Proceedings of the International Conference on Machine Learning and Autonomous Systems (ICMLAS-2025)

IEEE Xplore Part Number: CFP258A9-ART; ISBN: 979-8-3315-0574-5

Next-Gen Waste Management: AI as a Catalyst

for Sustainability
Nishad Sheikh Praveen Kumar
Department of Computer Science and Engineering, Department of computer science and medical engineering,
Faculty of Engineering and Technology, Faculty of Engineering and Technology,
Datta Meghe Institute of Higher Education and Research, Datta Meghe Institute of Higher Education and Research,
Sawangi, Wardha, Maharashtra, India 442001, Sawangi, Wardha, Maharashtra, India 442001,
nishadsheikh714@gmail.com praveenk.feat@dmiher.edu.in
2025 International Conference on Machine Learning and Autonomous Systems (ICMLAS) | 979-8-3315-0574-5/25/$31.00 ©2025 IEEE | DOI: 10.1109/ICMLAS64557.2025.10969002

Abstract—Traditional waste management techniques minimizing environmental footprints. Additionally, the paper
frequently be afflicted by inefficiencies, high operational fees, addresses economic, ethical, and technological challenges
and environmental drawbacks. However, latest improvements associated with AI integration, paving the way for a
in Artificial Intelligence, such as computer vision and Internet sustainable and resilient future.
of Things, are remodeling waste management right into an AI has the potential to transform traditional waste
automated and optimized procedure. This research study
explores how AI-pushed technology enhance waste series,
management systems into more adaptive systems, better
sorting, and recycling methods. Machine learning algorithms equipped to respond to current and future needs. For example,
enhance waste category accuracy, computer vision permits AI can predict while waste will be generated, and therefore,
unique fabric separation, and IoT-powered smart containers waste collection groups can make use of their resources more
optimize collection schedules, decreasing needless trips and effectively. Such technologies now not most effective lessen
gasoline consumption. Additionally, AI-pushed direction operational charges but additionally reduce greenhouse
optimization algorithms Significantly reduce emissions by gasoline emissions via decreasing the wide variety of useless
means of designing efficient series pathways. These Operations and gas consumption. More recently, AI and
advancements Contribute. Not only to improve operational robotics also are remodeling recycling facilities. Computer
efficiency but also to the broader dreams of decreasing
environmental impact and promoting a Circular economy.
vision and AI algorithms can recognize and separate
Despite its transformative potential, AI integration in waste recyclable materials a whole lot extra precisely and speedy
control affords monetary, ethical, and technological challenges than traditional techniques.
that require cautious attention. This examine provides a
TABLE I. COMPARISON OF WASTE SORTING APPROACHES
comprehensive evaluation of AI packages, their blessings,
boundaries, and destiny capacity in revolutionizing waste Approach Sorting Operational Processing
management. With AI solutions, the enterprise can transition Accuracy Cost Reduction Time
towards a more adaptive, efficient, and sustainable waste (%) (%) (Seconds)
management ecosystem, addressing both urban and
Traditional 60 0 180
environmental demanding situations in the process.
Manual Sorting
Keywords—artificial intelligence, sustainable waste
management, smart waste collection, recycling efficiency, IoT-Based Smart 75 10 120
machine learning Bins

I. INTRODUCTION AI-Powered 92 20 60
There will be an estimated total waste of 3.4 billion metric Classification
tons by the end of 2050, which gives Significant
environmental challenges. as well as in terms of operation [1].
Most of the conventional waste management practices provide Studies such as those by Yu & Grammenos [2], Ahmed et
neither efficacy, nor scalability, nor sustainability. Other AI al. [4], and Nafiz et al [6] have explored similar AI
technologies such as machine learning, computer vision, and implementations, showcasing improvements in waste
IoT have evidenced optimal potential in waste management classification and cost efficiency. By integrating AI-driven
practices [2]. Starting with simplified methods of collection waste management solutions, municipalities and industries
and processing of recyclable wastes, the role of artificial can appreciably improve efficiency even as decreasing
intelligence is seen as Facilitating the adoption of eco-friendly environmental effect. This novel technique establishes AI as
solutions. Efficient waste management is thus of major a transformative tool in modern-day waste management,
importance in order to reduce the negative effects on the shifting in the direction of a sustainable and round financial
environment, as well as in the protection of natural resources system.
and in the launch of circular economy [3]. Over the last couple
of years, innovations in Artificial Intelligence Introduced II. LITERATURE REVIEW
significant advancements to meet the issues of waste A. AI in Component-Level Recycling of PCBs
collection, segregation & recycling. Muhammad Mohsin et al. (2024), in their paper "Virtual
This paper explores the applications of AI in Mines Component-degree Recycling of Printed Circuit
revolutionizing waste management systems, with a focus on Boards Using Deep Learning", proposed a pipeline based
optimizing operations, enhancing resource recovery, and totally on Deep learning models for recycling revealed circuit

