See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332423832

The Use of Modern Technology in Smart Waste Management and Recycling:

Artificial Intelligence and Machine Learning

Chapter  in  Studies in Computational Intelligence · April 2019

DOI: 10.1007/978-3-030-12500-4_11

CITATIONS READS

0 473

8 authors, including:

Praveen Kumar Gupta Vidhya Shree

Rashtreeya Vidyalaya College of Engineering Rashtreeya Vidyalaya College of Engineering
47 PUBLICATIONS   68 CITATIONS    3 PUBLICATIONS   1 CITATION   

SEE PROFILE SEE PROFILE

Lingayya Hiremath Sindhu Rajendran

Rashtreeya Vidyalaya College of Engineering Rashtreeya Vidyalaya College of Engineering
28 PUBLICATIONS   9 CITATIONS    3 PUBLICATIONS   0 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Nanotechnology based sustained action drug delivery system View project

Environmental Impact and drug development View project

All content following this page was uploaded by Praveen Kumar Gupta on 15 April 2019.

The user has requested enhancement of the downloaded file.

The Use of Modern Technology in Smart
Waste Management and Recycling:
Artificial Intelligence and Machine
Learning

Praveen Kumar Gupta, Vidhya Shree, Lingayya Hiremath

and Sindhu Rajendran

Abstract Waste management is one of the primary problem that the world faces
irrespective of the case of developed or developing country. The key issue in the waste
management is that the garbage bin at public places gets overflowed well in advance
before the commencement of the next cleaning process. It in turn leads to various
hazards such as bad odor and ugliness to that place which may be the root cause
for spread of various diseases. The increase in population, has led to tremendous
degradation in the state of affairs of hygiene with respect to waste management
system. The spillover of waste in civic areas generates the polluted condition in
the neighboring areas. For eliminating or mitigating the garbage’s and maintains
the cleanness, it requires smartness based waste management system. The need of
proper waste management does not end with just collection and proper dispose of
garbage. It continues to the level of landfills and the amount that we can possibly
recycle. Recycling is estimated to be highly useful given that our dependency on raw
products reduces, besides the reduction of waste and subsequent landfills. Once the
recycling is done to sort metals, plastics, and glass articles, the use of biodegradable
waste can be extended beyond fertilizers and manure. The metals can be reused and
the plastics can be diverted from the landfills, which otherwise leads to choking of
the earth. The glass materials can be broken and melted back to form new articles
after deep cleaning. This chapter aims to understand the use of machine learning and
artificial intelligence in the most potential areas and the ultimate need to completely
replace the human interaction.

P. K. Gupta (B) · L. Hiremath

Department of Biotechnology, R. V. College of Engineering, Bangalore 560059, India
e-mail: praveenkgupta@rvce.edu.in
V. Shree
Department of Electronics and Instrumentation, R. V. College of Engineering,
Bangalore 560059, India
S. Rajendran
Department of Electronics and Communication, R. V. College of Engineering,
Bangalore 560059, India

© Springer Nature Switzerland AG 2019 173

R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence,
Studies in Computational Intelligence 823,
https://doi.org/10.1007/978-3-030-12500-4_11

er.ramankumar@aol.in
174 P. K. Gupta et al.

Keywords Location intelligence · Smart city · Graph theory optimization ·

Internet of things · Waste management · Automation · Artificial intelligence in
recycling · Graph theory · Graph optimization · Machine learning · Artificial
neural networks (ANN)

