You are on page 1of 40

Application of Artificial Intelligence and Machine

Learning approach in Waste Management System


A Presentation
Submitted in partial fulfillment of the requirement for the
award of Degree in Mechanical Engineering (ME-6299)

Archita Gogoi (D/17/ME/05)


Tapi Napa (D/17/ME/51)
Wasifur Rahman (D/16/ME/25)
Supervisor: Prof. M. Chandrasekaran
Department of Mechanical Engineering
North Eastern Regional Institute of Science and Technology
Overview
• Introduction

• Literature Review

• Design of a Smart Waste Collection System

• Development of Predictive Model for Probability of Waste collection by


SWBs

• Development of Optimization Route for waste collection from different


SWBs

• Conclusions

• References
Introduction to Waste Management

• Waste is unwanted, unusable or


discarded materials, having no or
little significance after use which
possess no or minor economic value.
With the generation of a million-ton
of waste everyday, the threat to
public health and environment
prevails undoubtedly.

• The term ‘waste management’


denotes the measures or steps
necessary for collection, treatment
and discarding of waste while
keeping in mind its impact on human
health and environment.
Fig.1: Waste Management System
Smart Approaches for Waste Management

• As the Government of India has taken an important


initiative to set up 100 smart cities, our main aim is
to accordingly come up with a smart approach to
tackle the issue of waste management.

• The approaches employed in a Smart Waste


Management System are:
1. Internet of Things (IoT),
2. Artificial Intelligence (AI), and
3. Machine Learning (MI).
Internet of Things (IoT)

• The Internet of Things (IoT) can be


defined as a system of interrelated
and inter-connected objects (called
‘things’) that has the ability to
collect and transfer data over a
wireless network without human
intervention.

• The term ‘things’ can refer to a wide


variety of devices ranging from
ordinary household objects to
advanced industrial equipments.

Fig.2: Internet of Things (IoT)


Artificial Intelligence (AI)

• Artificial Intelligence (AI) is


defined as a vast branch of
computer science that
endeavours to build smart
entities capable of performing
intelligent tasks without
human intervention.

• It is the simulation of human


intelligence in machines
granting them the capabilities
to mimic and behave like
humans. Fig.3: Artificial Intelligence (AI)
Machine Learning (ML)

• Machine Learning (ML) is a


subset of AI; in fact, it is simply
a technique for realising AI.

• ML is defined as a platform
which facilitates the learning of
the trends of a particular
system from the data
generated in the past, to make
an informed decision in the
future without manual
monitoring.
Fig.4: Machine Learning (ML)
Scope of the Project

• Two significant challenges for waste management are waste


collection and path collection.

• The concepts of IoT and AI are integrated in order to create


optimal changes in the conventional methodology of waste
collection using the large amount of data that is being
generated by the smart waste bin networks.

• In this work, the experimental datasets of work carried out


by Tran et al., [21] will be employed.
Literature Review on Intelligent Approaches
used in Waste Management
Author/Year of Waste Management Algorithm used Features
Publication System used

Delgado et al., Optimal waste Graph theory They used GSM/GPS system
2019 [17] collection and transmission module; No
hardware design and mobile
application program (APP)
were used.

Hannan et al., Optimal waste Heuristic They used transmission


2018 [6] collection module; No hardware design
and mobile application
program (APP) were used.

Jim et al., 2019 Management No They used mobile application


[7] program (APP); No hardware
design and transmission
module were used.
Author/Year Waste Algorithm used Features
of Publication Management
System used
Hong et al., Low cost, energy No They made use of LoRa Technology,
2014 [20] saving hardware design, transmission
module and mobile application
program (APP).

Tran et al., Optimal waste Graph theory They made use of LoRa technology,
2020 [21] collection, low cost, and Machine hardware design, transmission
energy saving Learning module and mobile application
program (APP).

Our Proposal Optimal waste Graph theory We will make use of GSM module
collection, low cost, and Machine for transmission, an effective
energy saving Learning hardware design and mobile
application program (APP).
Proposed Objectives

• In this work, it is planned to fulfil the following three objectives:

1. Design a microcontroller system with a sensor module to


measure the filling height of garbage in a waste bin and send
the information using GSM module to a mobile app.

2. Develop an IoT-based machine learning method for


predicting the probability of collecting waste from different
waste bins based on the historical input data.

