GEOMETRY AUGMENTATION FOR SPATIAL

DEVELOPMENT TO OPTIMIZE PEDESTRIAN FLOW

AND CIRCULATION.

A Dissertation Report

submitted in partial fulfillment of the requirements

for the award of the Degree of

BACHELOR OF ARCHITECTURE

By

NARITYA CHOUDHARY

(B. ARCH/10024/18)

DEPARTMENT OF ARCHITECTURE AND PLANNING

BIRLA INSTITUTE OF TECHNOLOGY MESRA
835215, RANCHI. 2022
 Naritya Choudhary | Dissertation 2022

CERTIFICATE

This is to certify that the work presented in the dissertation project

entitled “(Geometry augmentation for spatial development to
optimize pedestrian flow and circulation.)” by Mr.Naritya
Choudhary In partial fulfillment of the requirement for the award of
Degree Bachelor of Architecture of Birla Institute of Technology
Mesra, Ranchi is an authentic work carried out under my supervision
and guidance.

To the best of my knowledge, the content of this dissertation project

does not form a basis for the award of any previous Degree to
anyone else.

Dr. Prashant Prasad

Assistant Professor

The dissertation project as mentioned above is hereby approved

as a creditable study of project work and has been presented in a
satisfactory manner to warrant its acceptance as prerequisite to
the degree for which it has been submitted.

Dr. Smriti Mishra Dr. Manjari Chakraborty

Course coordinator Course coordinator

Head of the Department

ii
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

ACKNOWLEDGEMENT

I want to start by thanking my dissertation Supervisor, Dr. Prashant Prasad, who has
helped me with my research and motivated me to conquer every obstacle. I was able to
fit the puzzle together with the aid of his extensive knowledge and expertise.
Additionally, I would like to extend my sincere gratitude to my course coordinators,
Dr. Manjari Chakraborty and Dr. Smriti Mishra, for their faith in me, for their support
in both academic and personal endeavors, and for their superb oversight throughout this
process. I also want to express my gratitude to Ar. Aditi Nag for her unwavering
guidance and encouragement throughout the past four months. This report wouldn't
have been completed without their help and diligent participation in each step of the
process. I must appreciate Sharishma Manandhar, Keshav Jasoriya, and Satwik Singh
Rathore for their input during moments of perplexity and for being a consistent support
during the completion of my dissertation. I have so much thanks and admiration for
friendship that I am at a loss for words. Without the continuous support and affection
of my classmates Yogesh M. Singh, Razi Ahmed, Saurav Gaur, Aneesh Kumar, and
the rest, I would not have been able to finish. Most importantly, none of this could have
happened without my family. My parents and my brother. This dissertation stands as a
testament to their unconditional love and encouragement. Writing about this particular
subject has been challenging, rewarding and deeply motivating.

Thank you

Naritya Choudhary
B.Arch./10024/18

iii
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

CONTENTS
CONTENTS .............................................................................................IV
LIST OF FIGURES ..................................................................................VI
LIST OF TABLES.................................................................................. VII
ABSTRACT .......................................................................................... VIII
1.INTRODUCTION ................................................................................... 1
1.1 OVERVIEW......................................................................................................... 1
1.2 NEED ................................................................................................................... 2
1.3 JUSTIFICATION ................................................................................................. 3
1.4 AIM AND OBJECTIVE ...................................................................................... 4
1.5 SCOPE ................................................................................................................. 4
1.6 LIMITATIONS .................................................................................................... 5
1.7 METHODOLOGY & TOOLS ............................................................................. 5
2.LITERATURE REVIEW ........................................................................ 6
2.1 CROWD TYPOLOGIES AND TRIPS ................................................................ 6
2.1.1 CROWD BEHAVIOR PATTERNS ............................................................. 6
2.1.2 WALKING SPEED AND PATTERNS ........................................................ 8
2.1.3 LEVEL OF SERVICE ................................................................................. 11
2.1.4 IDENTIFICATION OF POSSIBLE TRIPS ................................................ 13
2.2 SIMULATION MODELS AND METHODS.................................................... 14
2.2.1 AGENT BASED SIMULATION ................................................................ 14
2.2.2 GENETIC ALGORITHM ........................................................................... 15
3.GOVERNING PARAMETERS ............................................................ 16
3.1 PHYSICAL PARAMETERS ............................................................................. 17
3.1.1 DIMENSIONS............................................................................................. 17
3.1.2 OBSTACLES .............................................................................................. 18
3.1.3 ACCESS NODES ........................................................................................ 19
3.1.4 COORDINATES ......................................................................................... 21
3.1.5 PHYSICAL PARAMETERS: PERMUTATIONS AND CASES .............. 22
3.2 ASSUMPTIONS AND CONSTANTS .............................................................. 23
4. SIMULATION SCRIPTING................................................................ 24
4.1 GEOMETRY AUGMENTATION .................................................................... 25

iv
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

4.2 PEDESTRIAN CROWD SIMULATION.......................................................... 26

4.3 PATH MODELING ........................................................................................... 27
4.4 HEAT MAP DERIVATION .............................................................................. 27
4.5 ITERATIVE FITNESS TEST............................................................................ 29
5. RESULTS OPTIMIZATION FOR CORRIDOR ................................ 30
5.1 DERIVED DATA .............................................................................................. 30
5.2 VALIDATION ................................................................................................... 31
5.3 COMPARISON OF ITERATIVE AND INDIVIDUAL ................................... 32
6. FURTHER SCOPE .............................................................................. 35
6.1 FURTHER REFINEMENT OF THE MODEL ................................................. 37
6.2 ADAPTATION FOR DIFFERENT MODELS ................................................. 36
6.3 POTENTIAL DERIVED USE-CASE ................ ERROR! BOOKMARK NOT
DEFINED.
7. CONCLUSION .................................................................................... 37
8. REFERENCES ..................................................................................... 38
9. BIBLIOGRAPHY................................................................................. 38