1 Introduction

In one of the recent surveys conducted by World Bank, about 1.3 billion of tons of
waste is generated each year. It is estimated that the numbers would rise to nearly
2.2 billion by 2025. These numbers have a significant impact on our societies and
on the ecology on the whole. The more the waste we generate, the more we waste
the resources and initiate hazardous situations for bio life. Waste management is
one of the primary problem that the world faces irrespective of the case of devel-
oped or developing country. Improper waste management has been proven to be
the sole reason for the hazardous air pollution leading to serious health issues like
COPD (Chronic Obstructive Pulmonary Disease) and Asthma, for people who have
their livelihood close to these Garbage bins. The failure to remove this accumulated
garbage is the sole reason for breeding of mosquitoes and houseflies, which is the
root cause of fatal diseases like Dengue, Malaria, etc. Nearly 235 million people are
victims of breathing illness due inhalation of air with foul smell [1]. While reducing
the level of human interaction in the waste management is equally essential, we need
to look for industrial automation using smart waste management techniques. This is
needed right from proper waste disposal, waste collection, prompt checking to pre-
vent over flow to safely disposing them. Recycling is yet another major innovative
step towards reduction of waste and to get back from the used products. Recycling
industry is now dominated by dumpster robots which makes the task of sorting and
recycling very easy and brings down the human intervention and exposure to waste.
Through this chapter we hope to establish the information based on the latest tech-
nology which has been proposed and practically implemented with optimization as
the future developments.

2 Smart Waste Management Using Artificial Intelligence

In order to increase the efficiency of cleaning processes, they are several methods
based on artificial intelligence and IoT, which effectively designed and implemented.
The aim of an integration of ‘Smart Cities’ is the ultimate goal of each developed and
developing country. With the integration of IoT and Data access networks, combina-
torial optimization, Geographic Information Systems (GIS), implemented through
electronic engineering, the waste management of any city can be organized easily
[2]. The IoT integrated trash cans can sense and send the data about the trash vol-
ume over to the servers through the internet. Using the graph theory optimization

er.ramankumar@aol.in
The Use of Modern Technology in Smart Waste Management … 175

techniques and associated algorithms, the data is processed to develop dynamic and
efficient waste management. The proposed system is simulated in a realistic city of
Copenhagen, using Open Data. This is pitched into the reality to attract third party
investments into the benefits attained by this model and result in innovative solu-
tions for smart cities. The established system depicts the involvement of Geographic
Information Systems (GIS), applied graph theory optimization, and some concepts
of machine learning.
The use of sensors is essential for sending the trash volume data to the servers,
which is processed and used to track and initiate the cleaning process accordingly
on a routine. This ensures that the trash is cleared well before it overflows and also,
prevents the unnecessary trash pickup when the trash cans are not filled. Open Data
is known to be the initiator of Big Data analytics which are enhanced with cyber
systems [3].

2.1 System Function

The system is completely based on the indication of waste levels in the trash can
by the embedded sensors. The data about the volume is transmitted to the servers
over the internet. The collected data after processing is used to optimize and direct
the municipality cleaning process. With this automated process, the cleaning team
has new routes each day, to visit and clear trash bins based on the need and also
ensure that there is no trash left unattended. The end results are a cleaner society
with reduced costs and manual efforts. The collected data is used to statistically
analyze the rate of filling up of these trash cans (Fig. 1).

Fig. 1 The proposed system in the smart city: Copenhagen [2]

er.ramankumar@aol.in
176 P. K. Gupta et al.

2.2 System Components

1. Sensors: The amount of waste is determined by the level of waste collected

from the top of the lid. The SONAR technique is used to sense the distance,
using Ultrasonic Ranging Module (HC-SR04). With an accuracy of 3 mm, this
technique can clearly indicate the distance in the range from 2 to 400 cm. The
temperature sensors, and weight sensors along with metal detectors are useful
for sorting the biodegradable, recyclable and reusable items from the trash cans
[2].
Working of Ultrasonic Ranging Module:
The Basic principle involved in the non-contact measurement is:
1. The IO trigger is used for a minimum of 10 μs of the high level signal.
2. The unit is designed to transmit eight 40 kHz of the triggering signal which is
detected once it is received back.
3. The time lapse in the event of emission and retrieval of the signal is used to
calculated and accordingly calibrated to estimate the distance of the level of the
trash filled (Fig. 2).
The angle between the sensors along with the distance from each other is important
to overcome the ‘crosstalk’. The target level is to be mounted at a perpendicular angle
from the base of the sensor [1].
2. Microcontroller: Arduino Uno, based on ATMega328, is used as the decision
making microcontroller in this prototype. With an operating voltage at 5 V and
optimized power consumption of about 40–50 mA, it is sufficient to collect data
and transmit it over the internet. The microcontroller has been chosen based on
the parameters like minimal power consumption, memory, processing and the
economic cost.
3. Access Network Interface: using the WiFi technology, the information from the
trash cans can be sent to the servers through internet. One of the methods have

