An Automated Crowd Management in Public Transport Using

Online Ticketing System

Krithikaa Venket V. S.

, S. Leelavathy

, S. Muthuselvan

, Rajes Kannan S.

and K. Rajakumari

Department of Information Technology, KCG College of Technology, KCG Nagar, Rajiv Gandhi Salai, Karapakkam,

Chennai – 600 097, Tamil Nadu, India

Department of Artificial Intelligence and Data Science, Panimalar Engineering College, Bangalore Trunk Road,

Varadharajapuram, Poonamallee, Chennai – 600 123, Tamil Nadu, India

Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, India

Keywords: Machine Learning, Overlapping Routes, ETA, Crowd Density Data Driven Decision Making, IR Sensor,

Passenger Counts, Route Demand.

Abstract: Due to the rapid increase in population, the usage of public transportation like local buses is increasing day

by day, especially in metropolitan cities, particularly in India, one densely populated country in the world.

Due to this, we are in need of proper crowd management in public transport, so we have developed automated

crowd management in public transport (MTC BUS) through an IR sensor integrated with an online ticket

booking system. Due to the unavailability of the estimated time of arrival (ETA) of the bus coming into the

bus stop for their destination, a high-demand route is congested with a high number of passengers when

compared with a low-demand route at a similar or slight variation of the ETA. This leads to improper usage

of public transport and an uneven distribution of crowd density. So, the crowd in the bus is estimated by fitting

an IR sensor in the door of the bus for calculating passenger count and classifying them into levels based on

crowd density, and the ETA is calculated based on the Machine learning algorithm. Decision-making on

allocating bus route numbers is based on ETA and demand for the buses on the selected route. All these

processes are connected with the online ticket booking system.

1 INTRODUCTION

Urban bus systems, an essential element of public

transportation, face many issues that detract from

passenger experience and system efficiency. One of

the most critical problems is over passenger

crowding, which affects comfort, safety and

operational efficiency. First, more crowded buses

lead to lower comfortability of passengers and also

potential safety concerns (Anju et al., 2022).

Crowding obstructs movement, blocks access to exit

doors, and raises the odds of accidents or injuries.

Second, the long ticket selling time at busy stops

greatly delays the bus time. But long queues of

passengers result in delayed departures, which in

turn reduces efficiency in all parts of the network

(Adithi S et al., 2022). Thirdly, increased congestion

and inefficient passenger flow due to uncontrolled

boarding. No separate queues or entry points, chaotic

boarding coupled with the passengers slowing down

the crowding process. Next, and more critically, the

conventional conductor-heavy ticketing systems,

especially in locations with high passenger volume,

are detrimental to operational efficiency. Fourthly,

boarding patterns are not coordinated (Adline Freeda

et al., 2016). Routes are often underutilized, as

passengers will board the next available bus. That

results in skewed demand across the routes some

buses are full while others sit a few seats empty,

compounding congestion on certain routes. All of

these challenges must be addressed in order to

improve the overall service and ridership of bus

systems. In this paper, we present a new

methodology to address as well as improve the public

transport journey for end-users (Ved Prakash Mishra

et al., 2019). Hence, ETA refers to the estimated time

of arrival of a vehicle to its destination.

V. S., K. V., Leelavathy, S., Muthuselvan, S., S., R. K. and Rajakumari, K.

An Automated Crowd Management in Public Transport Using Online Ticketing System.

DOI: 10.5220/0013926200004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 5, pages

259-265

ISBN: 978-989-758-777-1

259

2 RESEARCH OBJECTIVES

2.1 Towards a Dynamic and Efficient

Bus System

Build, and train a highly accurate machine learning

model to predict bus ETAs for individual routes.

Integrate historic data from previous trips with live

traffic statistics, and route dependent factors like

conditions like weather, and flow of construction. 1-

Analyze the data for different aspects and categorize

the bus routes into either 1-5 demand levels based on

the analysis. In the topical identification phase,

relevant data, including passenger volume, peak hour

variations, geographical distribution, and others, must

be considered (Yu, B., Yang, Z., Yao, J., 2010).