v
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

LIST OF FIGURES

FIG 1: CROWD WAITING AT A PLATFORM AT DADAR CENTRAL STATION

| SOURCE: TRACK.IN ................................................................................................. 1
FIG 2: CROWD SIMULATION ILLUSTRATION | SOURCE: AUTHOR ................ 2
FIG 3: SHORTEST PATH ILLUSTRATION | SOURCE: AUTHOR ......................... 3
FIG 4: PEDESTRIAN SPACING ILLUSTRATION | SOURCE: EDWARD HALL
AND ROBERT SOMMER ............................................................................................ 7
FIG 4: PEDESTRIAN SPEED GRAPHICAL | SOURCE: TCQM 2015 ..................... 9
FIG 5: PEDESTRIAN FLOW GRAPHICAL | SOURCE: TCQM 2015 ...................... 9
FIG 6: PASSENGER LOAD | SOURCE: TCQM 2015 .............................................. 11
FIG 7: LEVEL OF SERVICE ILLUSTRATION | SOURCE: AUTHOR .................. 13
FIG 8: GOVERNING PARAMETERS FLOWCHART | SOURCE: AUTHOR ....... 16
FIG 9: GOVERNING PARAMETERS: `DIMENSION’ | SOURCE: AUTHOR ...... 18
FIG 10: GOVERNING PARAMETERS: `OBSTACLE’ | SOURCE: AUTHOR ...... 19
FIG 11: GOVERNING PARAMETERS: `ACCESS NODE’ | SOURCE: AUTHOR 20
FIG 12: GOVERNING PARAMETERS: `COORDINATE’ | SOURCE: AUTHOR. 22
FIG 13: WORKING FRAMEWORK SCRIPT | SOURCE: AUTHOR ...................... 25
FIG 14: GEOMETRY AUGMENTATION| SOURCE: AUTHOR ............................ 25
FIG 15: PEDESTRIAN CROWD SIMULATION| SOURCE: AUTHOR ................. 26
FIG 16: PEDESTRIAN CROWD SIMULATION SCRIPT| SOURCE: AUTHOR ... 26
FIG 16: PATH TRAJECTORY MODELING | SOURCE: AUTHOR ....................... 27
FIG 17: HEAT MAP GENERATION | SOURCE: AUTHOR ................................... 28
FIG 18: HEAT MAP GENERATION | SOURCE: AUTHOR ................................... 28
FIG 19: HEAT MAP GENERATION SCRIPT | SOURCE: AUTHOR ..................... 29
FIG 19: ITERATIVE TEST 1 | SOURCE: AUTHOR ................................................ 29
FIG 20: ITERATIVE TEST 2 | SOURCE: AUTHOR ................................................ 30
FIG 21: ITERATIVE TEST 3 | SOURCE: AUTHOR ................................................ 30
FIG 22: REAL TIME DATA COLLECTION | SOURCE: AUTHOR ....................... 31
FIG 23: HEAT MAP VALIDATION | SOURCE: AUTHOR/(HWANG, 2019) ....... 32
FIG 24: ITERATIVE TEST QUESTION | SOURCE: AUTHOR .............................. 33
FIG 24: ITERATIVE TEST QUESTION SAMPLE | SOURCE: AUTHOR ............. 33
FIG 25: HEAT MAP GENETIC ALGORITHM | SOURCE: AUTHOR ................... 34

vi
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

LIST OF TABLES

TABLE 1.1 METHODOLOGY & TOOLS | SOURCE: AUTHOR ......... 5

TABLE 2.1 LEVEL OF SERVICE | SOURCE: TCQM 2015 ................ 12
TABLE 3.1 POSSIBLE TRIPS | SOURCE: AUTHOR .......................... 13
TABLE 4.1 GOVERNING PARAMETERS | SOURCE: AUTHOR ..... 16
TABLE 4.2 PHYSICAL PARAMETERS | SOURCE: AUTHOR.......... 17
TABLE 4.3 DIMENSION PARAMETER | SOURCE: AUTHOR ......... 18
TABLE 4.4 OBSTACLE PARAMETER | SOURCE: AUTHOR ........... 19
TABLE 4.5 ACCESS NODE PARAMETER | SOURCE: AUTHOR .... 20
TABLE 4.6 COORDINATES PARAMETER | SOURCE: AUTHOR ... 21
TABLE 4.7 LIST OF VARIABLES| SOURCE: AUTHOR.................... 22
TABLE 4.8 PERMUTATION CASES | SOURCE: AUTHOR............... 23
TABLE 5.1 ASSUMPTIONS AND CONSTANTS | SOURCE:
AUTHOR .................................................................................................. 24
TABLE 6.1 ADAPTAION PARAMETERS | SOURCE: AUTHOR ...... 36

vii
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

ABSTRACT
Simulation models are becoming increasingly popular in the AEC industry and have
had massive impacts on the performance and predictability of designed systems and
structures. The study attempts to establish a framework for a model that acts like a
design tool for designers and planners of transit nodes and corridors. An accurate
assessment of pedestrian trajectories is required to determine demand at various spots
on the platform. Because of the diverse origins and destinations, planning transit
platforms may be a difficult issue. To build transit stations to suit present and future
demand, a thorough understanding of predicted pedestrian movements on the platform
is required.

Currently, the needed fundamentals are not available at the desired depth. In rare
situations, certain stations, particularly their platforms, may have hit capacity
restrictions. As a result, multiple transit platform additions are planned. As a result,
multiple transit platform additions are planned. To build transit stations to suit present
and future demand, a thorough understanding of predicted pedestrian movements on
the platform is required. The amount of people on the platform and the time until the
next transit arrived were other crucial factors in determining the waiting spot.
The investigation provides a baseline & a framework for a genetic algorithm-based
simulation model for interventions in transit nodes, and successfully demonstrates
potential for further exploration and refinement. It performs significantly better than
human input design and has a high enough confidence value for execution and
implementation.