Fig. 2 The working of the ultrasonic ranging module [1]

er.ramankumar@aol.in
The Use of Modern Technology in Smart Waste Management … 177

implemented the CC3000 Shield with the on-board antenna as it is known to

have exceptional coupling with the microcontroller chosen.
4. Battery output: Since the working of this system is based on a daily report gen-
eration, the battery is chosen to facilitate a long duration of use.
5. Database: The data collected from the trash cans are organized into the data
systems, using the software such as MySQL.
6. Artificial Intelligence: from the data collected, the statistical analysis helps in the
determination of the possible rates of filling the trash cans and the appropriate
routes to be followed regularly by the cleaning team. Also, the timings of the data
collected from trash cans emphasizes on the locations to be given priority during
the different times of the day. Some places might see a regular filling of trash cans
in the mornings while others in the late nights. Also, there are some locations like
the schools, offices, houses which need immediate cleaning compared to open
fields with lesser human interaction. So once the artificial intelligence is boosted
to the system, the routes are defined and optimized based on the timings and the
corresponding need [4].
7. Optimization techniques: with the recorded data, the system uses the artificial
intelligence and optimization techniques [5] to reduce the transportation expen-
diture of costs and time, by directing the best routes to empty the filled trash cans.
This is optimized keeping in mind the traffic, and the driving distance and time.
The end of the optimization results in a highly efficient route to be followed by
the truck drivers, enhanced with the GPS location services [6].

2.3 The Smart Dustbin

The cylindrical structure is associated with a piston that is useful for compression of
the garbage. The trash can plate is attached to the cylinder and the leaf switch is to be
suspended upside down [7] through the side hole. The leaf switch level is placed at a
point lower than the maximum level. This is essential for the precautionary measures
of the garbage overflow, in cases of fault from the cleaning team. The compressing
plate can reach down to press the switch. Once the threshold level is reached, the
garbage is prevented to be dumped inside the trash can in order to avoid overflow
(Fig. 3).

2.4 The Optimization Techniques and the Corresponding

Algorithms Opted

1. Shortest path Spanning Tree (SPST): The SPST technique is used to find the
shortest distance between two trash cans in the pre entered graph data of the
streets and lanes in the outlay of the Smart city. The streets are associated with

er.ramankumar@aol.in
178 P. K. Gupta et al.

Fig. 3 The design of the smart dustbin [1]

the edges and the joining points are considered as the vertices. This is essential
to optimize the routes for trash pickup from the data received from all the trash
cans [8].
2. Genetic Algorithms (GA): The optimization of the trash collecting cycles is a
combinatorial problem, which is to be duly similar to the ‘The minimum Linear
Arrangement Problem’ [9] and the popular ‘The Traveling Salesman Problem’
[10]. The GA techniques are quite efficient when it comes to highly optimized
driving routes for the trash clearance [11].
3. K-Means: another NP-Hard problem, it is especially complex when it comes to
solving hard clustering problems. K-means is highly efficient method to be used
as solution in situations of huge complex cluster parameters.

2.5 The Evidence of Practical Output of the Proposed Model

With our simulated city of Copenhagen, Denmark, the working was tested with the
real data from GIS of the streets and trash can locations: 3046 trashcans with about
18 division of trash cleaning teams. The dots represent the location of the trash cans
and the different colors are used to indicate the different division teams, in charge of
those trash bins (Fig. 4).
The trashcans were modelled based on the filling in the Poisson’s distribution.
The individual trash can i, is assigned as per the assigned time to reach Fi, with a
stipulated λ  7 for 7 regular days of the week. Also, the trashcan i, can be assigned
a daily filling volume of about DFi, in accordance to the poisons distribution with

er.ramankumar@aol.in
The Use of Modern Technology in Smart Waste Management … 179

