Solid waste classification: An approach based on reinforcement learning and differential evolution

Si Li; Si Li

doi:10.3934/environsci.2026019

AIMS Environmental Science

2026, Volume 13, Issue 3: 458-486. doi: 10.3934/environsci.2026019

Previous Article Next Article

Solid waste classification: An approach based on reinforcement learning and differential evolution

Si Li ^,

Liaoning institute of science and engineering, Jinzhou, Liaoning, 121000, China

Received: 11 December 2025 Revised: 24 May 2026 Accepted: 10 June 2026 Published: 29 June 2026

As cities grow and intelligent urban areas expand, effectively managing escalating waste volumes, from their generation and sorting to their final disposal, becomes increasingly essential. This investigation introduces a sophisticated deep-learning (DL) approach for categorizing solid waste into multiple waste categories, including glass, metal, paper, plastic, cardboard, and trash. Our model addresses issues overlooked in previous studies, including class imbalance and sensitivity to hyperparameters in classification tasks. It uses wide, dilated convolutional layers that adeptly identify and combine key features for precise classification. Addressing class imbalance, we implement a reinforcement learning (RL) approach, where the agent evaluates each sample individually and classifies it. For every accurate classification, the agent earns rewards, whereas inaccuracies lead to penalties, with greater penalties/rewards applied to the less prevalent class. This system enables the agent to develop an optimal strategy guided by specific reward functions and a well-defined learning environment. To enhance hyperparameter optimization, the proposed approach improves the differential evolution (DE) algorithm using a clustering-guided mutation strategy based on k-means clustering. This strategy clusters the candidate hyperparameter population using k-means, selects the cluster with the lowest average objective value, and uses the best candidate within that cluster to guide the mutation process. A unique method revitalizes the candidate solutions throughout the population, advancing the hyperparameter adjustment process. Exhaustive evaluations on the TrashNet and Trash datasets demonstrate the efficacy of our model, achieving a high classification accuracy of 89.908% on TrashNet and 87.438% on Trash. These findings highlight the system's capacity to properly handle particular difficulties in solid waste classification, especially data imbalance and hyperparameter sensitivity.
- solid waste management,
- reinforcement learning,
- imbalanced classification,
- differential evolution,
- hyperparameter optimization
Citation: Si Li. Solid waste classification: An approach based on reinforcement learning and differential evolution[J]. AIMS Environmental Science, 2026, 13(3): 458-486. doi: 10.3934/environsci.2026019

Related Papers:

Abstract

As cities grow and intelligent urban areas expand, effectively managing escalating waste volumes, from their generation and sorting to their final disposal, becomes increasingly essential. This investigation introduces a sophisticated deep-learning (DL) approach for categorizing solid waste into multiple waste categories, including glass, metal, paper, plastic, cardboard, and trash. Our model addresses issues overlooked in previous studies, including class imbalance and sensitivity to hyperparameters in classification tasks. It uses wide, dilated convolutional layers that adeptly identify and combine key features for precise classification. Addressing class imbalance, we implement a reinforcement learning (RL) approach, where the agent evaluates each sample individually and classifies it. For every accurate classification, the agent earns rewards, whereas inaccuracies lead to penalties, with greater penalties/rewards applied to the less prevalent class. This system enables the agent to develop an optimal strategy guided by specific reward functions and a well-defined learning environment. To enhance hyperparameter optimization, the proposed approach improves the differential evolution (DE) algorithm using a clustering-guided mutation strategy based on k-means clustering. This strategy clusters the candidate hyperparameter population using k-means, selects the cluster with the lowest average objective value, and uses the best candidate within that cluster to guide the mutation process. A unique method revitalizes the candidate solutions throughout the population, advancing the hyperparameter adjustment process. Exhaustive evaluations on the TrashNet and Trash datasets demonstrate the efficacy of our model, achieving a high classification accuracy of 89.908% on TrashNet and 87.438% on Trash. These findings highlight the system's capacity to properly handle particular difficulties in solid waste classification, especially data imbalance and hyperparameter sensitivity.