Keywords: Pedestrian simulation, Agent based modeling, Genetic based

algorithm, Crowd optimization, Iterative model, Generative design,
Transit nodes, Heat map.

viii
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

1.INTRODUCTION
Modern urban cities' growth is governed by public transportation; more and more cities
are developing walkable infrastructure and have identified public transportation as an
essential mode to achieve so. Many transit nodes and corridors are currently
overcrowded and were not designed to accommodate the population growth caused by
increased public transportation ridership. At the same time, AI, simulation models, and
generative design are churning out more efficient design solutions than us architects
and planners, and with time, all of us will need to get along with the technology to make
more informed decisions and make more accurate projections. The study attempts to
establish a framework for a model that acts like a design tool for designers and planners
of transit nodes and corridors, enabling them to make data-informed design decisions
before, during, and after the design and planning phases. The tool does not aim at
replacing the designer or the planner, but to provide him with more accurate trajectories
and projections, for them to then formulate their design upon. The data provided by the
simulation model can have a wide spectrum of use cases with multiple possible
permutations and combinations.

Fig 1: Crowd waiting at a platform at Dadar central station | Source: track.in

1.1 OVERVIEW
Simulation models are becoming increasingly popular in the domain of the AEC
industry and have had massive impacts on the performance and predictability of

1
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

designed systems and structures. Data-driven decision making thus becomes an

essential part of the process, enabling simulation models to perform better and assist us
in making better decisions and projections, and AI and machine learning are ever-
growing contributors to the same. Fig: 2 illustrates an abstract image of crowd
simulation.

Fig 2: Crowd simulation illustration | Source: Author

1.2 NEED
Because of the diverse origins and destinations, as well as the mix of waiting and
strolling pedestrians, planning transit platforms may be a difficult issue. An accurate
assessment of pedestrian trajectories is required to determine demand at various spots
on the platform. The expected walking distances and the sources and destinations are
important factors in this. According to the literature, it is unclear how people pick their
destinations on the platform and hence their itineraries. One common assumption is that
passengers, particularly regular commuters, already walk to the place on the platform
at their origin station to try and minimize the distance to the exit at their destination.

2
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

This makes it unclear for a designer/planner to accurately judge the trajectories without
any simulation data or real-time monitoring systems.

Fig 3: Shortest path illustration | source: Author

1.3 JUSTIFICATION
The number of passengers has expanded significantly as a result of considerable
improvements in the railway network during the past few decades. This resulted in the
current scenario, in which certain railway stations are used by more passengers during
peak hours than they were built for. In rare situations, certain stations, particularly their
platforms, may have hit their capacity restrictions. As a result, multiple transit platform
additions are planned.
To build transit stations to suit present and future demand, a thorough understanding of
predicted pedestrian movements on the platform is required. Currently, the needed
fundamentals are not available at the desired depth. Although today's counting
equipment can establish the number of people at a cross section (e.g., platform
entrances or transit doors), data on platform movement is rarely accessible.
Nonetheless, this information is essential to establishing the load on certain portions of
the platform and, as a result, the platform's required breadth and size.

3
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Platforms are commonly utilized for both waiting and walking. It is therefore critical
for the design to determine the distribution of waiting pedestrians as well as the
trajectories of the walking pedestrian. The location with the highest pedestrian
concentration may be calculated and utilized to construct the platform using this
knowledge and the projected number of people at the platform.

Aspects of the distribution of waiting pedestrians on transit platforms were previously

examined (Bosina et al., 2015). Observations revealed that waiting pedestrians choose
sites where they may lean against a wall, obstruction, or railing. The amount of people
on the platform and the time until the next transit arrived were other crucial factors in
determining the waiting spot. The impacts on the trajectories of walking pedestrians
will be investigated in depth in this contribution.

1.4 AIM AND OBJECTIVE

The aim of the study primarily focuses on "formulating a model to optimize pedestrian
flow and circulation in polymorphous transit matrices through generative design and
spatial character augmentation." In the process of achieving this, the following
objectives were listed:

● To determine pertinent crowd flow typologies in a transit matrix.

● To establish parameters, formulations & assumptions affecting pedestrian flow
& circulation in a transit node.
● To create a base spatial geometry based on a simulation-derived crowd matrix.
● To formulate a model as a framework to optimize pedestrian flow and
circulation for interventions in existing and new design workflows.

1.5 SCOPE
The exploration targets transit-oriented nodes and corridors and helps designers make
more informed decisions. It has a wide spectrum of use cases, which have been defined
in the research. The paper formulates a framework for future development and
optimization for more accurate implementation and monitoring. This research tries to
combine Agent based simulation and genetic algorithm-based simulation to optimize
pedestrian flow and circulation

4
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

1.6 LIMITATIONS
In this study, a pedestrian simulation model is evaluated in relation to architectural
design. It was found that the simulation model’s limitations are connected to the
creators' skills. The user must decide if the variable they want to examine can be
separated into the few behaviors that the program does dictate, since the behavior of
the subjects in the model is mostly driven by the script that the user writes. If this is the
case, the simulation model may nevertheless be useful to architects and designers in a
number of circumstances. The simulation model would be useful in any circumstance
where the available space is the variable, or if there is a defined area but another
factor—such as the number of ticket booths, cashiers, or loading docks—is the variable.
As a result, this simulation model is helpful in many circumstances for which designers
may have wished for it, but they must carefully consider the context and the particular
variable in order to decide how to research it.
1.7 METHODOLOGY & TOOLS
Various tools and software were adopted to achieve the results and construct a
framework; the tools are subject to change upon future development. The methodology
and tools used are listed below in Table 1.1

TABLE 1.1 METHODOLOGY & TOOLS | SOURCE: AUTHOR

5
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

2.LITERATURE REVIEW
Numerous books, journals, and articles were used to define the parameters to base the
simulation model on. These parameters help us identify the variables and standards the
models will be based on. Selection of parameters is what defines the simulation model
and is the basis of genetic based iterative test that help us determine the fitness
parameters.