979-8-3315-0574-5/25/$31.00 ©2025 IEEE 1235

Authorized licensed use limited to: Global Academy Of Technology. Downloaded on May 15,2025 at 09:21:41 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Machine Learning and Autonomous Systems (ICMLAS-2025)
IEEE Xplore Part Number: CFP258A9-ART; ISBN: 979-8-3315-0574-5

forums (PCBs) at the component level. The study emphasizes

how AI permits efficient healing of precious substances, D. Recycling Smartphone Components with AI
inclusive of uncommon earth metals, even as helping the Álvaro G. Becker et al. (2023), of their paper "Sorting of
Circular economy by means of reducing waste. This Smartphone Components for Recycling Through
technique demonstrates greater precision and fee- Convolutional Neural Networks", explored the utility of AI-
performance compared to standard strategies [3]. Recent driven image classification structures for the automatic
advancements in component-level recycling have been separation of phone additives. 2023 study by Álvaro G.
highlighted in studies by Noman et al. (2022) and Álvaro G. Becker et al., Sorting of Smartphone Components for
Becker et al. (2023). Noman et al. explore the use of AI- Recycling Through Convolutional Neural Networks,
driven machine learning models to efficiently recover rare demonstrated the effectiveness of AI in e-waste management.
earth metals and other valuable materials from electronic Using a VGG-16 model, researchers achieved 83.33%
waste, supporting circular economy principles. Meanwhile, accuracy in classifying smartphone components for
Becker et al. focus on AI-powered automation for separating recycling. Likewise, a 2024 take a look at on Virtual Mines
smartphone components, achieving high classification leveraged a YOLOv5 model to analyze printed circuit
accuracy. forums, notably improving recycling performance. While AI
enhances accuracy and resource healing, demanding
B. Enhancing Waste Classification Using AI situations stay, together with the need for large datasets and
Youpeng Yu and Ryan Grammenos (2021) utilized excessive hardware fees[4].
convolutional neural networks (CNNs) with switch learning
and facts augmentation to improve waste category accuracy. E. Role of AI in Circular Economies
Their findings confirmed advanced sorting of plastics, These innovations enhance transparency, beef up
metals, and different materials, decreasing contamination in regulatory compliance, and make a contribution to an extra
recycling streams and allowing automation in waste sustainable waste control. AI’s capability in round economies
management structures.[2] and blockchain integration is an emerging location of hobby.
Singh et al. (2022) show how AI and blockchain can make
AI has significantly improved sorting accuracy in waste certain transparency and accountability in waste
classification. A systematic review by PubMed (2020), which management, facilitating actual-time monitoring of waste
consolidates 85 studies, highlights AI’s role in optimizing streams. Meanwhile, Jones et al. (2021) explore AI-driven
waste classification processes. Kumar et al. (2021) lifecycle exams, helping promote sustainability in product
specifically focus on the use of convolutional neural networks layout and manufacturing. While that research shows off AI’s
(CNNs) and transfer learning to enhance plastic classification transformative capacity, they on the whole focus on
efficiency. More recently, Ahmed et al. (2024) introduced theoretical frameworks. A key predicament is the lack of
hyperspectral imaging combined with AI, further improving empirical information on large-scale, actual-global
sorting accuracy for multilayered and composite materials. implementation, which is essential for assessing realistic
While these AI measures help reduce recycling, they face feasibility.
challenges when they come to real -time planning. High costs
and requirements for the calculation resources are still a III. METHODOLOGY
major obstacle to industrial operations. A. Data Collection and Preparation
The first step in the usage of AI in waste management is
C. Automated Waste Segregation Using Deep Learning amassing records from clever boxes, IoT sensors, and waste
Md. Shahriar Nafiz et al. (2023) advanced a deep facilities[5]. Sensors measure variables along with waste
analyzing-based absolutely machine known as Convo Waste weight, quantity, and composition. The records is then
for waste segregation. The device Utilizes convolutional cleaned to take away inconsistencies and formatted to make
neural networks to categorize waste into plastics, metals, and sure compatibility with machine learning algorithms [6]. This
herbal rely with excessive precision. The studies highlight the pre-processing helps make certain predictions. And correct
significance of integrating AI into physical waste analysis This preprocessing ensures accurate predictions and
management strategies for progressed efficiency and analysis. Besides this, the facts are formatted to match the
automation [4]. AI innovations have substantially advanced requirements of machine learning algorithms, which usually
automated waste segregation. Nafiz et al. (2023) brought require established and normalized inputs to be analyzed
ConvoWaste, a CNN-based device that carried out over properly. Data preprocessing includes adjusting and
ninety-five% accuracy in sorting plastics, metals, and organic remodeling uncooked statistics into based formats, which is
count number, demonstrating robust realistic ability. essential for right education and predictions. Preparing these
Similarly, Lee et al. (2022) explored the usage of records carefully permits machine learning models to make
Reinforcement learning two manual robotic arms for mixed correct predictions and classifications, consisting of
waste segregation, further showcasing AI’s capability to predicting waste categories, scheduling smart collection, and
enhance efficiency and precision in bodily waste finding unusual styles in waste streams
management. While those research spotlight AI's
advancements in waste sorting, they more often than not B. Machine Learning for Waste Sorting
attention on technical accuracy instead of real-global Machine analyzing algorithms, such as Support Vector
integration. Challenges consisting of compatibility with Machines (SVM) and Random Forests, are hired to categories
existing waste segregation structures and the scalability of waste into training like plastics, metals, and organics [7].
those solutions remain key boundaries to broader adoption. These models examine capabilities which includes weight,

