VIDYAVARDHAKA COLLEGE OF ENGINEERING

Autonomous institution affiliated to

Visvesvaraya Technological University, Belagavi

EcoSweep: Automated Waste Segregation and

Garbage Collection Robot

A Project Report submitted in partial fulfillment for the award of degree

BACHELOR OF ENGINEERING
IN
ELECTRONICS AND COMMUNICATION ENGINEERING
BY
INCHARA SATHISH S [4VV21EC053]
VARSHA S [4VV21EC172]
RAKSHITHA M [4VV22EC406]
MAHADEVA PRASAD K M [4VV22EC413]

Under the guidance of

SAHANA M S
Assistant Professor

DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

Accredited by NBA, New Delhi
VIDYAVARDHAKA COLLEGE OF ENGINEERING
Accredited by NAAC with A ‘Grade’
GOKULAM III STAGE, MYSURU-570 002
2024-25
 DEPARTMENT OF ELECTRONICS AND COMMUNICATION
ENGINEERING

CERTIFICATE

Certified that the project work entitled “Eco Sweep: Automated Waste Segregation and
Garbage Collection Robot” is a bonafide work carried out by Inchara Sathish S –
4VV21EC053, of Vidyavardhaka College of Engineering, in partial fulfillment for the
award of degree of Bachelor of Engineering in Electronics and Communication
Engineering of the Visvesvaraya Technological University, Belagavi during the year 2024-
2025.

It is certified that all corrections/suggestions indicated for internal assessment have been
incorporated in the report. The project report has been approved as it satisfies the academic
requirements in respect of Project work prescribed for the Bachelor of Engineering Degree.

Sahana M S Dr. Chandrashekar M Patil Dr. B Sadashive Gowda

Assistant Professor HoD Principal

External Viva

Name of the Examiners Signature with date

1 ………………………... ………………………

2 ………………………... ………………………
 DEPARTMENT OF ELECTRONICS AND COMMUNICATION
ENGINEERING

CERTIFICATE

Certified that the project work entitled “Eco Sweep: Automated Waste Segregation and
Garbage Collection Robot” is a bonafide work carried out by Varsha S – 4VV21EC172,
of Vidyavardhaka College of Engineering, in partial fulfillment for the award of degree of
Bachelor of Engineering in Electronics and Communication Engineering of the
Visvesvaraya Technological University, Belagavi during the year 2024-2025.

It is certified that all corrections/suggestions indicated for internal assessment have been
incorporated in the report. The project report has been approved as it satisfies the academic
requirements in respect of Project work prescribed for the Bachelor of Engineering Degree.

Sahana M S Dr. Chandrashekar M Patil Dr. B Sadashive Gowda

Assistant Professor HoD Principal

External Viva

Name of the Examiners Signature with date

1 ………………………... ………………………

2 ………………………... ………………………
 DEPARTMENT OF ELECTRONICS AND COMMUNICATION
ENGINEERING

CERTIFICATE

Certified that the project work entitled “Eco Sweep: Automated Waste Segregation and
Garbage Collection Robot” is a bonafide work carried out by Mahadevaprasad K M –
4VV22EC406, of Vidyavardhaka College of Engineering, in partial fulfillment for the
award of degree of Bachelor of Engineering in Electronics and Communication
Engineering of the Visvesvaraya Technological University, Belagavi during the year 2024-
2025.

It is certified that all corrections/suggestions indicated for internal assessment have been
incorporated in the report. The project report has been approved as it satisfies the academic
requirements in respect of Project work prescribed for the Bachelor of Engineering Degree.

Sahana M S Dr. Chandrashekar M Patil Dr. B Sadashive Gowda

Assistant Professor HoD Principal

External Viva

Name of the Examiners Signature with date

1 ………………………... ………………………

2 ………………………... ………………………
 DEPARTMENT OF ELECTRONICS AND COMMUNICATION
ENGINEERING

CERTIFICATE

Certified that the project work entitled “Eco Sweep: Automated Waste Segregation and
Garbage Collection Robot” is a bonafide work carried out by Rakshitha M –
4VV22EC413, of Vidyavardhaka College of Engineering, in partial fulfillment for the
award of degree of Bachelor of Engineering in Electronics and Communication
Engineering of the Visvesvaraya Technological University, Belagavi during the year 2024-
2025.

It is certified that all corrections/suggestions indicated for internal assessment have been
incorporated in the report. The project report has been approved as it satisfies the academic
requirements in respect of Project work prescribed for the Bachelor of Engineering Degree.

Sahana M S Dr. Chandrashekar M Patil Dr. B Sadashive Gowda

Assistant Professor HoD Principal

External Viva

Name of the Examiners Signature with date

1 ………………………... ………………………

2 ………………………... ………………………
 Declaration

We the members of the project team, studying in the VIII semester of Electronics and
Communication Engineering, Vidyavardhaka College of Engineering, hereby declare
that the entire project entitled “EcoSweep: Automated Waste Segregation and

Garbage Collection Robot” has been carried out by us independently under the
guidance of Sahana M S, Department of Electronics and Communication Engineering,
Vidyavardhaka College of Engineering. This project work is submitted to the
Visvesvaraya Technological University, Belagavi, in partial fulfillment of the
requirement for the award of the degree of Bachelor of Engineering in Electronics and
Communication Engineering during the academic year 2024-2025.

This dissertation has not been submitted previously for the award of any other degree
or diploma to any other institution or university.

Date:

Place: Mysuru

Name of the Students USN Signature

1. Inchara Sathish S 4VV21EC053

2. Varsha S 4VV21EC172

3. Mahadeva Prasad K M 4VV22EC406

4. Rakshitha M 4VV22EC413

i
 Acknowledgement
The satisfaction that accompanies the successful completion of any task would be
incomplete without the mention of people who made it possible and whose constant
guidance and encouragement crowned our efforts with success. We consider our
privilege to express the voice of gratitude and respect to all those who guided us and
inspired us in completion of this project.

First and foremost, we ought to pay due regards to our renowned institution, which
provided us a platform and an opportunity for carrying out this project work.

We wish to express our gratitude to Dr. B Sadashive Gowda, Principal, VVCE, for
providing congenial working environment.

We are thankful to Dr. Chandrashekar [Link], Professor and HOD, Dept. of ECE,
VVCE, for motivating us and also allowing us to use the facilities of the department
to complete this project successfully.

We express our sincere thanks to our guide Sahana M S, Assistance Professor, Dept
of ECE, VVCE for their constant co-operation, support, and invaluable suggestions.

We would like to thank our parents for their constant moral support throughout the
completion of this project.

Finally, last but not the least, we would like to extend our deep sense of gratitude to
our friends who always inspired us and encouraged us throughout the completion of
this project.