2.1 CROWD TYPOLOGIES AND TRIPS

Identification of crowd typologies and trips is essential to accurately and efficiently
defining the flow of the crowd and their social walking behavior. This section
specifically looks at the types of crowds, the behaviors, and the patterns observed in
previous research.

2.1.1 CROWD BEHAVIOR PATTERNS

Extensive research has been conducted on the subject of environment-behavior

interactions and the various disciplines that study it over a three-decade period. The
primary goal of the research was to generate rules that would aid in planning, as well
as the design and analysis of space.

It is common to see the emergence of distinct lanes, each of which is made up of

pedestrians traveling in the same way, among crowds of people walking in opposing
directions. This is also apparent if passing pedestrians choose to avoid one another, with
an equal likelihood on the left and right sides. The formation of lanes is caused by a
corresponding decline in the frequency of essential avoidance and deceleration
operations, which promote the effectiveness of passenger flow.

It is possible to define these lanes as "dynamically variable." The width of the entire
roadway and the volume of foot traffic are two elements that affect how many lanes
will be established. Traditionally, the interpretation of lane creation makes the
assumption that pedestrians generally choose to cross the street on the side designated
for motor traffic.

6
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

The following explanation helps understand how lanes form: Interactions between
pedestrians traveling against the flow of traffic or in places with mixed traffic directions
will be frequent and intense. Additionally, whenever two passengers are interacting,
they usually move slightly apart so they may pass one another. This sideways
movement helps to separate people walking in opposing directions. Additionally,
encounters between pedestrians traveling in uniform lanes will be less frequent and less
significant. As a result, when the variations are minor, it is thought that there is less
propensity to split up existing lanes. Furthermore, the state with the lowest interaction
rate, which is likewise connected to the motion's highest efficiency, does correlate to
the configuration that is the most stable. To understand it better, below is an illustration
of pedestrian spacing to understand repulsion and walking behavior.

Fig 4: Pedestrian spacing illustration | source: Edward Hall and Robert Sommer

An oscillation of passing directions is observed in areas with small pathways. When

one pedestrian can easily pass through the door, typically another can do so as well.
However, the pressure coming from the other side of the channel will force the
pedestrian flow that results to cease. A pedestrian will eventually be allowed to walk
through the door in the opposite direction, and the procedure will then repeat as before.

7
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

The following is a description of the mechanism that results in alternating flows: long
pathways clearly show how easy it is for other people traveling in the same direction to
follow once one of them passes a constriction (such as a door, a stairway, etc.). As a
result, there are fewer people waiting and pressing on this side of the constriction than
on the other, which increases the likelihood that the route will be filled. After some
time, this results in an impasse, and the direction of walking is changed. Short and wide
passages are easier to navigate, which results in more frequent changes in walking
direction.

At junctions, there is conflict between alternating, unstable, and transient collective

patterns of motion. For instance, the temporary roundabout traffic phases (shown
above) alternate with phases in which the intersection is traversed either vertically or
horizontally. Placing certain impediments in the middle of the junction will favor
smooth roundabout traffic over competing and inefficient patterns of motion, greatly
increasing the efficiency of the flow of people.

2.1.2 WALKING SPEED AND PATTERNS

The walkway level of service can be defined based on walking speeds and rates, as
shown below, which was derived from and interpreted from the literature. The walking
pedestrian walking speeds and its relation to pedestrian area help us parametrize how
pedestrian area is affected by walking speeds and how it affects the pedestrian flow in
multi-directional flows. (TCQM, 2015)

8
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 4: Pedestrian speed graphical | source: TCQM 2015

Fig 5: Pedestrian flow graphical | source: TCQM 2015

The flow, density, and pace of pedestrians are all connected. This interdependence
gives us a standard by which to determine the V value, or the ped flow per unit width.
The relationship between density, speed, and flow for pedestrians is described by the
following formula (TCQM 2015):

9
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

v=SxD, where (TCQM 2015)

v = peds flow per unit width (p/m/min),

S = peds speed (m/min), and

D = peds density (p/m. sq).

The expression's flow variable is written as a flow per unit of width. The reciprocal of
density may be used to get the following alternative expression:

v=S/M, where

v = peds flow per unit width (p/m/min),

S = peds speed (m/min), and

M = peds space (m2 /p), adjusted as appropriate for pedestrian characteristics. (LOS)

It has been demonstrated that free-flow walking speeds may range from 45 m/min to
145 m/mi. According to this definition, locomotion at speeds below 45 m/min is
restricted and shuffling, and locomotion at speeds above 145 m/min is running. A
pedestrian walking speed typically used for design is 75 m/min, which is
approximately 4.5 km/h. So, for the simulation model of base augmented geometry,
the pedestrian flow per unit width can be derived as -

v=S/M

v=75/m

v=75xP/(lxb), where,

v = pedestrian flow per unit width (p/m/min),

P= number of total pedestrians

l= length of the corridor

b= breadth of the corridor

10
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

now,

The computation of a pedestrian model can run for an infinite number of frames with
the simulation giving no concretized data to base decisions on. To tackle this constraint,
the following graph from TCQM, 2015 was examined:

Fig 6: Passenger load | source: TCQM 2015

For calculation and projection of design capacity load, the pedestrian flow of the peak
15 minutes of the peak 1 hour is taken into account (TCQM, 2015), and this same time
frame is used as a baseline for the duration to run the simulation on. The time in the
simulation model cannot be expressed in terms of seconds or minutes, so for the sake
of the simulation and data computation, frames are considered as a desirable measure
of time, and 30 frames are equalized as 1 second The peak load is calculated based on
the level of service a particular transit node desires to achieve.