Fig. 4 The street network and the location of the trashcans, under study in Copenhagen, Denmark
[2]

λ  1/Fi, where d corresponds to each day of the period under test. This helps us
determine the variance across the trash cans being filled.
The route creation for the division teams is associated with the optimization
techniques and algorithms adopted. For each individual team, a particular pint was
chosen to be set as the headquarters and all the routes assigned began from this point
and ended right here.
The economic restrictions were based on the collection costs, for which various
parameters like the number of workers, the driving speed, driving time, the fuel
consumption and the internet requirement sand one time set up costs are included
and analyzed against the efficiency of cleaning.
The performance parameters are evaluated based on different scenarios which
optimize the expenditure and also help the ultimate goal of a clean Smart city [2].

2.6 The Alternative Technology for Data Transfer

The other alternative technology used in the information transmission is using the
RFID tags and RFID readers. It is based on the preliminary radio technology. The
RFID tag is the combination of the antenna and the embedded microchip which store
an identical serial number to track the object or the person. With the RFID tag, the
radio frequency is transmitted over through the Reader’s Antenna. The RFID tag
modulated signals is received by the reader. The RFID tag is responsible to store the
trash data and transmit it to the RFID reader [12].
The active tags have longer range of service with their own internal battery to
support their power consumption. The Passive tags depend on the readers to suffice
the operating power and therefore have a lower range of just a few meters [13].
The RFID has the radio frequency transmitted once the reader in switched ON.
Once the RFID tag comes near to the reader antenna, the data is received and reverted

er.ramankumar@aol.in
180 P. K. Gupta et al.

back to the RFID reader. The most common reader used in the EM-18 and it is a
passive tag with the tag ID shifted to Arduino Microcontroller.
The Working of the RFID system:
The microchip contains a unique ID that integrates the circuitry for the proper
functioning of the tag. The internal EEPROM is essential to trace and keep the unique
ID. The tag used is passive so the antenna receives the power and the radio frequency
signals from the RFID reader. To recognize the RFID tag, the reader sends the data
back to the tag. The coil receives the signals in the form of alternating current and
passes it to the microchip [13] (Fig. 5).
The RFID tags and the readers are to be in the same frequency for the commu-
nication. The wide range of frequencies is left as the option for optimization and as
per the requirements. The low frequency is around 125 kHz and the high frequency
is about 13.56 MHz with the ultra-high frequencies varying between the 860 and
960 MHz. Most of the applications are also based on the 2.45 GHz frequency too.
The RFID tags are usually of CMOS IC EM4102, which is a passive RFID tag. The
reader generates a frequency of about 125 kHz and in occurrence of the 125 kHz
passive RFID tag, it will get energized from it.
The entire system is programmed to send a notification to the server once the trash
cans reach a certain level. The RFID reader is interfaced with the microcontroller
to enhance the process of verification. As the RFID tag interrupts the RFID reader,
the ultrasonic sensor calculates the level of trash filled in the dustbin. Based on the

Fig. 5 The working of the RFID tag [1]

er.ramankumar@aol.in
The Use of Modern Technology in Smart Waste Management … 181

level indicated the cleaning process is initiated. Usually the RFID tag is placed in
the trash bin and the RFID reader is placed with the antenna in the truck.

2.7 The Conclusion of Smart Waste Management

The efficient and proactive cleaning of the trashcans with the optimized technology
to save the manual effort, time and cost is highly appreciated and is in practice in most
societies. The model establishes a stronger waste collection system which reduces
the overflowing trash cans by the factor of 4. As the regular trash pickup is indicated
by the trash cans, the regularity of the check improves, thereby increasing the costs
by a margin of 13–25%. However, when the results are tallied across the required
goal of clean environments with regularized cleaning, the objective is achieved with
a high success rate.