1. Inchara Sathish S

2. Varsha S

3. Mahadeva Prasad K M

4. Rakshitha M

ii
 CONTENTS
List of Figures v
List of Tables vi
Abstract vii
Chapter - 1 Introduction 1-4
1.1 Background 1
1.2 Introduction 2
1.3 Motivation 3
1.4 Objective 4
Chapter - 2 Literature Survey 5-10
2.1 Sensor-Based Waste Segregation Systems 5
2.2 IoT-Integrated Waste Monitoring 6
2.3 Machine Vision and Deep Learning for Waste Classification 6-7
2.4 Robotic Automation in Waste Handling 7-8
2.5 Vision-Based Classification Using Machine Learning 8
2.6 Robotic Integration in Waste Collection 8-9
2.7 Limitations in Existing Systems 9
2.8 Summary and Technological Gap 9-10
Chapter - 3 Proposed Methodology 11-14
3.1 Waste Collection and Initial Handling 11
3.2 Visual Detection and Classification Using YOLO 12
3.3 Model Training and Performance 12
3.4 Robotic Arm Integration for Waste Sorting 12
3.5 Feedback Mechanisms and Error Handling 13
Data Management and System Monitoring Eco Sweep records
3.6 13
detailed metadata for every processed item
3.7 Summary of Methodological Workflow 14
3.8 Waste Classification and Localization 14
3.9 Robotic Sorting and Handling 14

iii
Chapter - 4 Hardware and Software Requirements 15-26
4.1 Hardware Requirements 15-21
4.1.1 ESP32 15-18
4.1.2 HC-05 Bluetooth Module 18-19
4.1.3 L298N MOTOR DRIVER 20-21
4.1.4 Robotic Arm 21
4.1.5 Laptop 22
4.2 Software Requirements 22-26
4.2.1 YOLO Model 22-23
4.2.2 ESP-IDF Framework 24
4.2.3 HC-05 Bluetooth communication protocol 24-25
4.2.4 Blynk – IoT applications 25-26
Chapter - 5 Results 27
Chapter - 6 Conclusion and Future Work 28-29
6.1 Conclusion 28
6.2 Application 28
6.3 Future Work 29
References 30-31
Appendix 32-35
List of publications 36

iv
 List of Figures
Figure Caption Page
No. No.
3 Block diagram of Methodology 11
4.1 Components circuit connection 15
4.1.1 ESP32 Pinout 16
4.1.2 HC-05 Bluetooth Module 18
4.1.3 L298N MOTOR DRIVER 20
4.1.4 MK1 Robotic Arm 21
4.2 4.2.1 Image Classifications 23
4.2.4 Blynk IoT mobile app Logo 26
5.1 Final Model 27
5.2 Pick and place the garbage in respective 27
Bins
5.3 Robotic arm picks the garbage 27

v
 List of Tables

Table Title Page

No. No.
4.1 L298N Motor Driver Pinout 20

vi
 ABSTRACT

For sustainable garbage management and environmental protection, separation of

garbage at the source steps to be effective. The project presents a robot known as Eco
Sweep. It is a fully automated garbage segregation and collection robot that makes use of
technology like machine learning, Internet of things and robotics to improve efficiency
along with accuracy. The system is based on using the YOLO (You Only Look Once)
machine learning model to live plastic, metal and glass recycling. Garbage image
processing runs on the laptop and classification data is sent to an ESP32 microcontroller
with the HC-05 Bluetooth module. The robotic system does all litter collection and
separator work and is certainly devoid of manual intervention. Through the Blynk IoT
platform, this garbage-mobile robot detects garbage and stops its movement also used to
deploy a robotic arm to pick up and put the trash in a bin. Automation saves human effort,
being faster and more accurate in sewage classification, especially when it comes to
garbage disposal. Wireless communication, real-time monitoring and IoT-implemented
control mechanisms make it a smart smart management solution.
Keywords: Sustainable garbage management, automated waste segregation,
YOLO, machine learning, IoT, robotics, ESP32, HC-05, Blynk platform. Plastic, metal,
glass detection, image processing, real-time monitoring, smart waste collection, robotic
arm, wireless communication. Garbage classification, source-level separation, mobility
robot, automation in disposal, environmental protection.

Salient Features: -
➢ AI-Driven Classifications: Fast and Smart Waste Identification using YOLO model
➢ Waste Collection Robotic Suite: Automated sorting and disposal with robotic arm +
mobile platform.
➢ Wi-Fi: ESP32 and Bluetooth module for on-the-fly data exchange.
➢ Remote Monitoring & Control: The IoT app Blynk focuses on live updates expanding
from as far away as you can be.

vii
Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

CHAPTER 1
EcoSweep: Automated Waste Segregation and
Garbage Collection Robot
INTRODUCTION

1.1 BACKGROUND

One of the major issues faced by today’s urban areas is managing waste efficiently,
especially as the amount and variety of household and industrial waste continue to grow.
Traditional manual methods of waste segregation are labour-intensive, unhygienic, and
inefficient, often resulting in improper handling and disposal. These issues have led to the
need for automation in waste segregation and collection systems, especially in the context
of smart city development.

To address these challenges, the Eco Sweep project introduces an automated waste
segregation and garbage collection robot that utilizes modern technologies such as machine
learning (YOLO), embedded systems (ESP32), Bluetooth communication (HC-05), robotic
automation, and IoT (Blynk platform). Designed for autonomy, the system sorts waste into
specific categories such as biodegradable, recyclable, and non-recyclable with minimal
human input.

The concept combines real-time object detection using the YOLO model, communication
between a laptop and ESP32 microcontroller via Bluetooth, and the physical operation of
a robotic arm mounted on a mobile robot. This robot navigates using IR or ultrasonic
sensors to detect waste, and the robotic arm performs the pick-and-place task based on the
classification data. A mobile application (Blynk) facilitates remote control and system
monitoring, ensuring user accessibility and real-time feedback.

This integrated system offers a promising solution for smart waste management, improving
hygiene, operational efficiency, and environmental sustainability.

Dept. of ECE, VVCE, Mysuru 1

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

1.2 INTRODUCTION

The global surge in population and industrial growth has resulted in an alarming increase
in the generation of solid waste. As urban areas expand and consumerism rises, cities are
producing more waste than ever before. Proper waste management is crucial in mitigating
the environmental impacts of this [Link] its importance, sorting waste at the source
is frequently neglected in waste management practices. When waste is not separated into
biodegradable, recyclable, and non-recyclable categories, it becomes difficult to manage it
effectively, leading to overburdened landfills and missed recycling opportunities.

Most traditional systems for sorting waste are based heavily on manual effort. These
methods are not only labor-intensive but also pose significant health hazards due to direct
exposure to toxic or infectious [Link] involvement in waste sorting can lead to
errors, which negatively impact the accuracy of classification and decrease recycling
efficiency.

To overcome these challenges, technology has emerged as a key driving force. The
convergence of Artificial Intelligence (AI), Machine Learning (ML), robotics, and Internet
of Things (IoT) has opened new possibilities in automating waste handling systems. These
smart technologies enable machines to detect, classify, and sort waste without human
intervention, making the process more efficient, reliable, and safer.

In this paper, we present EcoSweep, a comprehensive solution for robotic waste sorting.
EcoSweep leverages the YOLO (You Only Look Once) deep learning model for real-time
image classification of waste materials. The system combines a mobile robot, a robotic arm,
an ESP32 microcontroller, Bluetooth communication, and a mobile application for real-
time control and monitoring. The overarching goal is to provide a fully autonomous, eco-
friendly solution to modern-day waste management problems

Key Points:

• Rising urban waste requires smarter segregation techniques.

• Manual waste handling is risky, slow, and inaccurate.

• Automation using AI and IoT can revolutionize waste management.

• EcoSweep integrates classification, communication, and robotic execution.

Dept. of ECE, VVCE, Mysuru 2

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

1.3 Motivation

Waste segregation remains one of the most significant bottlenecks in achieving

efficient and sustainable waste management. Despite awareness campaigns and
government policies, improper disposal and classification of waste persist, especially in
developing countries. The motivation behind this project is rooted in the urgent need to
automate waste handling processes, reduce human risk, and optimize recycling practices.

Manual segregation is still widely practiced, especially in densely populated urban

areas. Waste collectors often deal with a hazardous mix of biodegradable and non-
biodegradable materials, including medical waste, broken glass, and toxic chemicals. This
exposes workers to health risks and leads to inefficient recycling due to contamination of
waste streams.

Furthermore, with growing interest in smart cities, there is a rising demand for
intelligent infrastructure. Automated waste segregators not only align with the vision of
smart urban ecosystems but also reduce operational costs, enhance labor productivity, and
contribute to environmental preservation.

The EcoSweep system addresses these concerns by introducing automation, real-

time waste identification, robotic segregation, and remote monitoring—all in a compact,
cost-effective prototype. The motivation is not just to improve technology but to create a
practical solution for a cleaner, safer, and greener society.