2.1.3 LEVEL OF SERVICE

There is no specific measure to determine the quality of service in a transit node. Level
of service is a parameter defined by the unit width flow and pedestrian space. It governs
the quality of a transit node based on pedestrian movement and egress. This helps in

11
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

the projection of design capacity and also helps determine measures and methods for
crowd optimization. (TCQM, 2015). According to TCQM 2015, it is broadly classified
into various types, namely LOS A, LOS B, LOS C, LOS D, LOS E and LOS F
For LOS A, walking speeds may be changed at will, making collisions with other
pedestrians uncommon. For LOS B, pedestrians may choose their own walking speeds,
and they are conscious of other people and react to their presence. For LOS C walking
speeds may be freely chosen; passing is feasible in unidirectional streams; however,
there will be small issues if one turn around or cross. For LOS D there is little freedom
to choose one's preferred walking pace and to pass other people; there is a significant
likelihood of disputes while turning around or crossing. In case of LOS E all pedestrians
are limited in their ability to walk quickly and pass other people; they can only go ahead
by shuffling; they can only travel backward with significant difficulty; and the amount
of traffic is beginning to exceed their ability to walk. Finally for LOS F walking speeds
are highly constrained; there is constant, unavoidable touch with other people; it is
nearly impossible to walk backwards or across; and the flow is unpredictable. The exact
numerical representation of different level of services is listed in table 2.1.

TABLE 2.1 LEVEL OF SERVICE | SOURCE: TCQM 2015

The same is illustrated below for a better understanding of LOS, and to visually
understand the metrics of the same.

12
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

LOS A LOS B LOS C

LOS D LOS E LOS F

Fig 7: Level of service illustration | source: Author

2.1.4 IDENTIFICATION OF POSSIBLE TRIPS

Based on existing research, potential trips in transit nodes were identified and then
simplified for the sake of simulation and execution. The main typology of trips was
classified and generalized for projection of trajectories and the distances of the possible
trips would serve as a baseline for fitness objectives They are listed in table 3.1.

TABLE 3.1 POSSIBLE TRIPS | SOURCE: AUTHOR

13
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

2.2 SIMULATION MODELS AND METHODS

In this part, we try to assess different simulation models and their use cases to identify
a potential base model for our simulation. For our simulation model, a combination of
genetic algorithm and agent based simulation was considered. Agent based simulation
was used for pedestrian simulation and genetic algorithm based iterative test was used
for execution of generative design,

2.2.1 AGENT BASED SIMULATION

Agent based simulation models can be understood as a computational model for

simulating the behaviors along with interactions of autonomous agents (individual and
collective entities), with the goal of better understanding how a system behaves and
what influences its results. It incorporates concepts from evolutionary programming,
complex systems, computational sociology, emergence, and game theory. In order to
comprehend the stochasticity of these models, Monte Carlo techniques are required.
Agent based simulation models are sometimes known as "individual-based
models."(Balaguer, 2015)
Agent based simulation modeling is comparable to but unlike from the idea of multi-
agent models or multi agent simulation in that the purpose of agent based simulation
model is to seek descriptive perception within the collective conduct of agents obeying
meek instructions, typically in natural systems, as opposed to designing agents or
addressing particular practical or engineering problems.
In an effort to recreate and foretell the emergence of complex phenomena, agent-based
models are a type of microscale model that simulates the concurrent operations and
interactions of several agents. It's an emerging process, which some people translate as
"the whole is larger than the sum of its parts." In other words, interactions between
lower-level subsystems lead to the emergence of higher-level system features. Or,
macro-scale state changes may result from the actions of small-scale agents.
Alternatively, complex behaviors are produced by simple behaviors (i.e., agents that
adhere to rules). (Balaguer, 2015)
Individual agents are often considered to be boundedly rational and are assumed to be
employing heuristics or straightforward decision-making principles to act in what they
believe to be their own safeties, for instance reproduction, monetary gain, or societal

14
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

standing. It is possible for Agent based simulation model agents to "learn," "adapt," and
"reproduce." (Balaguer, 2015)

2.2.2 GENETIC ALGORITHM

Complex sequential, allocation, and selection tasks can be solved using genetic
algorithms as optimization techniques. The operation of genetic algorithms is
analogous to that of biological evolution.
Charles Darwin, a famous naturalist, discovered that those who reproduced most
successfully were those who were most suited to their circumstances. An individual
may use and pass on desirable traits to its progeny after acquiring them, whether
through mutation or cross-breeding. These advantages in selection will help them and
their offspring, allowing all members of the relevant species to develop these traits over
time. Because individuals lacking these benefits will have fewer children as a result of
this selection disadvantage, their lack will result in extinction.
These similar ideas are applied by genetic algorithms. From a given initial collection
of suggested solutions, the most effective ones are chosen and used to the development
of new solutions. Then, better solutions are developed throughout future generations.
The so-called fitness value that a simulation model generates determines whether a
proposed solution has a selection advantage and whether it is then used to generate new
solutions.
One may need to optimize the sequence in which orders are loaded for a production
plant. Using a simulation model, one may determine the overall processing time for
each recommended loading sequence of orders. The evolutionary cycle will then
provide ever-improving sequences that process commands as soon as possible. Genetic
algorithms' key benefits are their independence from the task at hand and the excellent
quality of the answers they produce. Given this, genetic algorithms are especially well
suited to assisting one in a variety of simulation-based optimization activities.

15
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

3.GOVERNING PARAMETERS
This section outlines parameters affecting pedestrian flow and circulation, all the
domains and ranges with their variables, and how they affect pedestrian flow and
circulation. The parameters are then classified into 3 types, physical, behavioral, and
auxiliary which can be also be considered as variables, constants and limiting. Table
4.1 depicts the matrix of governing parameters:

Fig 8: Governing parameters flowchart | source: Author

TABLE 4.1 GOVERNING PARAMETERS | SOURCE: AUTHOR

16
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

3.1 PHYSICAL PARAMETERS

Physical parameters are parameters that govern the dimensionality and spatial character
of a space. These can be dimensions, positions, etc. Physical parameters can be directly
parametrized in the model and can be controlled by the user directly. A more detailed
list is listed below for better classification and understanding to then be incorporated in
the script in table 4.2.