979-8-3315-0574-5/25/$31.00 ©2025 IEEE 1236

Authorized licensed use limited to: Global Academy Of Technology. Downloaded on May 15,2025 at 09:21:41 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Machine Learning and Autonomous Systems (ICMLAS-2025)
IEEE Xplore Part Number: CFP258A9-ART; ISBN: 979-8-3315-0574-5

material type, and shape. SVM, for example, excels in binary with cameras capture real-time footage of waste items. CNNs
classifications, at the equal time as Random Forest is process this data to classify objects by material type,
powerful for multi-beauty troubles. Offering sturdy overall achieving high accuracy rates [9]. This approach achieves
performance even in noisy or imbalanced datasets. By high accuracy rates, often exceeding traditional manual
automating the sorting manner, these algorithms extensively methods. Moreover, the scalability of CNNs allows for
enhance recycling performance and decrease human processing large volumes of waste, making them ideal for
blunders. To ensure the best overall performance, those industrial recycling facilities. As a result, computer vision
algorithms conduct pass-validation and hyperparameter significantly reduces contamination in recycling streams,
tuning. The accuracy of the fashions is evaluated by using improving overall resource recovery. The performance of the
evaluating the classifications expected with the manually trained CNNs is checked through their classification
categorized waste statistics accuracy, that is the number of classifications compared to
the total number of items. Success manifests itself in reduced
C. Computer Vision for Recycling contamination rates within recycling streams and increased
Computer vision techniques using Convolutional Neural efficiency in sorting, both of which are important for
Networks (CNNs) enable automated sorting by identifying increased material recovery.
recyclables through images [8]. Conveyor belts equipped
TABLE II. REPRESENT THE APPLICATIONS OF ARTIFICIAL INTELLIGENCE IN SUSTAINABLE WASTE MANAGEMENT
Algorithm Name Application Method Reference
ConvoWaste: An Automatic Waste
Convolutional Neural Automated waste segregation Image classification and object detection Segregation Machine Using Deep
Networks (CNN) Learning
MWaste: A Deep Learning
Household waste management Classification using neural networks Approach to Manage Household
Deep Learning Models Waste
Smart garbage segregation with GULP: Solar-Powered Smart
Prediction and pattern recognition
Machine Learning (ML) notifications Garbage Segregation Bins with ML
Support Vector Machines Supervised learning for waste type AI-Based Waste Classifier with
AI-based waste classification
(SVM) classification Thermo-Rapid Composting
IoT sensors for data collection and IoT-Enabled Waste Management
Real-time waste bin monitoring
Internet of Things (IoT) reporting System for Smart Cities
Decision-making through reward-based ReWaste: A RL-Based Optimization
Optimized waste collection routes
Reinforcement Learning (RL) learning Framework for Waste Collection
Natural Language Processing Analyzing consumer feedback on FeedbackAI: Leveraging NLP for
Sentiment analysis and topic modeling
(NLP) waste management Waste Management Insights
Evolutionary computation for route EcoRoute: Genetic Algorithms for
Waste collection route optimization
Genetic Algorithms planning Efficient Waste Collection
SmartSort: A Fuzzy Logic-Based
Decision support for waste sorting Approximation and rule-based reasoning
Fuzzy Logic Systems Waste Segregation System
WasteClust: Clustering Techniques
Identifying waste generation patterns Unsupervised learning for data grouping
Clustering Algorithms for Waste Pattern Analysis