Implement an algorithm that is capable of providing

real-time recommendations on the best routes to take

for passengers. Use predicted ETAs, route demand

categories, and, if available, passenger preferences, as

well as live bus locations. Optimize passenger wait

times and throughput by favoring routes with less

demand and lower ETAs Create a display for

passengers and onboard buses at bus stops that shows

this information to promote transparency and

informed decision making when planning travel (S.

Rajaprakash et al., 2020).

3 RELATED WORK

3.1 Machine Learning for ETA

Prediction

Deep learning models like LSTMs have shown

promise in predicting passenger boarding/exiting

patterns and improving ETA accuracy (Gandhi et al.,

2023). Ensemble methods combining different

models like Random Forests and SVMs can further

enhance accuracy (Liu, R., Li, S., Sun, L., Li, F., &

Sun, Z. 2017).

3.2 Demand-Based Route Allocation

Dynamic optimization algorithms considering real-

time demand and bus locations have achieved

significant reductions in passenger waiting times

(Zhang et al., 2021). Decentralized multi-agent

systems offer flexible route assignment based on local

information.

3.3 Crowd Density Sensing

Infrared sensors provide accurate passenger count

data within buses (S. Muthuselvan et al., 2015).

Computer vision algorithms using cameras can

estimate crowd density and passenger behavior for

real-time monitoring (Patil, S. A., and Soman, S. S.

2017).

4 METHODOLOGY

4.1 Data Collection and Processing

Historical Data: Collect historical data on bus speed,

distance, and Estimated Time of Arrival (ETA) for

various routes (Correia et al., 2014). Pre-process the

data to ensure its quality and consistency. Real-Time

Data: Implement a system to gather real- time speed

data using GPS sensors on buses. Integrate infrared

(IR) sensors at bus stops to count passengers entering

and exiting. Store the real-time speed and passenger

count data in a Firebase Real-time Database for

immediate access.

4.2 Machine Learning Model for ETA

Prediction

Train a machine learning model on historical data to

predict bus arrival ETAs based on real-time speed

information. Utilize simulated speed data during the

training process to mimic real-life GPS data behavior

(Meghana Sarode et al., 2020). Deploy the trained

model to continuously generate predicted ETAs for

each bus stop and update them in the Firebase

database.

4.3 Demand Category Classification

Define a Demand Category classification system for

routes based on the following:

Sequence of Bus Stops: Analyze the sequence of

total bus stops before the source stop and the overall

sequence within the route. Higher sequence numbers

indicate potentially higher passenger volumes.

Assign lower Demand Category values (e.g.,

Category 1) to routes with lower passenger volumes

and higher values (e.g., Category 5) to routes with

higher volumes.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

260

4.4 Route Allocation Algorithm

• Develop an algorithm that considers the

following parameters when allocating routes

to buses:

• Demand Category: Prioritize routes with

lower Demand Category values to minimize

passenger crowding.

• Predicted ETA Difference: If multiple routes

have the same Demand Category, compare

predicted ETAs at the source stop (Mr.

Kruthik Gandhi H A et al., 2023).

• Select a route with a predicted ETA

difference below a pre-defined threshold

(e.g., 5-10 minutes) from the average

predicted ETA.

• Passenger Count: If there is a tie between the

demand category then use this passenger

count crowd level to decide if the best route

exists or there are no significant ETA

differences, consult the real-time passenger

count data.

• Allocate the route with a passenger count

below a specific threshold to distribute

passenger load across routes.

String crowd Level = "";

if (count >= 1 && count <= 10)

{

crowdLevel = "Seats Available";

}

else if (count >= 11 && count <= 20)

{

crowdLevel = "Less Crowded";

}

else if (count >= 21 && count<= 30)

{

crowdLevel = "Crowded";

}

else if (count >= 31 && count <= 40)

{

crowdLevel = "Highly Crowded";

}

else

{

crowdLevel = "Highly Dense

Crowded";

}

4.5 System Implementation

Integrate the data collection modules (GPS and IR

sensors) and the route allocation algorithm into a

central system using a Raspberry Pi Pico W or similar

microcomputer. Utilize the Firebase Real-time

Database to store and access the collected and

predicted data for real-time decision-making.