TABLE 4.2 PHYSICAL PARAMETERS | SOURCE: AUTHOR

3.1.1 DIMENSIONS
The ‘Dimension’ parameter is defined by the physical dimensions of the transit
corridor, it can be length, breadth, radius, diameter, angle etc. based on the chosen form
and number of segments. For our base simulation we consider a rectangular corridor
with length and breadth to be the governing parameters. The height of a space is a
limitation in this scenario as the behavioral changes in the flow and movement due to
the height cannot be standardized for the sake of parametrization. The breadth and the
length are directly controlled by a number slider with optimal minima and maxima
range, which would be then used as a domain for the iterative test. Table 4.3 lists down
the parameters.

17
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

TABLE 4.3 DIMENSION PARAMETER | SOURCE: AUTHOR

Fig 9: Governing parameters: `dimension’ | source: Author

3.1.2 OBSTACLES
The ‘Obstacle’ parameter is defined by the physical obstacles present in a spatial
geometry, as outlined in earlier literature (TCQM, 2015), obstacles affect the flow of
the crowd, for the simulation model the obstacles are considered as columns in the
space. They can be of a wide variety, columns, walls, billboards etc. The number of
segments, form, dimensions, orientation all can be controlled in the script and at the
same time can have a genetic seed-based randomization for iterative test. Table 4.4 lists
down the parameters.

18
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

TABLE 4.4 OBSTACLE PARAMETER | SOURCE: AUTHOR

Fig 10: Governing parameters: `obstacle’ | source: Author

3.1.3 ACCESS NODES

The ‘Access nodes’ parameter is defined by a point parameter defining the source of
crowd generation and crowd termination. There can be n-number of access nodes with
controlled parameters for bi-directional and unidirectional crowd simulation and can be
adjusted depending on the complexity of the circulation required. Similar to that of

19
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

dimension, the height again acts as a limiting factor as the height of an access node is a
limitation in this scenario as the behavioral changes in the flow and movement due to
the height cannot be standardized for the sake of parametrization. For the base
simulation model, 2 entry exit gates, concourse A and concourse B are considered as
the access nodes. The width and position of the access on the edges of the corridor are
directly controlled by a number sliders, MD-sliders and positional mapping with
optimal minima and maxima range, which would be then used as a domain for the
iterative test. Table 4.5 lists down the parameters.

TABLE 4.5 ACCESS NODE PARAMETER | SOURCE: AUTHOR

Fig 11: Governing parameters: `access node’ | source: Author

20
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

3.1.4 COORDINATES
The ‘Coordinate’ parameter is defined by a point parameter defining the source of point
of interest and position of different amenities on the corridor level. Level of social
interaction is the limiting parameter in this case as it cannot account for the delay and
congestion caused by social interactions of individuals and groups on the corridor level.
These are controlled and parametrized by a range of MD sliders, number sliders, point
mappers, and behavioral constants of probability of use case and adjacency. The
position of these points of interests are the primary target gene pool for iterative test.
The table 4.6 lists down the parameters.

TABLE 4.6 COORDINATES PARAMETER | SOURCE: AUTHOR

21
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 12: Governing parameters: `coordinate’ | source: Author

3.1.5 PHYSICAL PARAMETERS: PERMUTATIONS AND CASES

Physical parameters account for a major chunk of simulation data and thus can have
multiple permutations for multiple cases. First, listing down all the physical variable
parameters that can be directly controlled by the user in the table 4.7

TABLE 4.7 LIST OF VARIABLES| SOURCE: AUTHOR

22
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

We have a total of 8 individual parameters across various categories that can be

individually controlled through various input methods mentioned above. Now lets say,
we fix one of the 8 variables, this leaves us to freely control the other 7 parameters and
hence, run the iterative test accordingly. For example- The width of the corridor can be
fixed to a numeric value and the remaining parameters can be controlled individually.

Now, deriving from the different cases and permutations, a total of 255 possible cases
can be simulated in the model, running the fitness test on n- number of variables. The
detailed list of cases is tabulated in table 4.8.

TABLE 4.8 PERMUTATION CASES | SOURCE: AUTHOR

3.2 ASSUMPTIONS AND CONSTANTS

For the simulation model, numerous assumptions and constants were taken into account
to achieve a higher confidence in the model and simplification of computation.

23
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Pedestrian walking speed was taken as 75m/min as a mean of the range. Similarly
peripheral vision during walking speeds was considered as 120 degrees. Obstacle
repulsion of 1m was added to rectify potential geometry clipping issues and for the
design population, Level of service was the chosen matrix. Major assumptions and
constants are tabulated in table 5.1.

TABLE 5.1 ASSUMPTIONS AND CONSTANTS | SOURCE: AUTHOR

4. SIMULATION SCRIPTING
The final simulation model consists of various sections for different means of data
collection and computation. These mainly are the geometry augmentation, pedestrian
crowd simulation, path modeling and points of interest generation, genetic iterative test,
heat map visualization and monitoring and data computation. The full script is shown
in Fig: 13 and an enlarged version is attached in annexure 1.

24
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 13: Working framework script | source: Author

4.1 GEOMETRY AUGMENTATION

Geometry augmentation, defines a base geometry to run the pedestrian crowd
simulation on. The base geometry augmentation allows for the user to control all
physical aspects of the transit corridor. The base geometry is also used as a basis for
iterative fitness test. For this test, we set up a base of 34mx22m with pentagonal
obstacles with 4 access nodes on either side of the polygon with 4 major points of
interests: shop, EVM A, EVM B, and water. Every aspect of the base model is governed
by the user input. Fig; 14 is the base sim and the generated geometry of the same:

Fig 14: Geometry Augmentation| source: Author

25
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

4.2 PEDESTRIAN CROWD SIMULATION

The next part is pedestrian crowd simulation based on agent based simulation model
considering the crowd behavior patterns listed below. The crowd simulation for this
particular instance is being considered as bi-directional, but can be considered as
unidirectional for different scenarios and instances. Fig 15 shows the working of the
same and Fig: 16 depicts the script.