3 Sorting and Recycling of Waste

The proper trash collection does not end the problem of waste management. The trash
collection just ensures that the garbage does not affect the livelihood of people and
animals. But piles of trash are filling up the land mines and ending up in the oceans.
This is degrading the quality of life and the ecosystem. In order to efficiently dispose
the garbage, it is necessary to sort them into biodegradable, recyclable and non-
biodegradable substances. This ensures reutilization of the biodegradable substances
to enrich the soil and aggravate cultivation. The recycling of waste materials right
from the metals to the paper and plastics, is high on trend. Since manual sorting is
difficult and unhygienic, there are several methods which involve robots and concepts
of machine learning. This section deals with the techniques involved by the latest
technologies to help recycle the massive piled up unsegregated trash all over.

3.1 Introduction

The recycling of materials like the steel, glass, and plastics helps to reduce the depen-
dency and extraction of raw materials along with preventing the pile up of dump
yards. The latest research has resulted in the implementation of machine learning
in the strategies to be used for material segregation, using an induction algorithm
(AIMS). This section highlights the procedure detecting plastic, glass and metal
objects from huge masses of trash by using light transmission techniques. The real-
ization of machine learning in practical scenario with high intelligence automation
used in garbage sorting and recycling is the future of robotics. A general municipal
cleanup is unnoticed until the cost of disposal of the waste through landfills or other

er.ramankumar@aol.in
182 P. K. Gupta et al.

means increase. The revenue potential of recycling can be boosted with established
intelligence automation methods.

3.2 Machine Learning: A Boon to Automation

While machine learning is aiming at the ultimate goal of zero human involvement, it
is used in many ways to simplify lifestyle of all. The unforeseen and less debated part
of machine learning resorts to the problems of unemployment and the corresponding
financial crisis that the world needs to prepare for. Though machine learning is doing
exceptionally well with respect to the human errors, we need to focus on developing
technology in areas which needs to be away from human intervention. This could be
places like nuclear power plants, space shuttles, aircraft and turbine testing, and so on.
While we integrate automation with the recycling industry, the artificial intelligence
used in the background for these super sorting bots is getting smarter by the day. The
future of human civilization is exaggerated in science fiction films using teleportation,
time machines, giant holograms, flying cars and movable apartments. But the subject
of how to handle the waste that is slowly killing the ecosystem and indirectly us, is
often left without a thought.
The Materials Recycling Facilities (MRF) is the new innovative solution to it all.

3.3 Statistics

In the era of self-absorbed millennials, the landfills are the major cause of increasing
Greenhouse gas (GHG) emissions and the significant factor of drastic global climate
change. The ‘Great Pacific Patch’ has brought in immense damage to the marine life,
killing thousands of sea birds and turtles along with huge mammals like the blue
whale.
Even though the industries aim at innovative zero waste missions, it can only
reduce up to 70% of the resulting waste [14]. When analyzed about the huge costs to
eliminate wastes, it was discovered that the entire problem was centered at improper
sorting measures.
If plastics were completely diverted from landfills, a potential energy equivalent
to 60 million barrels of oils can be saved. The landfills can be reduced to a great extent
(About 20%). The sorted recycling yields better results with lesser dependency on raw
materials for future developments [14]. There exists innumerous methods to extract
and segregate metals like aluminum, copper, and steel along with compostable waste
products, like paper and food waste. But the ones which highest degrading factor to
the environment, plastic and glass articles, have not yet been segregated with good
success rates.

er.ramankumar@aol.in
The Use of Modern Technology in Smart Waste Management … 183

Incorporating recognition algorithms with artificial intelligence and machine

learning coupled together, can owe to be a great solution for an automated process
of post-consumer plastic segregation.

3.4 The System Description

The ultimate efficiency is dependent on two main factors. The first being the amount
of material segregated by the system. In terms of statistics, this can be well inter-
preted by Type I error, which is the error of identifying an object incorrectly and
not removing it from the trash. The second factor is dedicated to the accuracy of the
sorting, which is determined by the Type II error. This indicates the scenario when
an object is wrongly detected and placed in another category [15].
The use of Adaptive and Interactive Modelling Systems (AIMS) is to demonstrate
the sensor data flow and to output a highly efficient description of object features
that completely define the segregation strategies. The system is designed to tackle the
challenges of real time data acquisition from the sensors and deliver the information
about the machine sensed and learned processing speeds [16].