Key Points:

• Manual waste sorting is unsafe and inefficient.

• Urban growth increases the demand for automated systems.
• Smart cities require intelligent, sustainable infrastructure.
• EcoSweep contributes to cleaner cities and safer work environments.

Dept. of ECE, VVCE, Mysuru 3

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

1.4 Objective

The main objective of the EcoSweep project is to design, develop, and implement a fully
automated robotic system that can detect, classify, and segregate waste into appropriate
categories with minimal human involvement. The project aims to showcase how modern
technologies can be applied practically to address one of the oldest urban challenges—
waste management.

Specific objectives include:

1. Real-time Waste Detection: Utilize the YOLO machine learning model to detect
and identify objects as paper, plastic, metal, or glass from images captured via a
camera.
2. Wireless Data Transfer: Transmit classification data to the ESP32 microcontroller
using Bluetooth communication (HC-05 module) for seamless processing.
3. Autonomous Movement: Develop a mobile robot capable of navigating towards
waste, stopping on detection, and positioning itself for collection.
4. Robotic Waste Handling: Integrate a mechanical robotic arm to pick up the
identified waste and place it in the appropriate bin based on the classification result.
5. IoT Monitoring and Control: Implement a mobile application using the Blynk
IoT platform to allow users to control and monitor the system in real time.
6. System Optimization: Ensure high accuracy, minimal latency, and energy
efficiency to make the system suitable for practical use in public and private spaces.

By fulfilling these objectives, the project not only serves as an academic prototype but also
lays the groundwork for scalable deployment in smart city environments, educational
campuses, industrial zones, and residential communities.

Dept. of ECE, VVCE, Mysuru 4

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

CHAPTER 2
LITERATURE SURVEY
The challenge of managing increasing amounts of urban and industrial waste has motivated
extensive research into automated waste segregation systems. These systems have evolved
from simple sensor-based models to intelligent, vision-enabled platforms capable of real-
time classification and robotic handling. This literature survey presents a critical analysis
of existing systems, categorizing them by technology used, and highlights their limitations.
It also establishes the basis for proposing a more advanced and integrated solution, such as
EcoSweep, which combines machine learning, IoT, and robotics to address multiple
shortcomings found in past efforts.

2.1 Sensor-Based Waste Segregation Systems

Initial research in the field of automated waste segregation largely focused on sensor-based
techniques. These methods made use of inductive sensors (for metal detection), moisture
sensors (to distinguish wet from dry waste), capacitive sensors, and infrared (IR) sensors.
Systems using such hardware were cost-effective and relatively easy to implement,
particularly in small-scale applications.

Deeksha More et al. [4]The system includes a prototype that employs IR and metal sensors
to separate waste into wet, dry, and metallic categories. While effective in theory, this
system struggled to maintain accuracy when dealing with irregular waste shapes, multi-
material objects, or waste contaminated with moisture or grease. Additionally, these
systems lacked the intelligence to distinguish between materials with similar physical
properties, such as plastic and glass, making the segregation process prone to error.

Similarly, Sharanya et al. [9] designed a robotic arm-based system using proximity and
moisture sensors to detect and separate materials. This setup proved useful in low-traffic
areas like office bins but was not suitable for large-scale or dynamic environments due to
its limited sensor resolution and difficulty in adapting to real-world waste variation.

These sensor-dependent models often failed to handle composite materials or packaging

that did not exhibit strong inductive or moisture signals. Thus, although they automated
part of the sorting process, their real-world performance was limited by the type and
complexity of the waste stream.

Dept. of ECE, VVCE, Mysuru 5

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

2.2 IoT-Integrated Waste Monitoring

With the growing focus on smart infrastructure, IoT (Internet of Things) has become
a significant addition to waste management systems. Rather than only sorting waste, these
systems aimed to monitor the fill levels of bins, track waste collection frequency, and send
remote alerts for optimized collection.

Jessie R. Balbin et al. [1] introduced an IoT-enabled smart bin that tracked waste
volume and transmitted the data to a central monitoring app. While this increased the
efficiency of waste collection services, it did not address the problem of waste segregation
at the source, still leaving manual intervention necessary for sorting.

Margaret Richardson Ansah et al. [3] enhanced the system by integrating solar
power and LoRaWAN communication for low-power, wide-area remote access. Their
system allowed real-time waste level monitoring in remote or off-grid locations, making it
suitable for rural areas. However, similar to Balbin's model, it lacked the ability to
automatically classify the waste content.

Although these IoT systems contributed significantly to improving waste

monitoring and operational efficiency, they were not capable of actual segregation.
Therefore, the need remained for solutions that could combine real-time classification with
smart connectivity.

2.3 Machine Vision and Deep Learning for Waste Classification

Recent advances in machine vision and deep learning algorithms have dramatically
improved the potential of automated waste sorting. These systems use Convolutional
Neural Networks (CNNs) and object detection frameworks like YOLO (You Only Look
Once), SSD (Single Shot Detector), and Faster R-CNN to visually identify and classify
different types of waste based on images captured by a camera.

Nimisha S. Gupta et al. [10] explored the use of deep learning models to categorize
metal, plastic, and glass waste using real-time image data. Their results showed impressive
accuracy under controlled lighting conditions, especially when using high-performance
models like Inception-ResNet v2. However, their setup required a powerful GPU, making
it expensive and less suitable for lightweight mobile platforms.

Dept. of ECE, VVCE, Mysuru 6

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

Surendra Kumar Koganti and his team [11] implemented a Raspberry Pi-based
smart waste classifier using deep learning. Although their system significantly reduced
human input, it faced issues with processing speed and accuracy degradation in low-light
or outdoor conditions. Moreover, their model struggled to identify overlapping waste or
semi-transparent materials such as certain plastics.

Another promising approach was presented by Md. Nafis Raihan et al., [16] who
created a robot with an onboard camera, image detection algorithm (SSD Lite), and robotic
arm to identify and pick up garbage items. Despite good initial results, the system needed
further refinement in handling varying object sizes and integrating with IoT platforms for
large-scale use.

These machine learning-driven solutions marked a major leap in waste recognition

accuracy but highlighted the importance of efficient model training, system robustness, and
lightweight deployment for real-world applications.

2.4 Robotic Automation in Waste Handling

While many systems focused on classification, others explored robotic

manipulation to automate the physical handling of waste.

Ajay V P et al. [6] created a semi-automated system involving a conveyor belt and
multiple sensors to push items into designated bins. Though functional, it lacked
adaptability and required fixed infrastructure, limiting its deployment in mobile or outdoor
environments.

Ionut-Robert Badoi et al. [7] developed a GPS-enabled mobile robot with a camera,
microcontroller, and robotic arm. Their system could navigate autonomously, detect waste,
and sort it using a gripper. It performed well in identifying plastic and metal objects, but its
classification of paper was less reliable due to color and texture similarities with other
materials.

Chander Partap Singh et al. [8] focused on integrating waste bins with RFID and
ultrasonic sensors to monitor bin fill levels and sort waste types. However, the system
required pre-labeled waste and could not operate independently in uncontrolled settings.

Dept. of ECE, VVCE, Mysuru 7

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

These robotic systems proved that mechanical waste handling can be automated
effectively. Still, they depend heavily on accurate classification inputs—without which the
robotic actions lose their precision.

2.5 Vision-Based Classification Using Machine Learning

Despite the promising results of previous studies, many existing systems face
significant limitations:

• Inadequate classification for mixed waste: Sensor-based methods cannot reliably

distinguish between multi-material waste, such as juice boxes or laminated paper.
• Power-hungry vision systems: Deep learning models like YOLO and R-CNN need
significant computational power, which can be a limitation on mobile or embedded
platforms.
• Environmental sensitivity: Systems may work well in labs but falter under real-
world lighting, weather, and waste variability.
• Manual preprocessing: Some systems still rely on humans to place waste items
properly or pre-sort them into limited categories.
• Scalability challenges: Solutions designed for small bins or controlled settings don’t
scale easily to city-level or outdoor implementations.