Fig 15: Pedestrian crowd simulation| source: Author

Fig 16: Pedestrian crowd simulation script| source: Author

26
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

4.3 PATH MODELING

Path modeling is projection of trajectories for the iterative fitness tool to minimize
walking distances, at the same time minimizing congestion and crowd influx. The major
trajectories considered are through access nodes and points of interests serve as a detour
for a certain probable population of the crowd matrix. The script outlined is depicted in
Fig: 17.

Fig 17: Path trajectory modeling | source: Author

4.4 HEAT MAP DERIVATION

Heat map derivation is another aspect of the simulation model. The heat map is
integrated in the script and can be run in real time in tandem with the crowd simulation

27
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

model. Heat map helps the user identify potential gaps and points of conflict in the
derived base geometry and mitigate the results in real time. The heat map generation is
shown in Fig 18 and Fig 19

Fig 18: Heat map generation | source: Author

Fig 19: Heat map generation | source: Author

28
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 20: Heat map generation script | source: Author

4.5 ITERATIVE FITNESS TEST

The iterative test is based on a genetic based algorithm and works on achieving single
or multiple fitness objectives in tandem. For this particular instance, we have a look at
the iterative test defined by path trajectories and identification of access nodes to
achieve fitness objective of minimum travel path. The dimension parameters and
obstacle parameters are taken as constants for this particular test. The iterative test
provides us with fitness objectives, standard deviation in the fitness objectives of all
generated iterations and a 3D visualization of the sample space of iterations. For this
particular instance with only the access nodes being parametrized, 65e5 number of
iterations were generated

Fig 21: Iterative test 1 | source: Author

29
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 22: Iterative test 2 | source: Author

Fig 23: Iterative test 3 | source: Author

5. RESULTS OPTIMIZATION FOR CORRIDOR

The derive results, then can be linked together to identify potential design points of
interests and conflicts that a user might face, along with other data collection and
visualization methods to optimize pedestrian flow and circulation. The data can be
extracted for multiple use cases like existing stations which need intervention for
improvement of crowd flow and circulation quality.

5.1 DERIVED DATA

The simulation model is scripted in a way to provide designers with a set of real time
data which not only can be used for projection but also for monitoring of crowd. The
below image illustrates a prototype for data collection in real-time.

30
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 24: Real time data collection | source: Author

5.2 VALIDATION
For validation of the model, an existing study of an existing transit station was chosen.
New data analysis techniques are required to interpret the massive volumes of data
collected through the implementation of smart city technologies. Researchers analyzed
data from over 100 MAC (Media Access Control) ID sensors installed in a busy subway
station serving two subway lines to examine pedestrian transfer behavior using a
fingerprinting method in this study. The researchers then developed a model based on
the cumulative visibility of moving objects (CVMO), an AI-based technique, to present
the data in a form that might be used to address pedestrian flow issues in this practical
use of smart city technology. (Hwang, 2019)

The fingerprinting wireless location sensing method was utilized to map the MAC ID
location data collected over a three-month observation period in order to display the
congestion situation in real time. Furthermore, researchers developed a model that can
lead an instantaneous reaction to newly revealed conditions. Furthermore, researchers
devised a variety of behavioral economics-based methods for distributing traffic and
improving pedestrian flow, the efficacy of which was later demonstrated during a
following monitoring period. The proposed pedestrian flow analysis technique can help

31
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

reduce pedestrian congestion while also preventing accidents and maintaining law and
order. (Hwang, 2019)
The heat map generated by the earlier study was then compared to the one generated
by the simulation model adopted in this study. The current model demonstrates a
confidence of 79.2%.

Fig 25: Heat map validation | source: Author/(Hwang, 2019)

5.3 COMPARISON OF ITERATIVE AND INDIVIDUAL

For the sake of comparison, and to understand how better the simulation and iterative
test is compared to human design and input, a small survey was conducted where the
same design situation was provided to the simulation model and designers. All
parameters were fixed and only the position of access nodes and points of interests were
the variables the designer can select. The position of obstacles was fixed. The users
were given stickers which they had to place at their desired location, in an attempt to
optimize the pedestrian flow and circulation, and minimize commute distance at the
same time,

32
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 26: Iterative test question | source: Author

Fig 27: Iterative test question sample | source: Author

33
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Post collection of the data, average positions were taken from the survey results, same
result was generated in the simulation model using iterative test and the crowd
simulation results were compared.

Fig 28: Heat map Genetic algorithm | source: Author

It was noted that the genetic algorithm-based simulation model performed 15.5% better
than the human design version, and this difference would be even more significant if
more variables and parameters were introduced in the design problem.

34
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Fig 29: Heat map Human design| source: Author

6. FURTHER SCOPE
This study provides a baseline for a framework of a design tool that can be further
developed into a cohesive software package for designers and planners of transit nodes
and corridors, enabling them to make data-informed design decisions before, during,
and after the design and planning phases. The tool does not aim at replacing the designer
or the planner, but to provide him with more accurate trajectories and projections, for
them to then formulate their design upon. The data provided by the simulation model
can have a wide spectrum of use cases with multiple possible permutations and
combinations.

35
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

6.1 ADAPTATION FOR DIFFERENT MODELS

The model can be adjusted and implemented in various case scenarios for multiple
models. Following are certain models identified in Fig 30, the list of parameter
adaptation are listed on table 6.1.

Fig 30: Adaptation in different models | source: Author

TABLE 6.1 ADAPTAION PARAMETERS | SOURCE: AUTHOR

36
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

6.2 FURTHER REFINEMENT OF THE MODEL

Given the confidence level of the model and potential use cases, it presents the
opportunity for further refinement of the model. The model is still in nascent state and
has to be tested more extensively with different test scenarios to understand the gaps
better. The model also can be refined further to account for various parameters like
climatic factors, comfort level inside the spatial geometry, optimization of openings,
more complex social behavior of crowd simulation, inclusion for social interaction,
demographic based effect on walking patterns and speed, temporal effects of usage,
improvement measures for existing transit nodes, affect of illumination levels on
pedestrian flow just to name a few. The study is being carried forward for further
refinement and perusal.