4.6 Evaluation and Testing

Evaluate the effectiveness of the system by

comparing passenger wait times, overcrowding

levels, and overall system efficiency before and after

implementation (Sri Sindhuja Selvanayaki P et al.,

2018). Conduct simulations and field tests to validate

the accuracy of the ETA prediction model and the

overall functionality of the route allocation algorithm.

Figure 1: Points Scored by the Different Bus Routes.

To achieve highly accurate passenger inflow and

outflow data, we propose a multi-sensor approach at

each bus door's entry/exit points. Infrared (IR)

sensors will be strategically positioned to ensure

unobstructed views and protection from potential

damage or external interference. Additionally, by

employing more than one or two sensors in a

combined configuration, we can significantly

enhance passenger detection accuracy. Figure 1

shows that the points scored by the different routes.

This redundancy will mitigate the limitations of

individual sensors and account for potential

occlusions caused by luggage, backpacks, or closely

following passengers. Data Collection: Continuously

collect real-time passenger count data from each

sensor and update under each route and bus stops in

firebase real time database, ensuring data points are

time stamped for accurate timing and

synchronization. Implement data quality checks to

identify and handle potential anomalies or sensor

malfunctions.

Algorithm for Crowd Level Classification:

Establish data-driven passenger count thresholds for

each crowd level, considering: Average passenger

capacity of buses in your system. Comfort and safety

considerations for passengers. Potential variations in

passenger size (e.g., luggage, standing vs. seated)

4.7 Widget for Ticket Booking App

Displaying the assigned route ID, predicted ETA, and

crowd level information (e.g., "Level 3: Crowded") in

a widget within the ticket booking app. This

An Automated Crowd Management in Public Transport Using Online Ticketing System

261

information helps passengers make informed

decisions about route selection and travel time.

4.8 E-Ticketing and Passenger Flow

Data Collection

To expedite boarding and gather accurate passenger

data, we will utilize e-tickets displayed on

smartphones or electronic devices. These e-tickets

will showcase a scannable barcode (e.g., QR code), a

clear Ticket ID, and validity information. Passengers

will present e-tickets for scanning or visual

verification at entry points. Validated tickets grant

access, while invalid one’s trigger alerts. The e-

ticketing system automatically captures passenger ID,

boarding time, and entry point for analysis and route

optimization. This system can be integrated with the

IR sensor-based passenger counting system to

enhance accuracy, validate data, and generate

comprehensive passenger flow reports. The below

table 1 and table 2 shows that the SRT Tools and

Tidel Park stoppings respectively.

5 RESULTS AND DISCUSSION

Table 1: Route Details with Their ETA for Tidel Park

Stopping.

Serial No Route ID

Bus

Sto

ETA

Tidel

Par

7 minutes

Tidel

Par

9 minutes

Table 2: Route Details with Their ETA for SRP Tools

Stopping.

Serial

Route

Bus Stop ETA

102 S.R.P.Too

12 minutes

570 S.R.P.Too

10 minutes

19 S.R.P.Too

14 minutes

95 S.R.P.Too

9 minutes

Logic simulation of minimum ETA if the selected

source has more than one entry with ETA than give

least ETA routes id information and if there is tie or

it mean equal ETA in the database than return the

routeid information based on least value of category

entry among five routes collection than return than

only routeid information field in database if the

selected source has only one (Sharma D. et al.,2016).

The below figure 2 shows that the journey ticket

booking in public transport.

Figure 2: Journey Ticket Booking in Public Transport.