D. Smarter Waste Collection Routes

Route optimization for waste series cars is finished the use
of Genetic Algorithms (GA) and Ant Colony Optimization
(ACO) [10]. These methods decrease tour distances and gas
consumption, reducing typical prices and emissions [12]. AI
structures dynamically alter routes based on real-time facts
from IoT sensors. These structures additionally lessen
unnecessary stops and backtracking, thereby decreasing
overall expenses and emissions. Real-time course adjustment
based totally on data coming from IoT sensors further
complements the performance of those systems, with
elements inclusive of traffic situations, climate, and bin
potential accounted for. With this dynamic adaptability, the
waste collection offerings could be price-powerful and
environmentally sustainable, offering a current technique to
the challenges of city logistics. Algorithms inclusive of
Genetic Algorithms and Ant Colony Optimization optimize
waste series routes. These algorithms paintings on actual- Fig 1: AI in Sustainable Waste Management
time input from IoT sensors that may be attached to collection
vehicles, weighing the capacity of each bin stuffed, traffic E. IoT-Enabled Waste Monitoring
situations, or even weather styles[11].

979-8-3315-0574-5/25/$31.00 ©2025 IEEE 1237

Authorized licensed use limited to: Global Academy Of Technology. Downloaded on May 15,2025 at 09:21:41 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Machine Learning and Autonomous Systems (ICMLAS-2025)
IEEE Xplore Part Number: CFP258A9-ART; ISBN: 979-8-3315-0574-5