2.6 Robotic Integration in Waste Collection

Several researchers have attempted to enhance waste segregation systems by
integrating robotic arms and mobile robotic platforms to automate the pick-and-place
process.

Ajay V P et al. [6] built a robotic system with a conveyor belt and proximity sensors,
designed to segregate metal and non-metal waste. While this system enhanced speed and
repeatability, it remained constrained by its reliance on proximity and moisture readings
rather than advanced classification.

Ionut-Robert Badoi et al. [7] developed a mobile robotic unit equipped with a
camera, ultrasonic sensors, and a robotic arm for collecting and sorting waste. They
achieved an overall classification accuracy of 81%, with particularly strong results for
metal and plastic detection. However, performance for paper waste lagged, and the system
required environmental calibration.

Dept. of ECE, VVCE, Mysuru 8

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

Chander Partap Singh et al. [8] focused on a smart bin that used actuators and
sensors for sorting but lacked advanced vision systems. Their work primarily contributed
to real-time monitoring and automation rather than high-accuracy classification.

These systems confirm that robotic manipulation is essential for real-world

implementation of automated waste segregation. However, without accurate classification
inputs from vision models, robotic systems are limited in their effectiveness.

2.7 Limitations in Existing Systems

Despite substantial progress, several limitations remain in current waste management

technologies:

• Limited classification accuracy: Sensor-based systems struggle to identify

complex, mixed, or ambiguous materials.
• High power requirements: Vision-based deep learning models often need
powerful processors, which increases cost and limits portability.
• Environmental sensitivity: Many systems underperform in outdoor or
unstructured settings due to lighting, weather, or background variations.
• Manual intervention: Most systems still require some degree of human
involvement—either in feeding waste or interpreting data.
• Scalability issues: Systems designed for controlled environments often fail when
deployed in dynamic or high-volume areas.

2.8 Summary and Technological Gap

After evaluating a wide range of existing technologies, it is evident that while many
systems address parts of the waste management process—such as monitoring, sorting, or
robotic collection—few offer a complete end-to-end automated solution. Most are limited
either by detection accuracy, sensor range, cost, or operational flexibility.

Dept. of ECE, VVCE, Mysuru 9

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

The EcoSweep system aims to bridge this gap by:

• Combining YOLO-based vision for real-time classification.

• Using a robotic arm for autonomous pick-and-place actions.
• Integrating with the ESP32 microcontroller and Bluetooth for communication.
• Enabling remote monitoring via the Blynk IoT platform.
• Providing a scalable, adaptable platform for use in homes, schools, offices, and
public areas by learning from the limitations of prior systems and blending the
strengths of ML, robotics, and with the integration of IoT, EcoSweep delivers an
efficient, secure, and sustainable method for tackling modern waste management
issues.

Key Points:

• Past systems lacked real-time waste classification.

• IoT-based bins improved monitoring but not segregation.
• Robotic arms were underutilized or lacked intelligence.
• EcoSweep advances previous efforts by fully automating the entire process.

Dept. of ECE, VVCE, Mysuru 10

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

CHAPTER 3
SYSTEM METHODOLOGY

Fig.3: Block diagram of Methodology

The growing volume of municipal solid waste in urban areas demands smarter,
more efficient methods of segregation to improve recycling and reduce landfill
dependency. Traditional sensor-based segregation systems rely heavily on inductive,
capacitive, and moisture sensors, which often suffer from limited material coverage,
inaccurate classification, and high maintenance.
Eco Sweep introduces an advanced approach to waste segregation using computer
vision and robotic automation. By integrating the YOLO (You Only Look Once) deep
learning model with a precision robotic arm, Eco Sweep provides a highly accurate and
scalable solution for real-time waste detection, classification, and physical sorting. This
section outlines the complete methodology applied in the Eco Sweep system to achieve
efficient, contactless waste segregation.
3.1. Waste Collection and Initial Handling

The waste segregation process begins with the collection and delivery of waste
onto a moving platform, typically a conveyor belt. Waste items, such as plastic, metal,
and glass, are loaded onto the belt either manually or through an automated feeder.

Key considerations at this stage include:

• Ensuring uniform distribution of waste to avoid object overlap.

• Using an industrial-grade conveyor to maintain a steady speed.
• Installing a vibration mechanism or guiding rails to spread the waste for optimal
camera visibility

Dept. of ECE, VVCE, Mysuru 11

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

3.2. Visual Detection and Classification Using YOLO

The core of Eco Sweep’s intelligence lies in the application of YOLO, a deep
convolutional neural network (CNN) designed for real-time object detection.

3.2.1 Image Acquisition

• A high-definition camera mounted above the conveyor captures continuous

frames.
• Frames are passed to the onboard processing unit, which could be powered by a
GPU enabled embedded system (e.g., NVIDIA Jetson Nano or Xavier).

3.2.2 Image Processing and Detection

• The system processes each frame using the YOLOv5 or YOLOv8 model,
pretrained and finetuned on a custom waste classification dataset.
• YOLO detects individual items and assigns bounding boxes along with class
probabilities for categories such as: Plastic, Metal and Glass.

3.3 Model Training and Performance

• The model is trained on thousands of annotated waste images under varied

lighting and background conditions.
• Data augmentation techniques (e.g., rotation, blur, contrast adjustments) are
applied to improve generalization.
• YOLO’s single-stage detection architecture allows for high-speed inference,
making it suitable for real-time applications.

3.4. Robotic Arm Integration for Waste Sorting

Once the YOLO model classifies the waste, the positional data (bounding box centre and
class) is transferred to the robotic control unit.

3.4.1 Coordinate Mapping

• The image coordinate system is translated into physical coordinates on the

conveyor using a calibration matrix.
• This mapping allows the robotic arm to locate the exact position of the object.

Dept. of ECE, VVCE, Mysuru 12

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

3.4.2 Arm Movement and Object Handling

• A 6-degree-of-freedom (6-DOF) robotic arm is used to provide the flexibility

required for Page 2 of 5 precision picking.
• The arm is equipped with an adaptive gripper or suction-based end-effector,
depending on the object type.
• Once the object is grasped, it is deposited into one of several designated
segregation bins, each labelled for a specific waste category.

3.4.3 Motion Planning

• A motion planning algorithm, such as Inverse Kinematics with collision

detection, ensures the robotic arm moves smoothly without interfering with
nearby objects or itself

3.5. Feedback Mechanisms and Error Handling

To improve reliability, Eco Sweep incorporates real-time feedback and error-

handling mechanisms:

• Weight sensors in bins confirm successful object placement.

• If an item cannot be confidently classified or grasped, it is redirected to a manual
inspection bin.
• Feedback from failed detections is logged and used to retrain the model for
continuous performance improvement. This closed-loop system minimizes
classification errors and supports long-term scalability.

3.6. Data Management and System Monitoring Eco Sweep records

detailed metadata for every processed item:

• Waste type
• Confidence level of classification
• Timestamp
• Robotic arm action status

This data is stored locally and can be uploaded to a cloud-based dashboard for remote
access and analytics.

Dept. of ECE, VVCE, Mysuru 13

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

3.7. Summary of Methodological Workflow

Stage Description

Input Handling Waste fed via conveyor, camera captures images

Detection YOLO detects and classifies waste items in real-time Robotic Control Robotic
arm uses coordinate mapping for precise picking Sorting Waste placed into category-
specific bins Stage Description Feedback System logs action, handles uncertain cases
Monitoring Data stored and optionally uploaded for remote access

3.8. Waste Classification and Localization

After detection, each item of waste is:

• Assigned a category label (e.g., plastic, metal).