3.5 Artificial Intelligence Based Sorting Techniques

The sorting of waste materials is highly essential to give way to recycling. There are
multiple ways to sort the waste materials which significantly determine the cost effi-
ciency and the output levels. The Source segregation involves the separation of recy-
clable materials into the corresponding different criteria, right at waste production
site. It is considered as the best mode of waste segregation by manual self-awareness
schemes. The source segregation mode of waste processing show the possibility of
waste recovery has the potential to more than 25% with steadily increasing positives.
On the other hand, for the fully mixed waste to be processed is labor intensive con-
ventionally. Automating this process is an excellent use of the technology and skill in
order to prevent the human interference and the corresponding vulnerability to res-
piratory and skin prone diseases. These lie on the higher end of financial investments
with a waste recovery rate of mere 15% [17]. The machine learning induction algo-
rithm that we propose to use here is capable of changing the conventional methods
and efficiency rates, to a better output (Fig. 6).
The machine learning induction algorithm is capable of adapting to the changing
situations and incorporating new conditions and discoveries with those preloaded
findings. One of the major advantages of this lies in the possibility and potential to
combine even unrelated and discrete sensor data into the process of decision making.
With time, the performance of the inductive system can exceed and go beyond that
of an expert and super intelligent system which might be constrained by the hard
coded and less acquired knowledge base [15].

er.ramankumar@aol.in
184 P. K. Gupta et al.

Fig. 6 Schematic of the waste stream segregation waste [20]

The figure depicts a conceptual understanding of the prototype proposed as the

Waste segregation system, its components and the way they interact and communi-
cate with each other. The Sensor interrogator and the database assimilator consists of
the PC-Type computers which is handled by a single computer capable of perform-
ing the tasks independently. The configuration illustrates the mode of learning by
the system. The machine is trained to detect glass, metal articles and plastic, using
a variety of possible shapes, deformation levels, sizes, color, and different levels
of contamination. Each substance is made to pass through a systematically placed
array of unique sensors. Their corresponding responses are recorded and a combined
database is established which possesses the accumulated sensor data and the inter-
preted information. All the containers will now be analyzed and checked using the
AIMS. The performance parameters set by the user will determine the comparison
and use of different induction algorithms.
As soon as the proposed models have been developed and the sensor data will be
sent to the interrogator for assessment of the accuracy and stability [18].

er.ramankumar@aol.in
The Use of Modern Technology in Smart Waste Management … 185

A machine learning algorithm is designed to learn the ways to perform a given

task based on the preloaded data and preset examples. This is referred to as example
driven learning. It has the capability of learning on its own through feedback it takes
from the environment based on its direct interaction. This is known to be the goal
driven aspect of machine learning.
We would like to discuss the example driven learning that requires the user to set
the list of possible scenarios as examples during the training of the machine code. The
examples presented to the machine is usually a vector consisting of input variables
along with the output variables. The learning algorithm is designed to produce a
function or the model that is capable of mapping the input variables to the output
variables [17].
The physical attributes in this waste segregation model is the specifications of the
container, such as the size, color, acoustic features and the optical density. This acts
as the input variable to the model. Other possible input variables that can be used
to describe the container is the details regarding the objects origin and the date of
trash collection. These act as a direct link to the system output. The output is the
determination of the waste segregation bins to which each of the waste object is to
be sent.
After feeding the machine with adequate number of examples, the example driven
learning algorithm is utilized to create a classification function. The type of the
learning algorithm and the related parameters is decided by the system. This is
essential for the desired accuracy, speed and the stability of the output [19].

3.6 Overview of the Induction Algorithm and Selection

Criteria

The AIMS is a highly flexible and powerful system which can operate on both goal
and example driven learning. It is known to be the meta-learning algorithm as it
has the potential to decide on the right learning algorithm and the corresponding
parameter settings with high levels of optimization techniques.
This is achieved by a search tool that finds out the one learning algorithm with
maximum positivity to the objective of the user. This optimization technique is useful
for the fulfillment of one objective. In case of multiple objectives the AIMS chooses
a set of models that are Pareto optimal, which is relatively the best to the given
objectives.
The main advantage of the classification based Pareto optimal models is that
these have high levels of flexibility and the perfect idealized behavior under different
circumstances.
For example: Let us set the objectives to the AIMS as, High accuracy, Minimal
time, and minimal use of the sensors.

er.ramankumar@aol.in
186 P. K. Gupta et al.