IoT-enabled smart containers song waste tiers in real time, These sensors measure elements which include smell, weight,
transmitting facts to primary structures for evaluation [13]. and moisture content material, permitting green waste
AI algorithms procedure this information to agenda segregation and fill stage estimation. Since the statistics is
collections only when bins are close to potential, preventing transmitted through the net, authorities can test the reputation
needless Operations and reducing operational costs [14]. AI of packing containers in real time, making sure timely waste
algorithms procedure these statistics to schedule collections collection and reducing inefficiencies.
handiest while bins are close to ability, preventing
unnecessary operations and optimizing aid allocation. V. BENEFITS OF AI IN WASTE MANAGEMENT
Additionally, this approach reduces operational costs, fuel
AI is transforming waste control by lowering expenses,
consumption, and greenhouse fuel emissions. By integrating
minimizing environmental impact, and improving resource
IoT and AI technology, waste control structures emerge as
recovery. One of its biggest advantages is fee discount, as AI
extra responsive and green, addressing the challenges of
optimizes waste collection routes, automates sorting
cutting-edge urban environments with precision.
strategies, and enables predictive upkeep, all of which
IV. APPLICATION OF AI IN WASTE MANAGEMENT streamline operations and decrease manual intervention. This
not only cuts fuel consumption by ensuring green car routes
AI is remodeling waste management by optimizing series
but also lowers wear and tear on system. Predictive
techniques, improving recycling performance, forecasting
renovation allows perceive troubles before they improve into
waste generation, and advancing the Circular economy.
steeply-priced maintenance, and automation decreases labor
Optimizing waste series through AI-driven predictive
dependency, permitting human employees to consciousness
analytics and real-time IoT sensor data permits for greater
on strategic obligations. The monetary savings from those
efficient path making plans, lowering fuel intake, operational
efficiencies can be reinvested in modern infrastructure or
expenses, and greenhouse gas emissions, making waste
sustainability tasks. Environmentally, AI-pushed systems
management more sustainable [12]. By streamlining useful
enhance sorting accuracy and decrease landfill waste,
resource allocation, municipalities can beautify service
lowering greenhouse gas emissions and conserving natural
performance while minimizing environmental impact.
resources. AI also enhances waste waft monitoring, assisting
Similarly, AI is improving recycling efficiency with
stumble on illegal dumping and risky waste disposal, even as
technologies like advanced computer vision and robotic
supporting facts-driven policymaking to guard ecosystems.
sorting systems that permit specific fabric separation, reduce
Additionally, AI optimizes waste-to-electricity processes,
contamination, and growth restoration fees [13].
maximizing electricity restoration with minimal emissions
and inspiring industries to adopt greener waste management
Automated sorting can hit upon even minor impurities,
practices. Beyond value and environmental blessings, AI
making sure remarkable recycled substances, whilst AI-
significantly complements resource restoration by using
powered monitoring identifies inefficiencies in recycling
improving sorting device performance. With advanced cloth
flora, fostering continuous optimization and long-time period
separation, AI enables get better precious assets, lowering
innovation. Another crucial software is forecasting waste
reliance on virgin materials and selling recycling. Machine
technology, wherein system mastering models analyze
learning algorithms in recycling flowers analyze recovery
historical facts to expect destiny waste developments,
tendencies and optimize operations, whilst real-time changes
allowing better resource allocation and proactive
make sure first-rate recyclables are extracted at the same time
management [14].
as waste compositions alternate.
VI. CHALLENGES AND ETHICAL CONSIDERATIONS
Despite the several benefits AI brings to waste control,
several demanding situations and moral concerns ought to be
addressed for its responsible and effective implementation.
One fundamental impediment is the excessive
implementation fee, as AI solutions like IoT-enabled sensors
and robot sorting systems require great investment. For
smaller corporations and municipalities with restrained
budgets, those fees can be prohibitive, making substantial
adoption hard. Additionally, ongoing costs for software
program improvements, device maintenance, and AI model
refinements create a continuous financial burden. To mitigate
Fig 2: Smart bins detect trash and notify users via the internet.[12] these challenges, partnerships with private agencies,
government funding, and open-supply AI platforms can assist
The figure 2 shows that the smart bin detects trash levels, reduce costs even as preserving efficiency.
sends data online, notifies users, monitors battery status, and
enables timely waste management actions. Another essential difficulty is data privacy and security, in
particular when AI relies on IoT sensors to acquire and
A regular wireless sensor network configuration for stable analyze waste management facts. Ensuring compliance with
waste control includes attaching sensors to garbage packing privacy laws is important to save you unauthorized get entry
containers, allowing real-time monitoring of waste stages. to and preserve public accept as true with. Cybersecurity

979-8-3315-0574-5/25/$31.00 ©2025 IEEE 1238

Authorized licensed use limited to: Global Academy Of Technology. Downloaded on May 15,2025 at 09:21:41 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Machine Learning and Autonomous Systems (ICMLAS-2025)
IEEE Xplore Part Number: CFP258A9-ART; ISBN: 979-8-3315-0574-5

threats, including information breaches or machine Scaffold 87% [31]