• Located precisely using bounding box coordinates.
• The coordinates are converted from image space to real-world coordinates using a
camera calibration model. This allows the system to know exactly where on the
conveyor each item is located.

3.9. Robotic Sorting and Handling

This Is where the intelligent robotic arm comes into play:

• A robotic arm with a gripper or suction device receives the positional data of the
waste item.
• Using inverse kinematics and motion planning algorithms, the arm moves to the
location of the object.
• The arm picks up the item and places it into the correct bin (e.g., glass bin, plastic
bin). Features:
• Precision picking using coordinate mapping.
• Gripper adjustment for objects of different shapes and weights.
• Safety and collision detection to avoid errors. This process happens in real-time
and allows for continuous operation with minimal human oversight.

Dept. of ECE, VVCE, Mysuru 14

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

CHAPTER 4
HARDWARE AND SOFTWARE REQUIREMENTS
For a system to work successfully, its primary predetermined gear are its software
program and its hardware necessities. They make certain that the gadget and its working
context are well matched. overall performance is maximised and unexpected failure is
reduced when positive standards are met. They make price estimates and efficient resource-
focused making plans viable. while requirements are explicit, troubleshooting is short and
simple. these are critical to improvement fulfilment due to the fact they're scalable and
maintainable.

4.1 Hardware Requirements:

Fig. 4.1: Components circuit connection

4.1.1 ESP32
With an ESP32, the universal microcontroller system-on-a-chip built on a single
multicore processor with the internet of things-friendly peripherals, loads of memory and
integrated Wi-Fi as well as Bluetooth is indeed a complete deal of wonder and great
flexibility, low cost, and very reliable along with ultra-low power consumption. It is a
multipurpose product with dual-core processing, big memory, and many peripheral options
at disposal. These devices provide more capabilities and superior power efficiency +
flexibility than the majority of other microcontrollers.

ESP32 microcontroller is more user-friendly. It can be programmed in different

ways because it supports various programming environments like Arduino IDE, LUA,
MicroPython, JavaScript, Espressif IDF, and PlatformIO IDE.

Dept. of ECE, VVCE, Mysuru 15

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

Fig.4.1.1: ESP32 Pinout

Micro-USB, 5V, three.3V, and GND power pins:

• strength is furnished to the USB port through Micro-USB.

• This pin receives regulated 5V, that is alternatively regulated to 3.3V via the on-
board regulator, in general to supply energy to the board.
• The improvement board is powered by means of 3.3 volts which might be regulated
to this pin.
• GND stands for floor pin.

Enable (En): The microcontroller may be reset the use of this permit pin.

Analogue pins: Analogue pins, which measure analogue voltage between 0 and three.3V,
are ADC1_0 to ADC1_5 and ADC2_0 to ADC2_9.

Records conversion from virtual to analogue is accomplished thru the DAC pins (DAC1 &
DAC2).

T0 through T9 are capacitive contact pins, which might be typically used for capacitive
pads.

The number one characteristic of the RTC GPIO pins (RTCIO0 thru RTCIO17) is to evoke
the development board from deep sleep state.

Serial (Tx, Rx): Serial statistics is despatched and obtained via these pins.

Outside Interrupts (All GPIO): An interrupt may be because of any GPIO pin.

Dept. of ECE, VVCE, Mysuru 16

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

PWM (All GPIO): Any GPIO pin may be configured to operate as PWM at some stage in
the program.

VSPI (GPIO23 (MOSI), GPIO18 (CLK), GPIO5 (CS), & GPIO19 (MISO)): those pins are
mostly used for SPI-1 verbal exchange. PWM has 16 independent channels.

The number one motive of HSPI (GPIO13 (MOSI), GPIO14 (CLK), GPIO15 (CS), and
GPIO12 (MISO)) pins is SPI-2 verbal exchange.

IIC or I2C (GPIO21(SDA), GPIO22(SCL)): I2C communique is the primary motive for
these pins.

AREF: The number one purpose of those pins is to provide a reference voltage for enter
voltage.

Specifications of ESP32 Development Board

The specifications of ESP32 include the following.
• It functions a 32-bit LX6 microprocessor with a clock frequency of as much as 240
MHz, which may be both unmarried-core or twin-middle.
• It has 520 KB of SRAM, 448 KB of ROM, and sixteen KB of RTC SRAM.
• It has as much as a 150 Mbps of 802.11 wi-fi connectivity.
• it's miles compatible with both BLE and classic Bluetooth v4.2.
• It has as much as 18 channels of 12-bit SAR ADC, 2 channels of 8-bit DAC, and
sixteen channels of LED PWM in addition to 34 programmable GPIOs.
• It has 4 SPI, three UART, two I2C, and I2S serial connectors.
• It has one host controller for SD/MMC/SDIO and one slave controller for
SPI/SDIO. Its Ethernet MAC is generally used for bodily LAN communique.
• Its capabilities flash encryption and a secure boot.
• This microcontroller ordinarily helps AES, Hash, RSA, RNG, and ECC
cryptographic hardware acceleration.

How Does ESP32 Work?

To lessen the communication stack at the number one software processor, the
ESP32 improvement board can function as a slave tool for a bunch MCU or as a completely
independent gadget. via its interfaces, together with SPI/SDIO (or I2C/UART), this board
can quite simply engage with other structures to provide Wi-Fi and Bluetooth capability.

Dept. of ECE, VVCE, Mysuru 17

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

The Arduino UNO and an ESP8266 are not quite as good as the ESP32
improvement board. There are two fundamental methods this board can talk with the
outside global: stressed and wi-fi wireless. Bluetooth and wi-fi are principles that may be
used to provide wireless communique. 802.11 b/g/n/e/i WLAN MAC is functional in this
ESP32 board.

4.1.2 HC-05 Bluetooth Module

Bluetooth serial modules allow all serial enabled devices to communicate with each
other using Bluetooth.

Fig. 4.1.2: HC-05 Bluetooth Module

It has 6 pins,
1. Key/EN: It is used to bring Bluetooth module in AT commands mode. If Key/EN pin is
set to high, then this module will work in command mode. Otherwise by default it is in data
mode. The default baud rate of HC-05 in command mode is 38400bps and 9600 in data
mode.
HC-05 module has two modes,
1. Data mode: Exchange of data between devices.
2. Command mode: It uses AT commands which are used to change setting of HC-
05. To send these commands to module serial (USART) port is used.
2. VCC: Connect 5 V or 3.3 V to this Pin.
3. GND: Ground Pin of module.
4. TXD: Transmit Serial data (wirelessly received data by Bluetooth module transmitted
out serially on TXD pin)
5. RXD: Receive data serially (received data will be transmitted wirelessly by Bluetooth
module).
6. State: It tells whether module is connected or not.

Dept. of ECE, VVCE, Mysuru 18

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

Specification of HC-05 Bluetooth Module

• Bluetooth 2.0 with EDR (better data rate)
• 2.4 GHz ISM band frequency
• Class 2 (up to 4 dBm) transmit energy; GFSK (Gaussian Frequency Shift Keying)
modulation
• ordinary sensitivity: -80 dBm
• Supported profiles include SPP (Serial Port Profile), concealed (Human Interface
tool), and others.
• Range: more or less 10 meters (33 ft) in open air.
• Running voltage variety: DC three.3V to 5V
• Fewer than 50mA for working current, fewer than 2.5mA for standby current, and
less than 1mA for sleep cutting-edge.
• Interface: UART (established Asynchronous Receiver/Transmitter)
• Baud quotes: 1200, 2400, 4800, 9600, 19200, 38400, 57600, 115200, 230400, and
460800.
• Working temperature: -20°C to 75°C (-4°F to 167°F)

ESP32 Development Board interfacing with HC-05 Bluetooth

Through this interface, the HC-05 Bluetooth textual content may be read and printed
using an ESP32 on a serial show. right here, the ESP32 board receives the records thru
Bluetooth from the telephone and shows it above the serial reveal. to connect to the
Bluetooth module, you need to first download and set up the Bluetooth terminal app in your
telephone. using the Bluetooth app, facts is despatched from the phone.