References

[1]	Zhang H, Cao H, Zhou Y, et al. (2023) Hybrid deep learning model for accurate classification of solid waste in the society. Urban Clim 49: 101485.https://doi.org/10.1016/j.uclim.2023.101485 doi: 10.1016/j.uclim.2023.101485
[2]	Li N, Chen Y (2023) Municipal solid waste classification and real-time detection using deep learning methods. Urban Clim 49: 101462.https://doi.org/10.1016/j.uclim.2023.101462 doi: 10.1016/j.uclim.2023.101462
[3]	Chu X, Chu Z, Wang X, et al. (2023) Comparative analysis on the performances of implementing compulsory and advocative policies in municipal solid waste classification. Environ Impact Assess Rev 99: 106982.https://doi.org/10.1016/j.eiar.2022.106982 doi: 10.1016/j.eiar.2022.106982
[4]	Wu TW, Zhang H, Peng W, et al. (2023) Applications of convolutional neural networks for intelligent waste identification and recycling: A review. Resour Conserv Recycl 190: 106813.https://doi.org/10.1016/j.resconrec.2022.106813 doi: 10.1016/j.resconrec.2022.106813
[5]	Kosolwattana T, Liu C, Hu R, et al. (2023) A self-inspected adaptive SMOTE algorithm (SASMOTE) for highly imbalanced data classification in healthcare. BioData Min 16: 15.https://doi.org/10.1186/s13040-023-00332-2 doi: 10.1186/s13040-023-00332-2
[6]	Wang Z, Liu Q (2023) Imbalanced Data Classification Method Based on LSSASMOTE. IEEE Access 11: 32252–32260.https://doi.org/10.1109/ACCESS.2023.3262657 doi: 10.1109/ACCESS.2023.3262657
[7]	Yang J, El-Bouri R, O'Donoghue O, et al. (2023) Deep reinforcement learning for multi-class imbalanced training: applications in healthcare. Mach Learnhttps://doi.org/10.1007/s10994-023-06403-1
[8]	Danaei S, et al. (2022) Myocarditis diagnosis: a method using mutual learning-based abc and reinforcement learning, In: 2022 IEEE 22nd International Symposium on Computational Intelligence and Informatics and 8th IEEE International Conference on Recent Achievements in Mechatronics, Automation, Computer Science and Robotics (CINTI-MACRo), New York: IEEE, 000265-000270.https://doi.org/10.1109/CINTI-MACRo57952.2022.10029424
[9]	Ali YA, Awwad EM, Al-Razgan M, et al. (2023) Hyperparameter search for machine learning algorithms for optimizing the computational complexity. Processes 11: 349.https://doi.org/10.3390/pr11020349 doi: 10.3390/pr11020349
[10]	Franchini G (2024) GreenNAS: A Green Approach to the Hyperparameters Tuning in Deep Learning. Mathematics 12: 850.https://doi.org/10.3390/math12060850 doi: 10.3390/math12060850
[11]	Jiang J, Wen Z, Mansoor A, et al. (2024) Efficient Hyperparameter Optimization with Adaptive Fidelity Identification, In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 26181–26190.https://doi.org/10.1109/CVPR52733.2024.02476
[12]	Moravvej SV, Mousavirad SJ, Oliva D, et al. (2022) An improved de algorithm to optimise the learning process of a bert-based plagiarism detection model, In: 2022 IEEE Congress on Evolutionary Computation (CEC), New York: IEEE, 1–7.https://doi.org/10.1109/CEC55065.2022.9870233
[13]	Solano Meza JK, Orjuela Yepes D, Rodrigo-Ilarri J, et al. (2023) Comparative analysis of the implementation of support vector machines and long short-term memory artificial neural networks in municipal solid waste management models in megacities. Int J Environ Res Public Health 20: 4256.https://doi.org/10.3390/ijerph20054256 doi: 10.3390/ijerph20054256
[14]	Abu-Qdais H, Shatnawi N, AL-Alamie EA (2024) Intelligent system for solid waste classification using combination of image processing and machine learning models. J Exp Theor Artif Intell In press. https://doi.org/10.1080/0952813X.2024.2319241
[15]	Tao J, Gu Y, Hao X, et al. (2023) Combination of hyperspectral imaging and machine learning models for fast characterization and classification of municipal solid waste. Resour Conserv Recycl 188: 106731.https://doi.org/10.1016/j.resconrec.2022.106731 doi: 10.1016/j.resconrec.2022.106731
[16]	Adeleke O, Akinlabi S, Jen TC, et al. (2023) A machine learning approach for investigating the impact of seasonal variation on physical composition of municipal solid waste. J Reliable Intell Environ 9: 99–118.https://doi.org/10.1007/s40860-021-00164-8 doi: 10.1007/s40860-021-00164-8
[17]	Fatovatikhah F, Ahmedy I, Noor RM (2024) Waste Prediction Approach Using Hybrid Long Short-Term Memory with Support Vector Machine. Int J Comput Intell Syst 17: 103.https://doi.org/10.1007/s44196-024-00481-9 doi: 10.1007/s44196-024-00481-9
[18]	Shah KB, Visalakshi S, Panigrahi R (2023) Seven class solid waste management-hybrid features based deep neural network. Int J Syst Des Comput 1: 1–10.https://doi.org/10.55529/ijsdc.11.1.10 doi: 10.55529/ijsdc.11.1.10
