Artificial Intelligence and MCDA in Circular Economy: Governance Strategies and Optimization for Reverse Supply Chains of Solid Waste

Abstract

The integration of multi-criteria decision analysis (MCDA) and Artificial Intelligence (AI) is revolutionizing the governance of reverse supply chains for solid waste (RSCSW) within a circular economy framework. However, the existing literature lacks a systematic assessment of the effectiveness of these methods compared to traditional waste management practices. This study conducts a systematic literature review (SLR), following PRISMA guidelines and the P.I.C.O. framework, to investigate how MCDA and AI can optimize governance, operational efficiency, and the sustainability of RSCSW. After collecting 1139 articles, 22 were selected and used for analysis. The results indicate that hybrid MCDA-AI models, employing techniques, such as TOPSIS, AHP, neural networks, and genetic algorithms, enhance decision-making automation, reduce costs, and improve waste traceability. Nevertheless, regulatory barriers and technological challenges still hinder large-scale adoption. This study proposes an innovative framework to address these gaps and drive evidence-based public policies. The findings provide guidelines for policymakers and managers, contributing to the Sustainable Development Goals (SDGs) agenda and advancements in circular economy governance.

1. Introduction

The 21st century has brought unprecedented challenges to global industries, including rapid societal transformations driven by technological innovation, social changes, and evolving business models. However, this progress presents a critical paradox: the growing scarcity and depletion of natural resources. Among the most pressing issues is solid waste management, which intersects with the environmental, social, and economic dimensions on a global scale. While traditional supply chains are well studied and widely optimized, reverse supply chains for solid waste (RSCSW) remain fragmented and underdeveloped in governance. This condition results in significant gaps in waste recovery and resource reintegration, along with numerous logistical bottlenecks and trade-offs.

Several industries, such as automotive, petroleum, agriculture, and fisheries, generate waste streams with distinct challenges and opportunities. The automotive sector, for instance, faces difficulties in managing a variety of complex materials, including heavy metals, plastics, and rubber, while grappling with new trade-offs brought by electric vehicles (EVs), particularly in the recycling of lithium-ion batteries [1]. The petroleum industry, on the other hand, deals with hazardous by-products such as sludge, contaminated water, and used oils, which pose significant environmental risks but can also be transformed into secondary raw materials like synthetic fuels or asphalt modifiers [2]. In parallel, agricultural waste offers opportunities for the production of biogas, biofertilizers, and green hydrogen via syngas-based technologies, while fisheries waste can be repurposed into animal feed or fishmeal [3]. Despite advancements in recovery technologies, these industries continue to experience critical logistical challenges, from inadequate infrastructure to inefficient collection and sorting systems [4]. Addressing these gaps requires strengthened RSCSW governance, technological innovation, and integrated strategies that align with the principles of the circular economy, enabling the reintegration of waste into production cycles and reducing dependency on virgin resources.

The generation of solid waste represents one of the contemporary world’s most complex environmental, social, and economic challenges. Population growth, excessive consumption, and inefficient material management have intensified waste production, negatively affecting the quality of life, biodiversity, and climate [5,6,7,8,9]. The Food and Agriculture Organization (FAO) estimates that only 9% of aquaculture waste is properly managed, increasing the environmental footprint of the fishing sector [10]. This thinking is aligned with global decarbonization targets, the Sustainable Development Goals (SDGs) [11]. Studies by Wilson et al. [12] and Ghisellini, Cialani, and Ulgiati [13] also highlight specific challenges in waste management in sectors such as healthcare, construction, and the automotive industry. In particular, the construction industry represents one of the most ecologically impactful sectors, due to its high demand for raw materials and massive generation of demolition waste. Materials such as concrete, steel, and ceramics can be reintegrated into new production cycles through circular practices. The use of Artificial Intelligence (AI) techniques, such as object recognition, material classification algorithms, and predictive models, can support waste triage, quality assurance, and logistics planning in construction waste recycling chains. In this context, the circular economy emerges as an essential paradigm to address resource scarcity and ineffective waste management. Unlike the linear model of “take, make, and dispose,” the circular economy seeks to keep resources in use for as long as possible, reintegrating waste into the production cycle as secondary raw materials [14,15,16,17]. The circular economy has been identified as a crucial approach to mitigating the negative impacts of solid waste generation globally, promoting resource optimization and reducing dependence on the extraction of new raw materials [15]. Governments and private sectors have been implementing circular economy-based strategies, such as recycling incentives, more sustainable production models, and the creation of more resilient supply chains [14]. However, significant challenges remain in the practical implementation of these initiatives, requiring new technological and managerial approaches to ensure their effectiveness. This model demands the optimization of solid waste supply chains (RSCSW), which play a fundamental role in the recovery of discarded materials. Nevertheless, RSCSW management is complex and faces significant challenges, including logistical disruptions, fragmented legislation, unforeseen events, failures in waste digitization and tracking, and cyberattacks [17,18,19,20,21,22].