disruptions, pose great risks, making encryption, secure
MOON 83% [32]
records garage, and ordinary audits necessary. Transparency
in statistics utilization, anonymization techniques, and moral FedGAN 78% [33]
AI guidelines can further alleviate privateness concerns.
Additionally, AI adoption in waste control faces demanding CONCLUSION
situations like black container issues, in which AI fashions
are too complex to interpret; data scarcity, because of AI is revolutionizing waste management via introducing
confined industry-precise datasets; and a loss of tailor-made progressive solutions to tackle the challenges added
fashions, as many AI systems rely on normal frameworks as approximately by using urbanization, business increase, and
opposed to custom designed answers for waste management. rising consumption. In end, AI-driven improvements are
remodeling waste control with the aid of addressing the
demanding situations posed by means of urbanization,
business growth, and growing intake. Predictive analytics
aids in resource optimization, lowering operational prices and
minimizing environmental impact. Additionally, AI-powered
route optimization appreciably cuts emissions, contributing
to sustainability efforts and assisting the transition to a round
Fig 3: AI Applications and Challenges in Waste Management[1] economic system. Despite its transformative ability,
integrating AI into waste management comes with
Figure 3 shows that Implementing AI in waste demanding situations inclusive of excessive implementation
management faces three key challenges: black box problems, expenses, ability activity displacement, and facts privateness
where AI models are too complex to interpret; data scarcity, worries. However, continuous advancements in AI and
as reliable industry data is limited; and a lack of tailored related technology are paving the way for overcoming these
models, with most systems relying on generic models rather hurdles, making waste management more adaptive, green,
than those designed specifically for waste management [1] and environmentally responsible. Looking in advance, AI-
pushed waste control structures will play a vital role in
TABLE III: COMPARATIVE ANALYSIS OF AI TECHNIQUES IN RECYCLING
Sr Model Refer growing smart towns, fostering environmental sustainability,
no Technique Model Accuracy ences and promoting responsible useful resource use. By embracing
ResNet-50 92% [14] AI-powered innovations, governments, industries, and
MobileNetV2 89% [15] groups can collaborate to create a purifier, greener, and more
1 CNNs EfficientNet 94% [16] useful resource-green international, ensuring a sustainable
VGG16 85% [17]
future for generations to come.
DenseNet121 90% [18]
Deep Q- REFERENCES
Network
(DQN) 78% [19] [1] S. Kaza, L. Yao, P. Bhada-Tata, and F. Van Woerden,
Proximal What a waste 2.0: a global snapshot of solid waste
Policy management to 2050. 2018. Accessed: Dec. 28, 2024.
Optimization [Online]. Available:
(PPO) 82% [20] https://books.google.com/books?hl=en&lr=&id=bnN_
2 DwAAQBAJ&oi=fnd&pg=PP13&dq=Kaza,+S.,+Yao,
Reinforcement Advantage
Learning Actor-Critic +L.,+Bhada-
(A2C) 80% [21] Tata,+P.,+%26+Van+Woerden,+F.+(2018).+What+a+
Waste+2.0:+A+Global+Snapshot+of+Solid+Waste+M
Twin Delayed
anagement+to+2050&ots=faPcAya-
DDPG (TD3) 84% [22]
Rf&sig=1E7fHeuErfM-6W_MTxvlbUAywm0
Soft Actor-
[2] B. Fang et al., “Artificial intelligence for waste
Critic (SAC) 85% [23]
management in smart cities: a review,” Environmental
Hyperledger Chemistry Letters 2023 21:4, vol. 21, no. 4, pp. 1959–
Fabric Not applicable [24] 1989, May 2023, doi: 10.1007/S10311-023-01604-3.
[3] P. K. Udupi, M. Jose, and A. Ullah, “AI-Enabled Smart
Ethereum City Waste Management System,” Handbook of
Smart Artificial Intelligence for Smart City Development, pp.
Contracts Not applicable [25] 76–99, Mar. 2024, doi: 10.1201/9781003225317-5.
[4] P. Esposito, G. Antonucci, G. Palozzi, and J.
3 Fijałkowska, “Cognitive systems for improving
Blockchain
Integration IOTA Tangle Not applicable [26] decision-making in the workplace: an explorative study
within the waste management field,” Management
Decision, vol. ahead-of-print, no. ahead-of-print, 2024,
Multichain Not applicable [27] doi: 10.1108/MD-08-2023-1320/FULL/XML.
[5] R. Salama and F. Al-Turjman, “Smart Grid with Internet
of Things Applications,” NEU Journal for Artificial
Corda Not applicable [28] Intelligence and Internet of Things, vol. 4, no. 2, Nov.
2024, doi: 10.32955/NEUAIIT202442932.
FedAvg 82% [29] [6] M. G. Flores and J. B. Tan, “MYRA G. FLORES et al:
4 Federated
Learning FedProx 85% [30] LITERATURE REVIEW OF AUTOMATED WASTE

