life, and Regional Development,” in Investigating Quality of
Urban Life, vol. 45, R. W. Marans and R. J. Stimson, Eds.
Dordrecht: Springer Netherlands, 2011, pp. 107–133.
[11] T. Dogan, S. Samaranayake, and N. Saraf, “Urbano: A new
tool to promote mobility-aware urban design, active
transportation modeling and access analysis for amenities and
public transport.,” in SimAUD 2018, 2018.
[12] M. Brewster, D. Hurtado, S. Olson, and J. Yen,
“Walkscore.com: A New Methodology to Explore
Associations Between Neighborhood Resources, Race, and
Health,” presented at the 137st APHA Annual Meeting and
Exposition 2009, 2009, [Online]. Available: https://apha.
confex.
com/recording/apha/137am/pdf/free/4db77adf5df9fff0d3caf5
cafe28f496/paper205082_1. pdf.
[13] T. Rakha and C. F. Reinhart, “Generative Urban Modeling: A
Design Workflow for Walkability-Optimized Cities,” 2012.
[14] D. Nagy, L. Villaggi, and D. Benjamin, “Generative urban
design: integrating financial and energy goals for automated
neighborhood layout,” in Proceedings of the Symposium on
Simulation for Architecture and Urban Design, Accessed:
Jan. 09, 2020. [Online]. Available:
https://dl.acm.org/doi/abs/10.5555/3289750.3289775.
[15] T. Wortmann, C. Waibel, G. Nannicini, R. Evins, T.
Schroepfer, and J. Carmeliet, “Are Genetic Algorithms Really
the Best Choice in Building Energy Optimization?,” in
SimAUD 2017, Accessed: Jan. 10, 2020. [Online]. Available:
https://www.researchgate.net/publication/317277483_Are_Ge
netic_Algorithms_Really_the_Best_Choice_in_Building_Ene
rgy_Optimization.
[16] T. Dogan, Y. Yang, and S. Samaranayake, “Urbano: A tool to
promote active mobility modeling and amenity analysis in
urban design,” TAD Translation, 2020.
[17] “OpenStreetMap.”
https://www.openstreetmap.org/#map=17/52.90112/-1.22824
(accessed Apr. 01, 2020).
[18] “Places | Google Maps Platform,” Google Cloud.
https://cloud.google.com/maps-platform/places (accessed
Apr. 01, 2020).
[19] ASHRAE, “ASHRAE Handbook—Fundamentals (SI),”
ASHRAE, Atlanta, 2013.
[20] SIA, 2024: Standard-Nutzungsbedingungen Für Die Energie-
Und Gebäudetechnik. Zürich: Swiss Society of Engineers and
Architects, 2006.
[21] Y. Yang, S. Samaranayake, and T. Dogan, “Using Open Data
to Derive Local Amenity Demand Patterns for Walkability
Simulations and Amenity Utilization Analysis,” in
Proceedings of ECAADe 37 / SIGraDi 23, Porto, Portugal,
2019.
202
Powered by TCPDF (www.tcpdf.org)