3. Optimize the paths of waste collection from the smart waste


bin placed in different areas.
Design of a Smart Waste Collection System

Fig.5: Block Diagram of proposed Smart Waste Bin (SWB) system


Different Components of the SWB System

1. Smart Waste Bin (SWB): It is a waste bin in which an ultrasonic sensor is


interfaced.
2. Ultrasonic Sensor (US): It is used to detect the level of the garbage filling
in the waste bin. The model of the sensor used is HC-SR04.
3. Node for Microcontroller Unit (NodeMCU): It provides connecting node
with ultrasonic sensor with microcontroller board wirelessly (i.e., by
using Wi-Fi).
4. Microcontroller Board: The microcontroller board processes the data
(waste loading) obtained from the ultrasonic sensor.
5. Global System for Mobile Communication (GSM) Module: It is used to
establish communication between the microcontroller board and the
network so as to transfer the processed data to the internet through
radio waves.
6. Mobile Application Program (APP): The output of GSM module is
displayed in mobile; and identifies the waste loading details of different
dust bins. Blynk mobile App is used in this work.
Different Steps of Waste Collection System

Dust collection begins with SWBs placed at different locations

Ultrasonic sensor in SWB monitors garbage filling level (i.e. ‘d’)

The real time data ‘d’ of various SWB is sent to microcontroller board.
(If d≤ 5cm; SWB is full
If d˃ 5cm; SWB is not full)
Data is sent to GSM module to establish communication with Blynk Mobile APP.

Blynk Mobile APP generates message about the current status of SWB (Full(1) or
Not Full(0))

If SWB is Full(1), then the garbage Datasets are stored for the
removal action will be initiated development of LR Predictive model

Dijkstra’s Optimal Algorithm is


applied

Optimised route for garbage


collection from different locations
Fig.6: Proposed smart waste bin embedded with an ultrasonic sensor to detect the level of
garbage (Sohag and Podder, [24])
Hardware Implementation

Fig.7: Circuit diagram showing the connections of Fig.8: Circuit showing the connections of the
the ultrasonic sensor and the GSM module with ultrasonic sensor and the GSM module with
the Arduino Uno board the Arduino Uno board
Fig.9: Circuit diagram showing the connections of the ultrasonic
sensor with the Node MCU

Fig.10: Circuit showing the connection of ultrasonic sensor with


NodeMCU (Anwar and Aasim, [23])
Expected Output

• The expected output of the


system should be a text
message showing the value of
the distance between the level
of garbage in the waste bin and
it’s top.

• The output will be shown on


the serial monitor of the
Arduino IDE as well as on the
serial monitor of the Blynk App.

Fig.11: Screenshot of output in Blynk


(Anwar and Aasim, [23])
Development of Predictive Model for Probability
of Waste Collection by SWBs

• The Logistic Regression(LR)


model is chosen as the
predictive modelling algorithm
to predict the probability of
waste collection in each waste
bin.

• The purpose behind choosing


the LR Model is that the output
variable is a binary classification;
that is, the status of the waste
bin can either be “collected(1)”
or “not collected(0)” Fig.12: Graph of Logistic Regression
Datasets Used

• The experimental work carried out by Tran et al. [21] is used in this
work.

• Two types of datasets were collected.

• The first type consists of the number of classes held in each


building of the university campus for a time period of 4 months.
There are 13 buildings in total which gives 13 datasets to train the
model.
Dustbin No. of No. of No. of No. of No. of No. of No. of
location in Monday Tuesday Wednesday Thursday Friday Saturday Sunday
Building classes classes classes classes classes classes classes

A 91 95 97 99 90 95 54
B 234 281 293 280 269 165 34
C 156 149 174 167 146 61 8
D 81 120 110 91 95 68 8
E 252 231 287 259 273 226 6
F 97 117 134 117 118 74 1
G 21 21 21 21 21 21 0
H 7 8 7 7 7 8 4
I 62 56 48 42 42 44 0
J 2 0 0 2 2 0 0
K 23 38 30 27 25 15 0
L 4 3 0 0 0 0 0
M 14 24 19 29 23 14 18

Table 1: Number of classes held for 4 months period


• The next type consists of datasets which states whether
the waste from the smart waste bins placed in each
building was “collected” or “not collected”.