The ESP32 WROOM and the HC-05 Bluetooth Module are the 2 primary elements
had to create this interface. considering that HC-05 Bluetooth is usually utilised for wi-fi
verbal exchange with smartphones, it's miles crucial to this interface. some AT instructions
may be used to trade the default Bluetooth module settings. The Bluetooth module's logic
stages could be 3.3 V for TX/RX, and it operates among 3.6 and 6 V. there may be no need
to change the TX/RX voltage level for the HC-05 because the ESP32 development board
can hit upon a 3.3V level.

Dept. of ECE, VVCE, Mysuru 19

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

4.1.3 L298N MOTOR DRIVER:

DC and stepper vehicles may be pushed with the aid of this high-electricity L298N
motor driving force module. This module is made of a 78M05 5V regulator and an L298
motor driving force IC. as much as four DC motors or two DC automobiles with path and
pace control may be controlled through the L298N Module.

Fig.4.1.3 L298N MOTOR DRIVER

L298N Module Pinout Configuration

Table 4.1: L298N Motor Driver Pinout

Pin Name Description

IN1 & IN2 Motor A input pins. Used to control the spinning direction of Motor A

IN3 & IN4 Motor B input pins. Used to control the spinning direction of Motor B

ENA Enables PWM signal for Motor A

ENB Enables PWM signal for Motor B

OUT1 & OUT2 Output pins of Motor A

OUT3 & OUT4 Output pins of Motor B

12V 12V input from DC power Source

5V Supplies power for the switching logic circuitry inside L298N IC

GND Ground pin

Dept. of ECE, VVCE, Mysuru 20

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

Features & Specifications

• Model of driving force: L298N 2A
• The Double H Bridge L298N driving force chip, the maximum motor deliver
voltage of 46V, the most motor deliver contemporary of 2A, and the common-sense
voltage of 5V.
• 5-35V is the driver voltage.
• Maximum power (W): 25W; driving force modern: 2A.
• Logical current: 0-36mAevery motor's modern-day sense.
• A heatsink for improved performance; and an energy-on LED indicator

4.1.4 Robotic Arm:

With its joints, motors, and segments, a robotic arm is a sort of mechanical arm that
features in addition to a human arm and may flow and carry out operations which include
selecting, putting, turning, and lifting objects.

Fig.4.1.4: MK1 Robotic Arm

Crucial factors:

1. Actuators
o Servo automobiles
2. The manage machine
o ESP32 microcontroller.
o Strength source: external battery p.c. or energy financial institution

3. Writing code

o ESP-primarily based systems the use of Python

4. Sensors (not required)

Cameras (for vision or obstacle detection)

Dept. of ECE, VVCE, Mysuru 21

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

5. Application: object sorting and choosing

4.1.5 Laptop
The laptop serves as the main processing unit inside the waste segregation gadget.
It makes use of a system studying version known as YOLO (You only look once) to method
pictures in real time and classify waste into 3 categories: recyclable, biodegradable, and
non-recyclable. The pc's camera takes snap shots of the waste substances, which the YOLO
model then interprets to identify the kind of waste. After type is completed, the computer
uses the HC-05 Bluetooth module to ship the results to the ESP32 microcontroller. This
makes it viable for the robot elements to respond in keeping with the class. The computer's
computational capability ensures brief processing and unique detection, making it an
essential a part of enabling the system's whole automation.

4.2 Software Requirements:

4.2.1 YOLO Model:
YOLO (You Only Look Once) is an actual-time object identity machine that
surpasses conventional two-level detectors in velocity and efficiency by predicting
bounding packing containers and sophistication chances for items in a picture the usage of
a unmarried convolutional neural network (CNN).

Here's a greater thorough analysis:

Key factors of YOLO:

• Unmarried-level detection: YOLO processes the whole photo in an unmarried pass,

instead of two-stage detectors like R-CNN, which want distinct tiers for classification
and region idea.
• Grid-primarily based approach: A grid is constituted of the enter picture, and the
version predicts a specific number of bounding containers and sophistication
possibilities for each grid cellular.
• Pace and performance: YOLO’s real-time object detection design makes
it appropriate to be used in surveillance structures and self-
using motors, amongst different applications. YOLOv1, YOLOv2, YOLOv3,
YOLOv4, YOLOv5, YOLOv6, YOLOv7, YOLOv8, and YOLOv11 are
the numerous iterations of YOLO that have been produced, and everyone
has progressed dealing with of small items, accuracy, and velocity.

Dept. of ECE, VVCE, Mysuru 22

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

• Ultralytics YOLO: Ultralytics is an enterprise that specialises in developing and

refining YOLO fashions and affords a range of resources and tools for practitioners
and researchers.
• YOLO11: this is the maximum recent new release of the YOLO version, and because
of its optimised structure and anchor-loose layout, it gives extremely
good performance and accuracy gains.
How YOLO Works:

Fig.4.2.1: Image Classifications

1. Input photo: An image is fed into the algorithm.
2. Grid department: A grid of cells is created from the image.
3. Prediction: The model forecasts the subsequent for every mobile:
• The probability that an object is in that cell.
• The item's bounding field coordinates (x, y, width, and peak).
• The object's class.
4. Output: to symbolize the diagnosed gadgets within the image, the model produces a
hard and fast of bounding containers with corresponding magnificence possibilities.
5. YOLO applications: real-Time Video Analytics: YOLO makes it viable to analyse
video content material in actual time.

Dept. of ECE, VVCE, Mysuru 23

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

4.2.2 ESP-IDF Framework

For creating apps on Espressif's ESP32, ESP32-S, ESP32-C, and ESP32-H range
of SoCs the usage of C and C++, the reputable, all-inclusive framework is known as ESP-
IDF, or Espressif IoT improvement Framework.

Here's a breakdown in more element:

• The goal of ESP-IDF is to offer an unbiased software improvement kit (SDK) for
developing apps on Espressif's internet of things SoCs to users.
• It is well matched with the programming languages C and C++.
• Goal structures: The ESP32, ESP32-S, ESP32-C, and ESP32-H households of SoCs are
the ones for which it's far intended.
• Functions: To expedite the development system, ESP-IDF comes with a toolchain,
APIs, components, and techniques.
• The framework referred to as ESP-IDF is open supply.
• Use instances: It makes it feasible to create a wide variety of community-connected
items, ranging from fundamental net of factors gadgets to commercial makes use of.

4.2.3 HC-05 Bluetooth communication protocol:

The Bluetooth 2.0 + EDR (better records charge) protocol is used by the HC-05
Bluetooth module for wi-fi communication. A UART (generic Asynchronous
Receiver/Transmitter) interface is used for serial records alternate. It makes use of GFSK
(Gaussian Frequency Shift Keying) modulation and operates within the 2.4 GHz ISM band.
Further info:

• Bluetooth Protocol: The Bluetooth 2.0 + EDR well-known serves as the inspiration
for the HC-05 module. In comparison to the authentic Bluetooth 2.0, this shows that it
offers more suitable information charge (EDR) capabilities, permitting quicker
statistics switch fees.
• Serial conversation (UART): The module uses a UART interface to connect to other
gadgets, like computer systems or microcontrollers. This interface makes connections
and statistics transfers easy by sending facts serially, step by step.
• Baud fee: To adjust the facts switch velocity, the module helps a number of baud fees,
including 9600, 19200, 38400, and others. typically, the default baud fee is 9600.
• The commercial, medical, and scientific (ISM) band, which is often utilised for wireless
conversation, is in which the HC-05 capabilities in phrases of frequency and

Dept. of ECE, VVCE, Mysuru 24

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

modulation. It makes use of Gaussian Frequency Shift Keying (GFSK) modulation,

which shifts the frequency of a provider wave to encode records onto it.
• Slave/master Mode: so, one can initiate or respond to connection requests, the HC-05
can be set up to function in both grasp and slave mode.
• AT commands: a set of AT commands transmitted thru the serial interface may be used
to alter the module's setup and settings.
• Variety: In outdoor, the HC-05's normal variety is up to 10 meters, or 33 feet.
• Power and contemporary: The module is suitable for battery-powered applications
because it typically runs on a voltage variety of three.3V to 5V DC and uses little
contemporary.