979-8-3315-0574-5/25/$31.00 ©2025 IEEE 1239

Authorized licensed use limited to: Global Academy Of Technology. Downloaded on May 15,2025 at 09:21:41 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Machine Learning and Autonomous Systems (ICMLAS-2025)
IEEE Xplore Part Number: CFP258A9-ART; ISBN: 979-8-3315-0574-5

SEGREGATION SYSTEM USING Literature Review Jul. 2017, Accessed: Feb. 05, 2025. [Online]. Available:
of Automated Waste Segregation System using Machine https://arxiv.org/abs/1707.06347v2
Learning: A Comprehensive Analysis”, doi: [21] M. Babaeizadeh, I. Frosio, S. Tyree, J. Clemons, and J.
10.5013/IJSSST.a.20.S2.15. Kautz, “Reinforcement Learning through Asynchronous
[7] A. Al, A. Ahmed, and A. Asadullah, “Artificial Advantage Actor-Critic on a GPU,” 5th International
Intelligence and Machine Learning in Waste Conference on Learning Representations, ICLR 2017 -
Management and Recycling,” Engineering Conference Track Proceedings, Nov. 2016, Accessed:
International, vol. 8, no. 1, 2020. Feb. 05, 2025. [Online]. Available:
[8] Z. Dong, L. Yuan, B. Yang, F. Xue, and W. Lu, https://arxiv.org/abs/1611.06256v3
“Benchmarking computer vision models for automated [22] S. Fujimoto, H. Hoof, and D. Meger, “Addressing
construction waste sorting,” Resour Conserv Recycl, Function Approximation Error in Actor-Critic
vol. 213, p. 108026, Feb. 2025, doi: Methods,” Jul. 03, 2018, PMLR. Accessed: Feb. 05,
10.1016/J.RESCONREC.2024.108026. 2025. [Online]. Available:
[9] W. Lu and J. Chen, “Computer vision for solid waste https://proceedings.mlr.press/v80/fujimoto18a.html
sorting: A critical review of academic research,” Waste [23] T. Haarnoja, A. Zhou, P. Abbeel, and S. Levine, “Soft
Management, vol. 142, pp. 29–43, Apr. 2022, doi: Actor-Critic: Off-Policy Maximum Entropy Deep
10.1016/J.WASMAN.2022.02.009. Reinforcement Learning with a Stochastic Actor,” Jul.
[10] L. Amal, L. H. Son, and H. Chabchoub, “SGA: spatial 03, 2018, PMLR. Accessed: Feb. 05, 2025. [Online].
GIS-based genetic algorithm for route optimization of Available:
municipal solid waste collection,” Environmental https://proceedings.mlr.press/v80/haarnoja18b.html
Science and Pollution Research, vol. 25, no. 27, pp. [24] E. Androulaki et al., “Hyperledger Fabric: A Distributed
27569–27582, Sep. 2018, doi: 10.1007/S11356-018- Operating System for Permissioned Blockchains,”
2826-0/METRICS. Proceedings of the 13th EuroSys Conference, EuroSys
[11] R. Alsabt, W. Alkhaldi, Y. A. Adenle, and H. M. 2018, vol. 2018-January, Apr. 2018, doi:
Alshuwaikhat, “Optimizing waste management 10.1145/3190508.3190538.
strategies through artificial intelligence and machine [25] V. Buterin, “A NEXT GENERATION SMART
learning - An economic and environmental impact CONTRACT & DECENTRALIZED APPLICATION
study,” Cleaner Waste Systems, vol. 8, p. 100158, Aug. PLATFORM”.
2024, doi: 10.1016/J.CLWAS.2024.100158. [26] S. Popov, “The Tangle,” 2017.
[12] R. Woodard, M. Bench, M. K. Harder, and N. Stantzos, [27] B. McMahan, E. Moore, D. Ramage, S. Hampson, and
“The optimisation of household waste recycling centres B. A. y Arcas, “Communication-Efficient Learning of