• The number (1) was assigned to denote “waste collected”


and to denote “not collected” the number (0) was
assigned.
Dustbin
location in Monday Tuesday Wednesday Thursday Friday Saturday Sunday
Building
A 1 1 1 1 1 1 0
B 1 1 1 1 1 1 0
C 1 1 1 1 1 1 0
D 1 1 1 1 1 1 0
E 1 1 1 1 1 1 0
F 0 0 1 1 1 1 0
G 0 0 0 0 0 0 0
H 0 0 0 0 0 0 0
I 1 0 0 0 0 0 0
J 0 0 0 0 0 0 0
K 0 0 0 0 0 0 0
L 0 0 0 0 0 0 0
M 0 0 0 0 0 0 0

Table 2: Actual output of waste collected (i.e., Filled SWB) or not collected (i.e., Not filled SWB)
1= Filled SWB; 0= Not filled SWB
Model Training and Development
• The first dataset obtained from Tran et al. [21] were used to train the
model.

• The source code for training the model was written in Jupyter Notebook
in the platform Colaboratory or ‘Colab’ which is a product from Google
Research.

• The Logistic Regression equation to predict the probability (P) of the


case “collected(1)” of the waste bin was obtained from the Scikit Learn
library which is:

• By executing the source code, the above equation is applied to each of


the 13 datasets.

• The output gives the value of probability of waste collection(1) from the
smart waste bins .
Dustbin Monday Tuesday Wednesday Thursday Friday Saturday Sunday
location in
Building

A 0.842 0.878 0.894 0.908 0.832 0.88 0.237


B 1 1 1 1 1 0.999 0.062
C 0.999 0.998 1 0.999 0.997 0.35 0.009
D 0.713 0.98 0.958 0.843 0.879 0.48 0.009
E 1 1 1 1 1 1 0.008
F 0.893 0.975 0.993 0.975 0.977 0.59 0.005
G 0.024 0.024 0.024 0.024 0.024 0.024 0.005
H 0.008 0.009 0.008 0.008 0.008 0.009 0.007
I 0.364 0.264 0.163 0.11 0.11 0.127 0.005
J 0.006 0.005 0.005 0.006 0.006 0.005 0.005
K 0.028 0.082 0.047 0.037 0.032 0.015 0.005
L 0.007 0.006 0.005 0.005 0.005 0.005 0.005
M 0.014 0.03 0.021 0.043 0.028 0.014 0.019

Table 3: Probability of collecting waste (1) from the smart waste bins
Predicted Output of Waste Collected from
SWB

• Based on the previous obtained output consisting of the values of


probability, the code is executed to predict the status of the waste
bin.

• If the value of probability obtained is greater than the threshold


value of 0.5, then the status of the waste bin should be updated
to “collected(1)”. Else it should show “not collected(0)”.
Dustbin
location in Monday Tuesday Wednesday Thursday Friday Saturday Sunday
Building

A 1 1 1 1 1 1 0
B 1 1 1 1 1 1 0
C 1 1 1 1 1 0 0
D 1 1 1 1 1 0 0
E 1 1 1 1 1 1 0
F 1 1 1 1 1 1 0
G 0 0 0 0 0 0 0
H 0 0 0 0 0 0 0
I 0 0 0 0 0 0 0
J 0 0 0 0 0 0 0
K 0 0 0 0 0 0 0
L 0 0 0 0 0 0 0
M 0 0 0 0 0 0 0

Table 4: Predicted output of waste collected (i.e., Filled SWB) or not collected (i.e., Not filled SWB)
1= Filled SWB; 0= Not filled SWB
Efficiency of the LR Model
• In order to determine the efficiency
of the Logistic Regression model,
Receiver Operating Characteristic
(ROC) curve is used.

• The code was first executed to


import the classification report of
the Logistic Regression model from
the Scikit Learn library. This gave the
optimum classification cut off value
to be 0.5 at TPR=0.91428571 and
FPR=0.035714291.

• And lastly by importing Matplotlib,


the ROC curve was plotted (Fig. 13)
which gave the Area under the curve Fig.13: ROC Curve
(AUC) to be 0.94 which implies the
efficiency of the model to be 94%.
Development of Optimization Route for
Waste Collection for Different SWB
• The predicted output of the status of waste collection (collected (1)
or not collected (0)) for each of the smart bins placed in all the 13
buildings of the university campus is used to build a graph theory
based algorithm for route planning of the waste pickup trucks.

• The chosen algorithm is Dijkstra’s algorithm as the objective is to


find the shortest distance for collecting waste from all full waste
bins.

• For testing the algorithm, the dataset giving the predicted output of
waste collected(1) or not collected(0) of the SWBs is used. The
source code for executing the algorithm was written in Jupyter
Notebook in the platform Colaboratory or ‘Colab’ which is a product
from Google Research.
Generation of Random Graph
• At first, a random position is
generated inside a circle of radius ‘r’
where ‘r’ is the coverage of the Wi-Fi
module which is 50m in this case and
the coordinate of the centre is (0,0).