4.2.4 Blynk – IoT applications

With the assist of the low-code IoT platform Blynk, you could create and regulate
cellular applications for controlling linked devices. It affords no-code functionality, a drag-
and-drop app builder, and some of capabilities like OTA updates and tool provisioning. an
internet portal for dealing with customers, devices, and information is any other function
of the Blynk platform.

Vital attributes:

• No-code App Builder: Use a drag-and-drop interface to create and alter cellular apps
without knowing any code.
• Communication: helps cell (GSM, 2G, 3G, 4G, LTE), Ethernet, WiFi, and LoRaWAN,
amongst different communication selections.
• Device management: This category covers features inclusive of OTA firmware
upgrades, WiFi provisioning, and tool provisioning.
• Protection: gives comfortable statistics shipping the use of give up-to-quit encryption.
• Scalability: Made to accommodate IoT tasks of all sizes, from small-scale pilots to large
rollouts.
• Blynk.360 Console: an online console for OTA updates, device, consumer, and facts
controls, in addition to different commercial enterprise operations.
• One packaged solution that makes it simpler to connect devices to the Blynk platform
is referred to as [Link].
• [Link]: Updates related devices' firmware over the air.
• Automations: set up alerts and push notifications as well as different product
automations.

Dept. of ECE, VVCE, Mysuru 25

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

• Community: assets and guide are supplied by means of a good-sized person and
development network.

How it operates:

1. join gadgets: to connect your devices (such as an Arduino or Raspberry Pi) to the
Blynk cloud, use the hardware libraries supplied through Blynk.
2. Create the App: Create the cell interface for coping with and maintaining an eye on
your devices the use of the drag-and-drop app builder.
3. three. App Deployment: submit the app to Google Play or the App store.
4. manage devices: you could manipulate your devices, users, and data the usage of
the Blynk console.

Fig.4.2.4: Blynk IoT mobile app Logo

Dept. of ECE, VVCE, Mysuru 26

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

CHAPTER 5
RESULT
The mobile robot navigates to the point of waste and comes to a halt for collection, while
the YOLO model running on a laptop communicates smoothly with the ESP32
microcontroller using the HC-05 Bluetooth module. Such wireless data transfer enables
smooth coordination between the machine learning model and the hardware. Additionally,
the Blynk IoT platform consists of an easy interface through which individuals can view
and manage the entire system using a mobile phone in remote areas. Waste Segregation
System reflects intensive technological development with the application of the YOLO
model in real-time for waste classification. It effectively distinguishes between the waste
as biodegradable, recyclable, and non-recyclable by name with more efficiency and lower
manual handling.

The system is well mechanized with less human involvement and maximum efficiency in
work. There are some challenges, particularly in optimizing the YOLO model to categorize
mixed waste types with great accuracy and to maintain reliability across varying
environmental conditions. These challenges suggest improvement areas such as model
scalability improvement, hardware resilience guarantee, and the training process
optimization for handling more complicated and diversified waste inputs. Overcoming
these limitations will be crucial to scaling the deployment of the system and ensuring long-
term sustainability and accuracy in practice.

Fig.5.1: Final Model

Fig.5.2. Pick and place the garbage in respective Fig 5.3. Robotic arm picks the garbage

bins

Dept. of ECE, VVCE, Mysuru 27

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

CHAPTER 6
CONCLUSION AND FUTURE WORK
6.1 Conclusion
Waste segregation is a significant environmental issue that the EcoSweep
Automated Waste Segregation and Garbage Collection Robot effectively illustrates how
cutting-edge technologies like computer vision, deep learning (YOLO-V8), and robotic
automation can be integrated to solve. Using a camera, the system can identify different
kinds of waste, classify them precisely using image processing, and physically sort them
with a robotic arm. This automated method improves safety and hygiene in waste-handling
environments by increasing waste management speed and accuracy while lowering the
need for human intervention.

The groundwork for intelligent, expandable, and sustainable waste management

solutions is laid by this project. EcoSweep promotes efficient recycling and environmental
preservation by focussing on waste materials that are frequently encountered, such as metal,
glass, and plastic. Future extensions, like adding more waste types or integrating with IoT
for data analytics and remote monitoring, are made possible by the system's modular
design. Overall, EcoSweep is a useful illustration of how technology can be used to promote
environmental sustainability, which makes it ideal for use in public waste collection areas,
industrial parks, and smart cities. In addition to meeting its technical objectives, the project
supports the larger vision of a cleaner, greener future.

6.2 Applications
• By automating the waste collection and segregation process, urban and rural areas
are kept clean and hygienic.
• Hospitals and medical facilities can use the robot to handle hazardous and
biomedical waste more safely, lowering human exposure and improving hygienic
and health compliance.
• To guarantee effective waste management, they could be placed in public spaces
like parks, streets, and campuses.

Dept. of ECE, VVCE, Mysuru 28

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

6.3 Future Work

Solar Panel Integration
Our purpose is to put in solar panels and a backup battery gadget on EcoSweep. The device
can be capable of characteristic independently of grid electricity thanks to this
improvement. In remoted and rural areas, it guarantees continuous waste segregation.
Sustainable operations and green strength efforts will benefit from the integration. With
this improvement, EcoSweep is placed as a smart, environmentally responsible trash
control answer.

Dept. of ECE, VVCE, Mysuru 29

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

REFERENCES
[1] Jessie R. Balbin, Irish Joy N. Maliban, Joshua Mark A. Marquez, “Automated Waste Segregation Bin
with IoTbased Mobile Monitoring Application”, 2021 IEEE International Conference on Automatic
Control and Intelligent Systems (26 June 2021, Shah Alam, Malaysia.

[2] Bh. Srinivas Sasikanth, Lingamsetty Naga Yoshita, G. Narasimha Reddy, Manitha P.V, “An Efficient &
Smart Waste Management System”, 2021 International Conference on Computational Intelligence and
Computing Applications (ICCICA).

[3] Margaret Richardson Ansah, Solomon Nchor Akansake, Adam Misbawu, “Solar Powered Automatic
Waste Management System Using LoRaWAN”, 2022 IEEE/IET International Utility Conference and
Exposition (IUCE).

[4] Deeksha More E.K, Divya S, G Kalyani, Gowthami R, “Automatic Waste Segregator Bin Using
Robotic Arm”, 2018 3rd IEEE International Conference on Recent Trends in Electronics, Information &
Communication Technology (RTEICT-2018), MAY 18th & 19th 2018.

[5] Vikas J. Nandeshwar, Vishwatej M. Shende, Arya B. Shinde, Rishi R. Shendre, Lokesh P. Shengolkar,
Darshan N. Shinde, “The Roadside Garbage Collection Robot”, 2024 4th International Conference on
Pervasive Computing and Social Networking (ICPCSN).

[6] Ajay V P, Pradeep Kumar M, Kishanth A, Vaishnavi Kumar, R. Santhiya Devi, Amirtharajan Rengarajan,
K. Thenmozhi, and Padmapriya Praveenkumar, “Automatic Waste Segregation and Management”, 2020
International Conference on Computer Communication and Informatics (ICCCI -2020).