Once the Pareto models are considered, there might be a variation with respect to
the efficiency and the precision of the output. For best results, the user could opt for
a slower but accurate processing.
On a similar line, the volume of the input waste can increase to which a faster and
a less accurate model can be employed with faster conveyer movement. If the input
sensor undergoes repair or damage, the model is designed to work without the data
from that particular sensor.

3.7 Results

In order to exhibit the practical working and the corresponding understanding of

the working model, we have defined two sorting tasks. The first task involves the
determination of the bottles based on the light sensor data.
The second task was designed to classify the container into metal, glass and
plastics. This is based on the sensor data obtained from a single acoustic sensor [20].
Process Discussion:
The Bottle sorting techniques involved the process of testing the bottle quality. Each
bottle had to pass through a solar cell and a red Led, whose light had to be obstructed
from reaching the cell. The corresponding voltage fluctuation created a waveform of
150 samples per millisecond. This is the input vector and the corresponding output
vector was to indicate 0 or 1. 0 corresponds to unclear bottle and 1 to a clear bottle.
The error of random guess of the clear substance would be about 40%.
The acoustic sensors work on the Fast Fourier Transforms that operate on the input
signal. The materials were segregated into metals by recording the sound waves. The
FFf calculated and formulated the first 200 harmonics of the power spectrum with a
specified bandwidth. These were the input parameters of the object.
In order to classify the waste into three different classes, the output vector is
composed of Boolean numbers: (o1, o2, o3) in which the metal, glass and plastic
corresponded to (1, 0, 0), (0, 1, 0) and (0, 0, 1) respectively. The error of random
guess would be about 33% and the error of classifying all clear objects as glass would
be 42%.

3.8 Comparison of Existing Learning Algorithms

With the explanation on AIMS techniques, we have discovered 3 other learning algo-
rithms which effectively bring down the predictive errors. They are: Simple Recursive
Splitting (SRS), Neural Networks (NN), and the instance based (IB) (Table 1).

er.ramankumar@aol.in
The Use of Modern Technology in Smart Waste Management … 187

Table 1 To demonstrate the results obtained through various models [20]

Experiment Learning Classification error (%) Evaluation Formulation
type algorithm Mean Std. time (ms) time (s)
deviation
Glass sorting SRS 20.5 1.5 0.41 115
NN 33.8 3.4 7.96 253
IB 22.9 1.6 9227 0
Material sort SRS 32.7 2.7 0.43 198
NN 7.3 1.6 8.67 88
IB 12.3 1.6 2046 0

4 Conclusion

The use of intelligent automation techniques in sorting the waste materials is highly
feasible and time efficient. The Control systems have derived the knowledge of
machine learning induction algorithms to enhance system adaptability. The prelim-
inary sensor data from acoustic tests and light transmission have the potential to
differentiate clear and unclear objects. Apart from this they possess the ability to
wisely segregate glass, metal, and plastics using emitted sound waves. Electromag-
netic sensors are to be implemented for preliminary metal sorting.
Many companies have been establish with waste recycling robots that have revo-
lutionized this field of machine learning.
Some of the potential companies with valid investment in technology and objec-
tives of automated waste cleaning are Zen Robotics, BioHitech and Intel.