• Similarly, another random point is


generated within the range of the
radius of the first circle (Wi-Fi range).
The coordinate of the generated point
is now the centre of the new circle.

• This process is repeated till all the 13


nodes are generated.

• By importing the Python Math and


Matplotlib libraries and after
executing the required code, the Fig.14: Random Generated Graph
random graph containing all the 13
nodes is generated.
Determination of the Coordinates of all
Nodes of the Random Generated Graph
Buildings (Nodes) X- coordinate Y-coordinate
A -22.81 -42.90
B -16.54 -64.99
C -2.71 -111.84

• After executing the D 14.50 -71.85


E 52.61 -88.24
required code, the
F 83.45 -89.17
Cartesian coordinates of
G 112.04 -69.67
all the 13 nodes are
H 138.77 -87.58
obtained.
I 104.70 -109.18
J 65.50 -115.87
K 19.71 -108.91
L 47.22 -83.45
M 73.86 -61.22

Table 5: Cartesian coordinates of the nodes of the


Random Generated Graph
Calculating distance between all pairs of nodes of
the random generated graph
Distance between buildings
Distance (in m)
(Pair of Nodes)
AB 22.95
BC 48.85
CD 43.54
DE 41.48
EF 30.84
FG 34.60
• The required code is executed for GH 32.17
calculating the distance between all pairs HI 40.34
IJ 39.76
of nodes. And the output is provided in JK 46.31
the table. KL 37.48
LM 34.70
MA 98.40
AC 71.81
CE 60.15
EB 72.96
BD 31.80
DG 97.56
GJ 32.17
JM 55.28
MI 57.01
IF 29.18
FL 36.67
LH 91.64
Application of Dijkstra’s Algorithm

• Node A is considered to be the Path Shortest distance (in m) Route


source node. AB 22.95 AB
AC 71.81 AC

• The distances between the AD 54.76 ABD

various nodes are considered as AE 95.92 ABE


weights based on which the AF 126.77 ABEF
Dijkstra’s Algorithm will calculate AG 152.32 ABDG
the shortest possible path AH 184.49 ABDGH
between the source and the
AI 155.96 ABEFI
destination nodes.
AJ 184.49 ABDGJ
AK 230.81 ABDGJK
• By importing the SYS Library in AL 163.45 ABEFL
Python and after executing the AM 98.40 AM
algorithm, the required shortest
path for waste collection will be
obtained. Table 7: Shortest path for waste collection
given by Dijkstra’s Algorithm
Conclusions
• In this work a microcontroller system based on IoT architecture has been proposed for waste
collection system.

• The system proposes a ‘smart waste bin (SWB)’ provided with ‘ultrasonic sensor’ to monitor
waste filling level in the waste bin (or) dust bin. The ultrasonic sensor measures the garbage
filling level (d) of different SWB located at various locations.

• The collected data is served as input information to develop Logistic Regression model for
predicting the probability of the waste collection (filling rate) from different smart waste bins.
The datasets of Tran et al., [21] obtained from 13 SWBs located in various buildings is
employed in this work.

• After that, a practical use-case of the waste collection process at the Ton Duc Thang University
was tested by computing the area under the Receiver Operating Characteristic (ROC) curve.

• Finally, Dijkstra’s algorithm was applied to optimize the shortest route for waste collection
between two filled waste bins.

• This entire system when integrated together provides better operations for optimizing
operating costs and collecting the wastes on time and hence eliminating the problem of over
flowing waste bins.
References
• [1] Dubey, S., Singh, P., Yadav, P., and Singh, K.K., (2019) Household Waste
Management System Using IoT and Machine Learning, International Conference
on Computational Intelligence and Data Science 167(2020) pp. 1950-1959.

• [2] Badar, M., Shamsi, S., Aldalbahi, A., and Haque, M.M., (2020) Applications of
AI and ML in IoT, Integration of WSNs into Internet of Things: A security
perspective.

• [3] Ahmed, M., and Memon, M., (2010) Integrated solid waste management
based on the 3R approach, Journal of Material Cycles and Waste Management
12(1) pp. 30-40.

• [4] Agarwal, R., Chaudhary, M., and Singh, J., (2015) Waste Management
Initiatives in India for human well being, European Scientific Journal 11(10) pp.
106-127.