The absence of robust governance practices negatively impacts the efficiency and benefits of reverse supply chains for solid waste (RSCSW). Governance involves regulatory structures, coordination mechanisms, and incentives for different stakeholders to operate effectively. According to Coase [23] and Williamson [24], the efficient coordination of economic activities is directly related to the ability to minimize transaction costs and optimize decision making. Furthermore, Jessop [25], Meckling and Jensen [26], and Yang and Lu [27] emphasize that effective governance depends on control and transparency mechanisms to ensure collaboration among stakeholders. Additionally, the G20/OECD Principles of Corporate Governance [28] provide a comprehensive framework to strengthen governance structures, highlighting the importance of transparency, accountability, and stakeholder participation in promoting sustainable and resilient systems. Moreover, effective governance in reverse supply chains and solid waste management is crucial to ensuring the efficiency, resilience, and sustainability of these systems [29,30,31,32]. The complexity of these supply chains, which involve the allocation and commitment of financial resources, demands a systematic and evidence-based approach to making informed decisions [33]. In this context, the use of decision support tools, such as multi-criteria decision analysis (MCDA) and, more recently, Artificial Intelligence (AI) methods, can mitigate the risk of decisions that do not reflect maximum efficiency and resilience [34]. Considering these aspects of the need for robust governance as a form of resilience, managerial alternatives, such as the use of decision support models (DSMs), have been increasingly employed to optimize RSCSW management. Among these models, multi-criteria decision analysis (MCDA), Analytic Hierarchy Process (AHP), Goal Programming (GP), and Fuzzy Logic stand out, enabling the evaluation of trade-offs between multiple criteria and balancing economic, environmental, and social aspects [18,19,20,21]. In another analytical dimension, the advancement of Artificial Intelligence (AI) has revolutionized logistics and supply chain management, introducing advanced forecasting, optimization, and automation techniques, such as machine learning, neural networks, and genetic algorithms [35,36,37,38]. Thus, this study aims to (1) identify integrated optimization strategies using MCDA and AI to improve governance in reverse solid waste supply chains (RSCSW); (2) evaluate how these strategies contribute to operational efficiency, cost reduction, and material recovery across different sectors (e.g., construction, agriculture, urban management); (3) propose a structured decision support framework to guide the implementation of AI-MCDA integration in RSCSW governance; (4) analyze governance, regulatory, and technological barriers, offering recommendations to support evidence-based public policies and sustainable logistics practices.

2. Materials and Methods

The increasing complexity of reverse supply chains for solid waste (RSCSW) demands innovative solutions to optimize material recovery and promote the circular economy. Although multi-criteria decision analysis (MCDA) and Artificial Intelligence (AI) have shown potential to enhance RSCSW management, a knowledge gap remains regarding the effective integration of these approaches. This systematic review aims to bridge this gap by synthesizing the existing literature on the integrated application of MCDA and AI in RSCSW optimization while identifying best practices and challenges for implementing sustainable solutions. To address this knowledge gap, a systematic literature review (SLR) was conducted, following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) methodology [39]. Although PRISMA originated in clinical health research for identifying evidence, its application has expanded significantly in recent years, including fields such as management and engineering, where it has proven effective in synthesizing evidence and ensuring methodological rigor [40,41]. The PRISMA guidelines [39] were employed to ensure transparency and reproducibility in the review process. Additionally, the P.I.C.O. (Population, Intervention, Comparison, Outcome) framework was used to formulate the research question and structure the review, further enhancing methodological rigor [42].

Table 1. Key studies on MCDA and AI in RSCSW governance.

Study	Year	Main Findings
Kirchherr et al. [14]	2017	Identifies barriers to the circular economy but does not propose advanced technological solutions.
Geissdoerfer et al. [15]	2017	Explores circular economy models without analyzing MCDA and AI in RSCSW governance.
Govindan et al. [18]	2015	MCDA improves efficiency in reverse logistics but lacks integration with AI.
Tseng et al. [19]	2020	Highlights technological innovations without an integrated approach to MCDA and AI.
Culot et al. [35]	2020	Presents AI applications in logistics without a direct connection to RSCSW.
Nasiri and Huang [36]	2011	Applies fuzzy logic to waste collection routing without a multi-criteria decision structure.
Mardani et al. [37]	2015	Comprehensive review of MCDA techniques, providing a robust theoretical basis for integration with AI.
LeCun et al. [38]	2015	Introduces deep learning, offering insights into advanced AI applications in RSCSW governance.

summarizes the main studies on MCDA and AI in RSCSW management, highlighting advances, gaps, and the current state of research in this area.