References

1. Kumar, N.S., Vuayalakshmi, B., Prarthana, R.J., Shankar, A.: IOT based smart garbage alert
system using Arduino UNO. In: IEEE Region 10 Annual International Conference, Proceed-
ings/TENCON (2017). https://doi.org/10.1109/TENCON.2016.7848162
2. Gutierrez, J.M., Jensen, M., Henius, M., Riaz, T.: Smart waste collection system based on
location intelligence. Procedia Comput. Sci. (2015). https://doi.org/10.1016/j.procs.2015.09.
170
3. Mitton, N., Papavassiliou, S., Puliafito, A., Trivedi, K.S.: Combining cloud and sensors in a
smart city environment. Eurasip J. Wirel. Commun. Netw. https://doi.org/10.1186/1687-1499-
2012-247
4. Sinha, T., Kumar, K.M., Saisharan, P.: Smart dustbin. Int. J. Ind. Electron. Electr. Eng. (2015)
5. Vicentini, F., Giusti, A., Rovetta, A., Fan, X., He, Q., Zhu, M., Liu, B.: Sensorized waste
collection container for content estimation and collection optimization. Waste Manag. (2009).
https://doi.org/10.1016/j.wasman.2008.10.017
6. Kim, B.-I., Kim, S., Sahoo, S.: Waste collection vehicle routing problem with time windows.
Comput. Oper. Res. https://doi.org/10.1016/j.cor.2005.02.045
7. Hannan, M.A., Arebey, M., Basri, H., Begum, R.A.: Intelligent solid waste bin monitoring and
management system. Aust. J. Basic Appl. Sci. (2010). https://doi.org/10.1007/s10661-010-
1642-x

er.ramankumar@aol.in
 188 P. K. Gupta et al.

8. Spira, P.M., Pan, A.: On finding and updating spanning trees and shortest paths. SIAM J.
Comput. (1975). https://doi.org/10.1137/0204032
9. Petit, J.: Experiments on the minimum linear arrangement problem. J. Exp. Algorithmics (JEA)
(2003). https://doi.org/10.1145/996546.996554
10. The traveling-salesman problem: Mathematics in Science and Engineering (1977). https://doi.
org/10.1016/S0076-5392(08)61182-0
11. López, J.G., Imine, M., Rumín, R.C., Pedersen, J.M., Madsen, O.B.: Multilevel network char-
acterization using regular topologies. Comput. Netw. https://doi.org/10.1016/j.comnet.2008.
04.014
12. Hunt, V.D., Puglia, A., Puglia, M.: RFID-A Guide to Radio Frequency Identification. https://
doi.org/10.1002/9780470112250
13. Ali, M.L., Alam, M., Rahaman, M.A.N.R.: RFID based e-monitoring system for municipal
solid waste management. In: 2012 7th International Conference on Electrical and Computer
Engineering, ICECE 2012 (2012). https://doi.org/10.1109/ICECE.2012.6471590
14. Torres-García, A., Rodea-Aragón, O., Longoria-Gandara, O., Sánchez-García, F., González-
Jiménez, L.E.: Intelligent waste separator. Computacion y Sistemas (2015). https://doi.org/10.
13053/CyS-19-3-2254
15. Tcheng, D.K., Lewis, M.R., Newell, T.A.: Applications of machine learning to sorting of
recyclable containers. In: 215 Conference on Solid Waste Research in the Midwest, Chicago,
Illinois, Oct 1990
16. Lewis, M.R., Newell, T.A.: Development of an automated clear/color sorting system for recy-
cling containers. In: Conference on Solid Waste Research in the Midwest, Chicago, Illinois,
Oct 1990
17. Tcheng, D.K., Lambert, B.L., Lu, S.C.-Y., Rendell, L.A.: Building robust learning systems by
combining induction and optimization. In: Proceedings of the International Joint Conference
on Artificial Intelligence, San Mateo, California. Morgan Kaufmann (1989)
18. Apotheker, S.: Garbage In, But What Comes Out? Resource Recycling, October 1990
19. Tcheng, D.K., Lambert, B.L., Lu, S.C.-Y., Rendell, L.A.: AIMS: an adaptive interactive mod-
eling system for supporting engineering decision making. In: Proceedings of the Eighth Inter-
national Workshop on Machine Learning, San Mateo, California. Morgan Kaufmann (1991)
20. Tarbell, K.A., Tcheng, D.K., Lewis, M.R., Newell, T.Y.A.: Applying machine learning to the
sorting of recyclable containers. University of Illinois at Urbana-Champaign, pp. 209–215,
Urbana, Illinois

er.ramankumar@aol.in
View publication stats