• [5] Ahn and Choi, J., (2019) Internet of vehicles and cost-effective traffic signal
control, Sensors 19 pp.1275.
• [6] Hannan, M., Akhtar, M., Begum, R.A., Basri, H., Hussain, A., and Scavino, E., (2018)
Capacitated vehicle-routing problem model for scheduled solid waste collection and
route optimization using PSO algorithm, Waste Management 7(1) pp.31-4.

• [7] Jim, A.A.J., Kadir, R., Mamun, M.A.A., Nahid, A.A., and Ali, M.Y., (2019) A noble
proposal for Internet of Garbage Bins (IoGB), Smart Cities 2(2) pp. 214-229.

• [8] Zantalis, F., Koulouras, F., Karabetsos, S., and Kandris, D., (2019) A review of
machine learning and IoT in smart transportation, Future Internet 11(4) pp. 94.

• [9] Sahu, R.M., Godase, A., Shinde, P., and Shinde, R., (2016) Garbage and Street
Monitoring System using Internet of Things, International Journal of Innovative
Research in Electrical, Electronics, Instrumentation and Control Engineering 4(4).

• [10] Shubho, M.T.H., Hassan, M.T., Hussain, M.R., and Neema, M.N., (2013)
Quantitative Analysis of Spatial Pattern of Dustbins and its Pollution in Dhaka City—A
GIS Based Approach, Asian Transactions on Engineering (ATE ISSN: 2221-4267) 3(4)
pp.1-7.

• [11] Sharma, N., Singha, N., and Dutta, T., (2015) Smart Bin Implementation for Smart
Cities, International Journal of Scientific and Engineering Research 6(9) pp. 787-791.
• [12] Monika, K.A., (2016) Smart dustbin—an efficient garbage monitoring system, International
Journal of Engineering Science and Computing 6(6) pp.7113-7116.

• [13] Anitha, A., (2017) Garbage monitoring system using IoT, 14th IOP Conference Series: Materials
Science and Engineering 26(3) pp.1-12.

• [14] Shyam, Krishna, Gopal, Kumar, Sunil, Manvi, S., and Bharti, P., (2017) Smart waste management
using Internet of Things (IoT), 2nd International Conference on Computing and Communications
Technologies IEEE pp. 199-203.

• [15] Yusof, N.M., Jidin and Rahim, (2017) Smart garbage monitoring system for waste management
in MATEC Web of Conferences (EDP Sciences) pp. 97-01098.

• [16] Kumar and Sathish, N., (2016) IoT based smart garbage alert system using Arduino UNO, IEEE
Region 10 Conference (TENCON) pp. 1028-1034.

• [17] Bueno-Delgado, M.V., Romero-Gázquez, J.J.L., Jimenez, P., and Pavón-Marino, P., (2019) Optimal
path planning for selective waste collection in smart cities, Sensors 19 pp. 1973.

• [18] Pardini, K., Rodrigues, J., Kozlov, S.A., Kumar, N., and Furtado, V., (2019) IoT based solid waste
management solutions: A survey, Journal of Sensor and Actuator Networks 8(1) pp.5.

• [19] Najm, I.A., Hamoud, A.K., Lloret, J., and Bosch, I., (2019) Machine learning prediction approach
to enhance congestion control in 5G IoT environment, Electronics 8(6) pp.607.
• [20] Hong, I., Park, S., Lee, B., Lee, J., Jeong, D., and Park, S., (2014) IoT based
smart garbage system for efficient food waste management, The Scientific World
Journal, Article ID 646953 pp. 13.

• [21] Tran, A.K., Cao, H.P., Pham, D.L., and Le Mai, B.N., (2020) Waste
Management System using IoT based Machine Learning in University, Wireless
Communication and Mobile Computing pp. 1-13.

• [22] Sunny, S.H., Dipta, D.R., Hussain, S., Faruque, H.M.R., and Hussain, E., (2019)
Design of a Convolutional Neural Network based Smart Waste Disposal System,
1st International Conference of Advanced Science, Engineering, Robotics and
Technology pp. 1-5.

• [23] Anwar and Aasim, (2018) IoT based garbage monitoring using Arduino, Diss.
West Bengal University of Technology.

• [24] Sohag, M.U., and Podder, A.K., (2020) Smart garbage management system
for a sustainable urban life: An IoT based application, Internet of Things Vol. 11
pp. 100255.
Supervisor: Prof. M. Chandrasekaran (Ex-H.O.D)

Project Coordinator: Dr. K. K. Mandal


H.O.D: Prof. S. Mahto


Faculty Members

Department of Mechanical Engineering, NERIST


You might also like