These studies highlight significant advances but also reveal important gaps. Despite the growing interest in circular economy practices, there is a significant gap in the literature regarding the integration of MCDA and AI in governance models for RSCSW. This review seeks to address this gap by systematically analyzing existing studies and proposing a framework for future research.

Thus, the current state of research on governance for reverse solid waste supply chains (RSCSW), particularly those optimized through the integration of multi-criteria decision analysis (MCDA) and Artificial Intelligence (AI), remains underexplored. By identifying and synthesizing evidence on the integration of MCDA and AI-based optimization tools, emphasizing decision making through circular economy governance practices for solid waste management, this study seeks to examine the dynamics of RSCSW in a circular economy context. This examination aims to bring clarity to managerial aspects, sustainability policies, and the key stakeholders involved in the decision-making process.

To ensure methodological rigor and reproducibility, this study followed the guidelines of PRISMA, as established by Page et al. [39]. PRISMA is a widely adopted protocol for systematic reviews, ensuring transparency in the selection, eligibility, and data extraction processes. The review process involved four fundamental stages of PRISMA: identification—a structured search in databases (Scopus, Web of Science, IEEE Xplore, and Capes), from August 2024 to January 2025, along with the use of the Covidence application [43]; screening—the exclusion of duplicate articles and initial analysis of titles and abstracts; eligibility—a thorough evaluation of full texts based on inclusion/exclusion criteria; inclusion—the final selection of articles that meet the methodological and theoretical requirements of the study.

Additionally, the P.I.C.O. model was used to structure the research question. Thus, the population is reverse solid waste supply chains; the interventions are integrated application of MCDA and IA in RSCSW governance; the comparison is traditional waste management methods with this one; the outcome is the improvement in operational efficiency, traceability, and governance. This approach ensures that the review follows a rigorous structure, enabling objective comparisons between studies and reliable synthesis of findings and ensuring a focused and comprehensive answer to the central research question:

“How can the integration of Multi-Criteria Decision Analysis (MCDA) and Artificial Intelligence (AI) optimize reverse supply chains within the circular economy of solid waste, adhering to governance practices, principles, or mechanisms, in terms of operational efficiency, cost reduction, and material recovery, when compared to traditional waste management practices?”

2.1. Eligibility and Exclusion Criteria

Regarding the eligibility criteria, the publications were included if they met the following criteria:

-

Publication Period: published between 2015 and 2024. This timeframe was selected to capture recent advancements in AI and MCDA applications within solid waste management, driven by evolving regulations and the growing emphasis on the circular economy. It also reflects the maturation of AI technologies, such as deep learning and genetic algorithms, making them more accessible for complex waste management challenges.

-

Context: studies addressing circular economy and solid waste management with robust optimization methods for decision making and well-documented results.

-

Governance: studies focusing on governance practices, policies, principles, or mechanisms through the coordination and control of reverse solid waste supply chains within circular economy frameworks.

-

Study Type: empirical research articles and case studies published in peer-reviewed journals.

-

Methodology: studies employing MCDA, AI, or the integrated application of MCDA and AI for the optimization of solid waste management in reverse supply chains.

-

Relevance: studies demonstrating clear relevance to the identification of evidence concerning the research question.

In terms of the exclusion criteria, the articles were excluded if they met any of the following criteria:

-

Publication Date: studies published before 2015;

-

Publication Type: non-peer-reviewed articles or publications in journals with low impact;

-

Scope: systematic literature reviews with a similar scope to the current review;

-

Methodology: studies not involving MCDA or AI applications in solid waste management or studies with poorly described or weak methodologies;

-

Context: studies not applicable to the circular economy or solid waste management;

-

Focus: studies lacking a specific focus on solid waste optimization in reverse supply chains.

2.2. Search Strategy

The search strategy was developed using Boolean operators to ensure comprehensive coverage of the relevant literature. Key concepts were combined to create a robust search string, which included terms related to reverse supply chains, solid waste, circular economy, governance, multi-criteria decision analysis (MCDA), and Artificial Intelligence (AI). The combined search string used was as follows:

(“urban solid waste” OR “municipal waste” OR “industrial waste” OR “agricultural waste”) AND (“circular economy” OR “circularity” OR “closed loop” OR “recycling” OR “upcycling” OR “valorisation”) AND (“multi-criteria decision analysis” OR “MCDA” OR “decision-making models” OR “optimization models”) AND (“artificial intelligence” OR “machine learning” OR “AI integration” OR “predictive analytics”) AND (“reverse supply chain governance” OR “reverse logistics” OR “stakeholder collaboration”) AND (“governance models” OR “decision-making governance” OR “management frameworks”) AND (“biotechnology” OR “green chemistry” OR “waste-to-energy technologies” OR “recycling technologies”) AND (“public policies” OR “private sector intervention” OR “community-based waste management” OR “socio-economic impacts” OR “sustainability policies”).