[7] Ionut-Robert Badoi, Ioan Lie, “Automatic Waste Segregation System”, 2022 International Symposium
on Electronics and Telecommunications (ISETC).

[8] Chander Partap Singh, Manisha, Pao-Ann Hsiung, Shivani Malhotra, “Automatic Waste Segregator as
an integral part of Smart Bin for waste management system in a Smart City”, 2019 5th International
Conference on Computing Communication Control and Automation.

[9] Sharanya.A, U. Harika, N. Sriya, Sreeja Kochuvila, “AUTOMATIC WASTE SEGREGATOR”,

[10] Nimisha S Gupta, Deepthi V, Mayakunnath, Rejeth Pal S, Badsha T S, Nikhil Binoy C, “Automatic
Waste Segregation”, Proceedings of the Second International Conference on Intelligent Computing and
Control Systems (ICICCS 2018).

[11] Surendra Kumar Koganti, G Purnima, Pechetti Bhavana, Y Veera Raghava, Resmi, “Deep Learning
based Automated Waste Segregation System based on degradability”, Proceedings of the Second
International Conference on Electronics and Sustainable Communication Systems (ICESC-2021).

Dept. of ECE, VVCE, Mysuru 30

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

[12] Nandhini.S, Mrinal Sharma. S, Naveen Balachandran, KK. Suryanarayana D. S. Harish Ram
“Electronically assisted automatic waste segregation”, Proceedings of the Third International Conference
on Trends in Electronics and Informatics (ICOEI 2019).

[13] Megalan Leo.L, Yogalakshmi.S, Jerrin Simla.A, R. Thandaiah Prabu, Sathish Kumar P, Sajiv. G, “An
IoT Based Automatic Waste Segregation and Monitoring System”, Proceedings of the Second
International Conference on Artificial Intelligence and Smart Energy (ICAIS-2022).

[14] Md. Shahariar Nafiz, Shuvra Smaran Das, Md. Kishor Morol, Abdullah Al Juabir, Dip Nandi, “Convo
waste: An Automatic Waste Segregation Machine Using Deep Learning”, 2023 3rd International
Conference on Robotics, Electrical and Signal Processing Techniques (ICREST).

[15] Nurisha Hania Kamarudin, A’zraa Afhzan Ab Rahim Faculty of Electrical Engineering, Noor Ezan
Abdullah, Ili Shairah Abdul Halim, Siti Lailatul Mohd Hassan, “Development of Automatic Waste
Segregator with Monitoring System”, 2019 4th International Conference on Information Technology
Information System and Electrical Engineering (ICITTSEE).

[16] Md. Nafis Raihan, Mahbubur Rahman, Fahad Bin Alam, Ekra Bin Syed Mojib, “A Novel Approach
for Waste Collection: Automated Waste Collecting Robot with Advanced Image Recognition
Technology and Onboard Robotic Arm”, 2020 IEEE Region 10 Symposium (TENSYMP) 5-7 June 2020,
Dhaka, Bangladesh.

Dept. of ECE, VVCE, Mysuru 31

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

Appendix
Python Source Code
import cv2
from ultralytics import YOLO
import serial
import time
import pyttsx3

engine = [Link]()

# Load the trained YOLO model

model_path = "[Link]" # Path to your trained model
model = YOLO(model_path)

# Open the live camera feed

camera_index = 0 # Use 0 for the default camera; adjust if you have multiple cameras
cap = [Link](camera_index)

if not [Link]():
print("Error: Could not access the camera.")
exit()

# Set up the serial port (adjust to your serial port)

serial_port = 'COM4' # Replace with your actual serial port (e.g., '/dev/ttyUSB0' on Linux)
baud_rate = 9600
ser = [Link](serial_port, baud_rate)

# Frame tracking
detection_counts = {'glass': 0, 'metal': 0, 'plastic': 0}
frame_threshold = 15 # Number of frames to track the same object

# Initialize a frame counter for prediction skipping

frame_counter = 0
prediction_frequency = 6 # Process every 3rd frame

Dept. of ECE, VVCE, Mysuru 32

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

while True:
ret, frame = [Link]()
if not ret:
print("Failed to grab frame.")
break

# Skip frames to improve performance

if frame_counter % prediction_frequency != 0:
frame_counter += 1
continue

# Perform prediction on the frame

results = [Link](source=frame, save=False, show=False, conf=0.8) # Adjust conf
as needed

# Get the detected class names and corresponding confidence scores

detected_classes = results[0].names # class names
boxes = results[0].[Link] # Bounding box coordinates
confidences = results[0].[Link] # Confidence scores

# Reset detection counts if no objects are detected in this frame

detection_found = False

# Process detections for each object

for i, box in enumerate(boxes):
confidence = confidences[i].item() # Get the confidence score
if confidence >= 0.8: # Adjust threshold as needed
class_id = int(results[0].[Link][i].item()) # Class ID of detected object
class_name = detected_classes[class_id] # Class name corresponding to the ID
print(class_name)

if 'glass' in class_name:
print("glass")
detection_counts['glass'] += 1
detection_found = True

Dept. of ECE, VVCE, Mysuru 33

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

elif 'metal' in class_name:

print("metal")
detection_counts['metal'] += 1
detection_found = True
elif 'plastic' in class_name:
print("plastic")
detection_counts['plastic'] += 1
detection_found = True

# If any object has been detected continuously for `frame_threshold` frames

try:

for class_name, count in detection_counts.items():

if count >= frame_threshold:
if class_name == 'glass':
print("sent 1")
[Link]("glass detection sent")
[Link]()
[Link](b'1') # Send 1 via serial for glass
elif class_name == 'metal':
print("sent 2")
[Link]("metal detection sent")
[Link]()
[Link](b'2') # Send 2 via serial for metal
elif class_name == 'plastic':
print("sent 3")
[Link]("plastic detection sent")
[Link]()
[Link](b'3') # Send 3 via serial for plastic
# Reset the count after sending
detection_counts[class_name] = 0
except:
print("port error")

# Visualize the predictions (you can add labels and boxes)

Dept. of ECE, VVCE, Mysuru 34

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

annotated_frame = results[0].plot() # Draw boxes and labels on the frame

# Display the annotated frame

[Link]("YOLO Live Detection", annotated_frame)

# Increment the frame counter

frame_counter += 1

# Break loop on 'q' key press

if [Link](1) & 0xFF == ord('q'):
break

# Release resources
[Link]()
[Link]()
[Link]() # Close the serial connection

Dept. of ECE, VVCE, Mysuru 35

Eco Sweep: Automated Waste Segregation and Garbage Collection Robot
2024-25

List of publications

Our project’s research paper titled “Eco Sweep: Automated Waste Segregation and
Garbage Collection Robot”, was presented at the 3rd IEEE International Conference on
Knowledge Engineering and Communication Systems (ICKECS-2025), SJC Institute of
Technology, Chikkaballapur, India, April 2025.

Dept. of ECE, VVCE, Mysuru 36

 The Report is Generated by DrillBit Plagiarism Detection Software

Submission Information

Author Name Team 24

Title EcoSweep
Paper/Submission ID 3568923
Submitted by hodec@[Link]
Submission Date 2025-05-02 [Link]
Total Pages, Total Words 32, 7846
Document type Project Work

Result Information

Similarity 11 %
1 10 20 30 40 50 60 70 80 90

Sources Type Report Content

Student
Paper Words < Quotes
0.43% Internet 14, 1.77%
3.4% 3.35%

Journal/
Publicatio
Ref/Bib
n 7.16%
7.67%

Exclude Information Database Selection

Quotes Not Excluded Language English

References/Bibliography Not Excluded Student Papers Yes
Source: Excluded < 14 Words Not Excluded Journals & publishers Yes
Excluded Source 0% Internet or Web Yes
Excluded Phrases Not Excluded Institution Repository Yes

A Unique QR Code use to View/Download/Share Pdf File