Decision making of allocating route id for same

ETA for same source on different route id

Example 1:

Source = S.R.P.Tool, ETA"6"

category"1"

source"S.R.P.Tool" route 91 and

another ETA"6"

category"2"

source "S.R.P.Tool" rout95

In this same ETA which the allocation is based on

category which has least value as per logic it is

routeid91 has 1 which is least

Decision making of allocating route id for

difference in wide range of ETA for same source on

different route id

Example 2:

Source ="Kandhanchavadi" ETA"11"

Category"2" 95 routeid

source"Kandhanchavadi"

And another ETA"8"

Category"1" 91 routeid

source "Kandhanchavadi" a

another ETA"58" Category"4" 570 routeid

source "Kandhanchavadi"

The allocation of route id is 91 which has least

ETA 3. Decision making of allocating route id for

selected source has only has single routeid

Example 3:

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

262

source=parrys ETA"1"

category "5"102 routeid

source “Parrys" allocated route id is

102

The output of widget displaying ETA, Route ID,

Crowd Level shows in the figure 3

Figure 3: Output for Identification of Crowd Level.

The E_Ticket Display with Ticket ID and Barcode is

shown in this figure 4.

Figure 4: E_Ticket Display with Ticket ID and Barcode.

Accurately classifying crowd density with sensor

data demands meticulous attention to detail. The type

and placement of sensors play a crucial role, with bi-

directional infrared detectors strategically positioned

at entry/exit points offering reliable passenger count

data. Robust algorithms trained on diverse datasets

further enhance accuracy and adaptability. Studies

showcase promising results, with some exceeding

90% accuracy in real-time estimation using sensors

and computer vision. Figure 5 gives the device

activation image.

Nonetheless, there are limitations that must be

addressed. Sensor readings can also suffer from noise

due to passenger behaviors, luggage sizes, and

movement patterns (D. Darsena et al., 2017).

Downtime due to external factors that can hinder the

sensor or disrupt data interpretation like obstacles,

lighting and environmental noise.

Figure 5: Device Activation.

The sensor data mainly shows the number of

passengers in the bus, but lacks information on their

behavior (standing vs. sitting). Additionally,

inaccurate classifications can result for certain

demographics or scenarios due to biases in training

data or algorithm design.

Nonetheless, it is important to note that sensor-

based crowd density classification could still be

extremely valuable if tailored intelligently (Lavanya

S et al., 2017). Following proper sensor selection and

arrangement, using powerful algorithms trained on

abundant and diverse data, and doing everything to

take into main consideration the impact of external

factors and limitations, this can turn out to be a strong

tool to manage public spaces and maximize

efficiency.

6 CONCLUSIONS

Increased convenience and reduced frustration:

Improved bus arrival time estimates (ETAs) allow

better planning and reducing waiting times at bus

stops, creating a better and more predictable travel

experience. Greater comfort and safety: details of

crowd numbers in real time allow passengers to board

less crowded buses, improving comfort and reducing

the risk of safety. Informed travel decisions: Access

to ETA, demand and crowd data allows passengers to

make decisions on which bus to take or a different

mobility option, optimizing their journey. Balanced

system and reduced overcrowding: Allocating routes

according to demand categories (1-5) can lead to a

more balanced system with passenger supply

matching routes capacity to improve efficiency and

An Automated Crowd Management in Public Transport Using Online Ticketing System

263

reduce crowding. IR sensor-based passenger counting

system is assisted with to improve the accuracy of the

system.

Optimized route-planning– Through real-time data,

operators can understand demand per route, thus

deploying buses themselves and efficiently

optimizing operational costs, as well as reducing

congestion. Better resource utilization: Real-time

passenger volume data allow operators to deploy only

those buses that are in demand, preventing empty or

overcrowded vehicles and effective resource

utilization.

Utilizing historical data and real-time data

analysis for strategic decisions (route expansion, fare

changes, and even improving infrastructure)

promotes a data-driven decision-making method of

public transportation management.