[19]	Chen C, Liang R, Song M, et al. (2024) Noise-assisted data enhancement promoting image classification of municipal solid waste. Resour Conserv Recycl 209: 107790.https://doi.org/10.1016/j.resconrec.2024.107790 doi: 10.1016/j.resconrec.2024.107790
[20]	Jayaraman V, Lakshminarayanan AR (2024) MSW-Net: A hierarchical stacking Model for automated municipal solid waste classification. J Air Waste Manage Assoc.https://doi.org/10.1080/10962247.2024.2367123
[21]	Niu B, Feng Q, Yang J, et al. (2023) Solid waste mapping based on very high-resolution remote sensing imagery and a novel deep learning approach. Geocarto Int 38: 2164361.https://doi.org/10.1080/10106049.2022.2164361 doi: 10.1080/10106049.2022.2164361
[22]	Murugan PR, Muneeswaran V, Vardhan NVS, et al. (2024) Deep Learning-based Hybrid Image Classification Model for Solid Waste Management, In: 2024 International Conference on Cognitive Robotics and Intelligent Systems (ICC-ROBINS), New York: IEEE, 168–172.https://doi.org/10.1109/ICC-ROBINS61647.2024.10557452
[23]	Prakash TS, Patnayakuni SP, Shibu S (2023) Municipal solid waste prediction using tree hierarchical deep convolutional neural network optimized with balancing composite motion optimization algorithm. J Exp Theor Artif Intell.https://doi.org/10.1080/0952813X.2023.2254924
[24]	Yudhana A, Fahmi M (2023) Improving Waste Classification Using Convolutional Neural Networks: An Application of Machine Learning for Effective Environmental Management. Rev D'Intelligence Artif 37: In press.https://doi.org/10.18280/ria.370415
[25]	Lilhore UK, Simaiya S, Dalal S, et al. (2024) A smart waste classification model using hybrid CNN-LSTM with transfer learning for sustainable environment. Multimedia Tools Appl 83: 29505–29529.https://doi.org/10.1007/s11042-023-16518-2 doi: 10.1007/s11042-023-16518-2
[26]	Karl F, Pielok T, Moosbauer J, et al. (2023) Multi-objective hyperparameter optimization in machine learning—An overview. ACM Trans Evol Learn Optim 3: 1–50.https://doi.org/10.1145/3611387 doi: 10.1145/3611387
[27]	Bean JC (1994) Genetic algorithms and random keys for sequencing and optimization. ORSA J Comput 6: 154–160.https://doi.org/10.1287/ijoc.6.2.154 doi: 10.1287/ijoc.6.2.154
[28]	Mousavirad SJ, Schaefer G, Korovin I, et al. (2021) An enhanced differential evolution algorithm using a novel clustering-based mutation operator, In: 2021 IEEE international conference on systems, man, and cybernetics (SMC), New York: IEEE, 176–181.https://doi.org/10.1109/SMC52423.2021.9658739
[29]	Deb K (2005) A population-based algorithm-generator for real-parameter optimization. Soft Comput 9: 236–253.https://doi.org/10.1007/s00500-004-0371-x doi: 10.1007/s00500-004-0371-x
[30]	Yang M, Thung G (2016) Classification of trash for recyclability status. CS229 Project Report 2016: 3.https://cs229.stanford.edu/proj2016/report/ThungYang-ClassificationOfTrashForRecyclabilityStatus-report.pdf
[31]	Ochecolo DI, Trash dataset. Kaggle, 2024. Available from: https://www.kaggle.com/dimonisochecolo/trash-dataset.
[32]	Mousavirad SJ, Schaefer G, Ebrahimpour-Komleh H (2021) The human mental search algorithm for solving optimisation problems, In: Enabling AI Applications in Data Science, Cham: Springer, 27–47.https://doi.org/10.1007/978-3-030-51552-2_3
[33]	Tubishat M, Ja'afar S, Alswaitti M, et al. (2021) Dynamic salp swarm algorithm for feature selection. Expert Syst Appl 164: 113873.https://doi.org/10.1016/j.eswa.2020.113873 doi: 10.1016/j.eswa.2020.113873
[34]	Yang XS, Deb S (2009) Cuckoo search via Lévy flights, In: 2009 World congress on nature & biologically inspired computing (NaBIC), New York: IEEE, 210–214.https://doi.org/10.1109/NABIC.2009.5393690
[35]	Yang XS (2010) Firefly algorithm, stochastic test functions and design optimisation. Int J Bio-Inspired Comput 2: 78–84.https://doi.org/10.1504/IJBIC.2010.032124 doi: 10.1504/IJBIC.2010.032124
[36]	Yildizdan G, Baykan ÖK (2020) A novel modified bat algorithm hybridizing by differential evolution algorithm. Expert Syst Appl 141: 112949.https://doi.org/10.1016/j.eswa.2019.112949 doi: 10.1016/j.eswa.2019.112949
[37]	Vakilian S, Moravvej SV, Fanian A (2021) Using the artificial bee colony (ABC) algorithm in collaboration with the fog nodes in the Internet of Things three-layer architecture, In: 2021 29th Iranian Conference on Electrical Engineering (ICEE), New York: IEEE, 509–513.https://doi.org/10.1109/ICEE52713.2021.9528250

Reader Comments

Your name:*

Email:*
© 2026 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)