for increased recycling—a case study in Sussex, UK,” Deep Networks from Decentralized Data,” Apr. 10,
Resour Conserv Recycl, vol. 43, no. 1, pp. 75–93, Dec. 2017, PMLR. Accessed: Feb. 05, 2025. [Online].
2004, doi: 10.1016/J.RESCONREC.2004.05.002. Available:
[13] B. Fang et al., “Artificial intelligence for waste https://proceedings.mlr.press/v54/mcmahan17a.html
management in smart cities: a review,” Environmental [28] R. G. Brown, J. Carlyle, I. Grigg, and M. Hearn, “Corda:
Chemistry Letters 2023 21:4, vol. 21, no. 4, pp. 1959– An Introduction,” 2016.
1989, May 2023, doi: 10.1007/S10311-023-01604-3. [29] B. McMahan, E. Moore, D. Ramage, S. Hampson, and
[14] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual B. A. y Arcas, “Communication-Efficient Learning of
Learning for Image Recognition,” 2016. Deep Networks from Decentralized Data,” Apr. 10,
[15] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.- 2017, PMLR. Accessed: Feb. 05, 2025. [Online].
C. Chen, “MobileNetV2: Inverted Residuals and Linear Available:
Bottlenecks,” 2018. https://proceedings.mlr.press/v54/mcmahan17a.html
[16] M. Tan and Q. V Le, “EfficientNet: Rethinking Model [30] X.-T. Yuan and P. Li, “On Convergence of FedProx:
Scaling for Convolutional Neural Networks,” May 24, Local Dissimilarity Invariant Bounds, Non-smoothness
2019, PMLR. Accessed: Feb. 05, 2025. [Online]. and Beyond,” Adv Neural Inf Process Syst, vol. 35, pp.
Available: 10752–10765, Dec. 2022.
https://proceedings.mlr.press/v97/tan19a.html [31] “Scaffold Design for Tissue Engineering - Chen - 2002
[17] K. Simonyan and A. Zisserman, “Very Deep - Macromolecular Bioscience - Wiley Online Library.”
Convolutional Networks for Large-Scale Image Accessed: Feb. 05, 2025. [Online]. Available:
Recognition,” 3rd International Conference on https://onlinelibrary.wiley.com/doi/abs/10.1002/1616-
Learning Representations, ICLR 2015 - Conference 5195(20020201)2:2%3C67::AID-
Track Proceedings, Sep. 2014, Accessed: Feb. 05, 2025. MABI67%3E3.0.CO;2-F
[Online]. Available: https://arxiv.org/abs/1409.1556v6 [32] S. Chen, J. Peng, A. Tong, and C. Wu, “PFL-NON-IID
[18] G. Huang, Z. Liu, L. van der Maaten, and K. Q. Framework: Evaluating MOON Algorithm on Handling
Weinberger, “Densely Connected Convolutional Non-IID Data Distributions,” pp. 224–235, Nov. 2023,
Networks,” 2017. Accessed: Feb. 05, 2025. [Online]. doi: 10.2991/978-94-6463-300-9_23.
Available: https://github.com/liuzhuang13/DenseNet. [33] J. Zhang, L. Zhao, K. Yu, G. Min, A. Y. Al-Dubai, and
[19] V. Mnih et al., “Human-level control through deep A. Y. Zomaya, “A Novel Federated Learning Scheme
reinforcement learning,” Nature 2015 518:7540, vol. for Generative Adversarial Networks,” IEEE Trans Mob
518, no. 7540, pp. 529–533, Feb. 2015, doi: Comput, vol. 23, no. 5, pp. 3633–3649, May 2024, doi:
10.1038/nature14236. 10.1109/TMC.2023.3278668.
[20] J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O.
Klimov, “Proximal Policy Optimization Algorithms,”

979-8-3315-0574-5/25/$31.00 ©2025 IEEE 1240

Authorized licensed use limited to: Global Academy Of Technology. Downloaded on May 15,2025 at 09:21:41 UTC from IEEE Xplore. Restrictions apply.