6.1 Additional Potential Results

Environmental benefits: Optimization of routes and

reduction of congestion can result in reduced fuel

consumption and lower emissions, therefore leading

to a cleaner environment. More members on the bus:

A more reliable, comfortable, data-driven bus can win

over newcomers, and encourage existing riders to

ride more often, promoting sustainable transport

choices. Increase the public perception: the machine-

centric system which utilize real-time information to

improve travel experience can enhance public’s view

about public transportation, provoking its more

extensive use.

6.2 Limitations

Although our selected model showed promising

results in predicting ETA (bus arrival time), we

believe there can be a better approach. Advancing to

more complex Models such as XGBoost or design

specialized Sequential models such as LSTMs, &

Extending the Data Set further back in time to include

weather patterns & special events are a few of the

ways in which we could hone in more accuracy and

address for corner cases. And while bi-directional

infrared sensors are good for reliable passenger

counts, they also have limitations such as possible

interference and lack of information about individual

behavior. Exploring other technologies, such as

computer vision with depth cameras or radar sensors,

may help create more complex datasets and help with

more nuanced crowd characterization, such as

separating standing from seated passengers. Such

deeper insights can facilitate dynamic capacity

modifications and further ensure enhanced resource

allocation. Our ongoing efforts towards increasing the

accuracy of our model and the most insightful ways

of data collection we can achieve will allow us to

further optimize the performance of our system so

we can provide an even more time- effective and

comfortable experience for end users and public

transport operators.

6.3 Future Work

Our work highlights the promise of real-time data and

automated systems in improving public

transportation experiences. However, continuous

development is crucial to maximize its impact. Here

are key areas for future exploration

6.4 Enhanced ETA Prediction

Advanced Models: Consider the benefits of

integrating XGBoost, LSTMs, or ensemble

approaches to provide a more resistant and precise

arrival time forecast, accommodating unusual cases

and intricate circumstances. Grow Data Sources:

Introduce more periods including climatic patterns,

exceptional events and real-time traffic conditions to

supply a fuller picture about possible delays in a

specific area make model further adaptability. Hyper

parameter tuning: Once the model has been chosen,

hyper parameters for the model should be optimized

and updated over time to maximize performance

against particular conditions within your system.

6.5 Advanced Crowd Characterization

More than a head count: They also can go beyond

just counting passengers by using alternative

technology, such as computer vision coupled with

depth cameras or radar sensors. This would help

collect richer data sets in terms of enhanced crowd

characterization, for example passenger movement

patterns, real-time availability of seats and even

standing vs seated passengers. Granular insight: Use

these richer data sets to develop a better

understanding of crowd behaviour in buses. This

data could also be leveraged to enable dynamic

adjustment with respect to capacity, where resources

are targeted toward specific areas or passengers are

informed via mobile how to avoid congestion and

utilize alternate routes for a smoother journey.

By tackling these constraints and following up on

these exciting future directions, we can further

improve and advance our proposed system. This

will ultimately lead to a more efficient, reliable,

comfortable public transportation experience for all.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

264

REFERENCES

A. Anju, Barath. Maheshwaran, M. R, V. K and K. K. S R,

Sentimental Analysis for E-Commerce Website, 2022

10th International Conference on Emerging Trends in

Engineering and Technology - Signal and Information

Processing (ICETET-SIP- 22), Nagpur, India, 2022, pp.

1-4, doi: 10.1109/ICETET-SI2254415.2022.9791606.

Adithi S, Mahanth Sai M, Dhriti ruth Rajanna, Rekha.N, K

Rishika Ravi, Crowd Management Framework for

Departure Control in Bus Transport Service using

Image Processing, April 2022, Volume 8 Issue 11,

ISSN: 2349-6002 International Journal of Innovative

Research in Technology.

Adline Freeda, R., Sharmila, R.N (2016), A review of bulk

data dissemination protocols for reprogramming in

WSN, ICICES 2016,2016,7518937 International

conference on Information Communication and

Embedded Systems

Correia, A. I. C., Coelho, M. C., & da Silva, S.N. (2014).

Passenger Demand Management in Public Transport

Systems: A Review. Transportation Planning and

Technology, 37(1), 1-21.