7. CONCLUSION
The investigation of "Geometry augmentation for spatial development to optimize
pedestrian flow and circulation." provides a baseline & a framework for a genetic
algorithm-based simulation model for interventions in transit nodes, and successfully
demonstrates potential for further exploration and refinement. It performs significantly
better than human input design and has a high enough confidence value for execution
and implementation.

37
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

8. REFERENCES
Construction. Robotics and Automation in Construction. Board, T. R. (2022). Transit
Capacity and Quality of Service Manual, Third Edition (with supplemental material on
CD-ROM).

Hwang, Lee, & Kim. (2019). Real-Time Pedestrian Flow Analysis Using Networked
Sensors for a Smart Subway System. Sustainability, 11(23), 6560.

Patra, M., Sala, E., & Ravishankar, K. (2017). Evaluation of pedestrian flow
characteristics across different facilities inside a railway station. Transportation
Research Procedia, 25, 4763–4770.

9. BIBLIOGRAPHY
Appleyard, B. S., Taecker, M., Allen, C., Armusewicz, C., & Schroder, C. (2016).
Livable Transit Corridors: Methods, Metrics, and Strategies. Nap Edu.

Balaguer, C., & Abderrahim, M. (2008). Trends in Robotics and Automation in

Dell’Asin, G., & Hool, J. (2018). Pedestrian Patterns at Railway Platforms during
Boarding: Evidence from a Case Study in Switzerland. Journal of Advanced
Transportation, 2018, 1–11.

Gabbana, A., Corbetta, A., & Toschi, F. (2022). Modeling Routing Choices in
Unidirectional Pedestrian Flows. Collective Dynamics, 6, 1.

Hwang, Lee, & Kim. (2019). Real-Time Pedestrian Flow Analysis Using Networked
Sensors for a Smart Subway System. Sustainability, 11(23), 6560.

Interactive, W. (n.d.). Presidents Medals: The City Synthesiser.

http://www.presidentsmedals.com/Entry-53931
Jia, L. (2016). Modeling and Simulation of Passenger Flow Distribution in Urban Rail
Transit Hub Platform. Preprints.

38
Department of Architecture & Planning, BIT Mesra.
 Naritya Choudhary | Dissertation 2022

Khattak, A., Jiang, Y., Zhu, J., & Hu, L. (2017). A new simulation-optimization
approach for the circulation facilities design at urban rail transit station. Archives of
Transport, 43(3), 69–90.

Khattak, A., Yangsheng, J., & Abid, M. M. (2018). Optimal Configuration of the Metro
Rail Transit Station Service Facilities by Integrated Simulation-Optimization Method
Using Passengers’ Flow Fluctuation. Arabian Journal for Science and Engineering,
43(10), 5499–5516.

Khattak, A., Yangsheng, J., Lu, H., & Juanxiu, Z. (2017). Width Design of Urban Rail
Transit Station Walkway: A Novel Simulation-Based Optimization Approach. Urban
Rail Transit, 3(2), 112–127.

Ning, J., Zhou, Y., Long, F., & Tao, X. (2018). A synergistic energy-efficient planning
approach for urban rail transit operations. Energy, 151, 854–863.

Patra, M., Sala, E., & Ravishankar, K. (2017). Evaluation of pedestrian flow
characteristics across different facilities inside a railway station. Transportation
Research Procedia, 25, 4763–4770.

Shah, J., Joshi, G., & Parida, P. (2013). Behavioral Characteristics of Pedestrian Flow
on Stairway at Railway Station. Procedia - Social and Behavioral Sciences, 104, 688–
697.

Zhang, H., He, B., Lu, G., & Zhu, Y. (2022). A simulation and machine learning based
optimization method for integrated pedestrian facilities planning and staff assignment
problem in the multi-mode rail transit transfer station. Simulation Modelling Practice
and Theory, 115, 102449.

Zhang, Z., & Jia, L. (2020). Direction-Decision Learning Based Pedestrian Flow
Behavior Investigation. IEEE Access, 8, 15027–15038.

39
Department of Architecture & Planning, BIT Mesra.
 GEOMETRY AUGMENTATION FOR
SPATIAL DEVELOPMENT TO OPTIMIZE
PEDESTRIAN FLOW AND CIRCULATION
By Naritya Choudhary

WORD COUNT 5843 TIME SUBMITTED 02-DEC-2022 01:23PM

PAPER ID 93419674
GEOMETRY AUGMENTATION FOR SPATIAL DEVELOPMENT
TO OPTIMIZE PEDESTRIAN FLOW AND CIRCULATION
ORIGINALITY REPORT

5%
SIMILARITY INDEX

PRIMARY SOURCES

1 en.wikipedia.org
Internet 81 words —  1%
2 www.mdpi.com
Internet 62 words —  1%
3 www.gcu.pdx.edu
Internet 38 words —  1%
4 zh.scribd.com
Internet 27 words —  < 1%
5 nap.nationalacademies.org
Internet 26 words —  < 1%
6 www.coursehero.com
Internet 26 words —  < 1%
7 xfer.services.ncdot.gov
Internet 14 words —  < 1%
8 scholar.sun.ac.za
Internet 10 words —  < 1%
9 Hwang, Lee, Kim. "Real-Time Pedestrian Flow
Analysis Using Networked Sensors for a Smart
8 words —  < 1%
Subway System", Sustainability, 2019
 Crossref

10 Van-Minh Le, Yann Chevaleyre, Ho Tuong Vinh,

Jean-Daniel Zucker. "Hybrid of linear programming
6 words —  < 1%
and genetic algorithm for optimizing agent-based simulation.
Application to optimization of sign placement for tsunami
evacuation", The 2015 IEEE RIVF International Conference on
Computing & Communication Technologies - Research,
Innovation, and Vision for Future (RIVF), 2015
Crossref

EXCLUDE QUOTES ON EXCLUDE SOURCES OFF

EXCLUDE BIBLIOGRAPHY ON EXCLUDE MATCHES OFF