D. Darsena et al., Enabling and Emerging Sensing

Technologies for Crowd Management in Public

Transportation Systems: A Review (2017),

Transportation Research Procedia, Volume 25.

Lavanya S. Rani and Gayathri Binu, Smart Information

System for Public Transportation Using IoT (2017),

International Journal of Advanced Research in

Computer Science and Software Engineering, Volume

7, Issue 3.

Li, W., Yang, M., & Liu, Y. (2020). Bus arrival time

prediction with ensemble learning methods.

Transportation Research Part C: Emerging

Technologies, 111, 87-100

Liu, R., Li, S., Sun, L., Li, F., & Sun, Z. (2017). A Multi-

Sensor Approach for Passenger Counting in Public

Buses. IEEE Transactions on Intelligent Transportation

Systems, 18(8), 2277-2288.

Meghana Sarode et al., Automated Crowd Management in

Bus Transport Service (2020), Journal of Intelligent

Transportation Systems, Volume 24, Issue 5.

Mr. Kruthik Gandhi H A, Mr. Jerrin Joy, Dr. Udayabalan

Balasingam, Mr. Manish G Automated Bus Crowd

Management (2023), International Journal for Research

in Engineering Application & Management (IJREAM)

Vol-08, Issue-10, Jan 2023.

Patil, S. A., and Soman, S. S. (2017). Machine Learning for

Intelligent Transportation Systems: A Survey. Artificial

Intelligence Review, 49(1), 105- 138.

Rajaprakash S, Jaishanker N, Chan Bagath Basha, S

Muhuselvan, Athira Jayan, Aswathi A.B, sebastian,

Ginu, RBJ20 Cryptography Algorithm for Securing Big

Data Communication Using Wireless Networks,

Lecture Notes in Networks and Systems, Volume 334,

Pages 499 - 5072022 5th World Conference on Smart

Trends in Systems, Security and Sustainability, WS4

2021Virtual, Online29 July 2021.

S. Muthuselvan, and SomaSundaram K, A survey of

sequence patterns in data mining techniques,

International Journal of Applied Engineering Research,

Volume 10, Issue 1, Pages 1807 - 1815, 2015.

S. Rajaprakash, C. Bagath Basha, S. Muthuselvan,

Jaisankar N, Ravi Pratap Singh, RBJ25 cryptography

algorithm for securing big data, Journal of Physics:

Conference Series, Volume 1706, Issue 122 December

2020.

Sharma D. et al., A Review on Technological Advanceme

nts in Crowd Management System (2016), Internation

al Journal of Computer Applications, Volume 136,

Issue 1.

Sri Sindhuja Selvanayaki P, Yuvaraj N, An Extensive

Survey on IoT Architecture & Machine Learning

Algorithms for Crowd Detection in Public

Transportation (2016), International Journal for

Scientific Research & Development, Vol. 6, Issue 01,

2018.

Ved Prakash Mishra, Amna Rafi Chaudhry, Kajal Shah

Surname, Model for Crowd Distribution in Public

Transport Buses, ISSN: 2278-3075, Volume-8, Issue-

7C2, May 2019 International Journal of Innovative

Technology and Exploring Engineering (IJITEE)

Yu, B., Yang, Z., Yao, J., 2010. Genetic algorithm for bus

frequency optimization. Journal of Transportation

Engineering 136, 576-583.

Zhang, S., Li, J., Yang, Y., & Wang, Y. (2021). An efficient

bus route optimization algorithm based on dynamic

programming. IEEE Transactions on Intelligent

Transportation Systems, 22(10), 6405- 6415.

Zhang, Y., Li, Z., Yang, L., Lv, Y., & Chen, X. (2018).

Real-Time Bus Arrival Time Prediction with Bus

Bunching Consideration. IEEE Transactions on

Intelligent Transportation Systems, 19(11), 3392- 3402.

An Automated Crowd Management in Public Transport Using Online Ticketing System

265