These terms were chosen based on their relevance to the research topic of integrating MCDA and AI into models of governance for solid waste in reverse supply chains. They were combined using Boolean operators such as “AND” and “OR” to refine the search and ensure precision while maintaining broad coverage of related studies.

The scope was designed to capture a comprehensive view of the literature on MCDA and AI, the circular economy, technological innovations, governance practices, mechanisms, and models aligning with sustainable practices for global solid waste management.

2.3. Screening and Selection of the Studies

The initial selection of articles was based on reading the titles and abstracts, followed by reading the full text of the selected articles. Two independent reviewers conducted the title/abstract and full-text screening using Covidence [43]. Studies without unanimous agreement between reviewers were excluded. Disagreements were resolved by a third reviewer to ensure methodological rigor (PRISMA Item 8). Relevant data were extracted from these selected studies, including data on the optimization of RSCSW derived from solid waste, their applications in the circular economy through MCDA and AI, and the associated environmental and socio-economic impacts. The following outlines the step-by-step procedure: Process of selection: Two reviewers conducted the screening process. In cases of disagreement, a third reviewer was consulted to reach a consensus. This approach ensured objectivity and minimized bias in the selection process. Data collection process: Data extraction was performed independently by the two reviewers using a standardized form. To ensure accuracy, the reviewers contacted the authors of the studies whenever additional information or clarifications were needed. The extracted data included information on the study context, methodology, results, and conclusions. Assessment of research study quality: The quality of the included studies was assessed using the Critical Appraisal Skills Programme (CASP) [44] checklist. Two independent reviewers evaluated each study, focusing on criteria, such as clarity of the research question, appropriateness of study design, population selection, exposure and outcome measurement, control of confounders, and statistical analysis. Discrepancies between reviewers were resolved through discussion. Preparation of data for synthesis: The data were prepared for synthesis through the normalization of metrics and the identification of key themes. Heterogeneity among the studies was assessed qualitatively, considering differences in methodologies and contexts. Although a quantitative meta-analysis was not performed due to the heterogeneity of the studies, a narrative synthesis was conducted. Formal sensitivity analyses (e.g., subgroup analyses or exclusion of lower-quality studies) were not performed due to the small final sample size (n = 22), but the unanimous reviewer consensus and uniformly high quality of included studies (CASP 10/10) [44] ensure the robustness of findings (PRISMA Item 13f). Risk of bias assessment: Reviewers assessed the risk of bias in the included studies using the CASP checklist [44]. The evaluation focused on potential sources of bias, such as study design, data collection methods, and reporting of results. Discrepancies between reviewers were resolved through discussion. For the synthesis of results, a qualitative analysis was conducted, which included the normalization of metrics and the identification of key themes.

The heterogeneity among the studies was assessed qualitatively, and the findings were integrated into a coherent narrative.

Table 2. Quality assessment results of included studies based on the CASP [44] checklist.

Study	Authors	Total Score (CASP)	Quality Category
1	Meng et al. [45]	10/10	High
2	Kartal et al. [46]	10/10	High
3	Oyebode et al. [47]	10/10	High
4	Sengupta et al. [48]	10/10	High
5	Ma et al. [49]	10/10	High
6	Tseng et al. [19]	10/10	High
7	Al-Thani et al. [50]	10/10	High
8	Feitó-Cespón et al. [51]	10/10	High
9	Gardas et al. [52]	10/10	High
10	Oluleye et al. [53]	10/10	High
11	Hu et al. [54]	10/10	High
12	Xia et al. [55]	10/10	High
13	Darzi [56]	10/10	High
14	Hashemi [57]	10/10	High
15	Ghosh et al. [58]	10/10	High
16	Xu et al. [59]	10/10	High
17	Lv et al. [60]	10/10	High
18	Trochu et al. [61]	10/10	High
19	Huang et al. [62]	10/10	High
20	Santander et al. [63]	10/10	High
21	Gholizadeh et al. [64]	10/10	High
22	Chaube et al. [65]	10/10	High

presents a summary of the quality assessment of the included studies, conducted using the CASP [44]. All studies were classified as high quality as they met the criteria for clarity of research question, appropriateness of study design, population selection, exposure and outcome measurement, control of confounders, statistical analysis, presentation of results, discussion, and conclusion. To mitigate potential publication bias, comprehensive searches across multiple databases were conducted and included relevant gray literature, such as technical reports and theses, where applicable. Funnel plot analysis and Egger’s test were considered to assess the presence of publication bias but were not applied due to the heterogeneity of the included studies. A detailed CASP [44] assessment, including individual scores for each criterion, is available in the Supplementary Material (Table S1).

3. Results

Of the 1139 studies that emerged from the search process, 649 duplicates were excluded, leaving 490 for evaluation. Of these, 319 were excluded, 171 studies were assessed for eligibility, and 149 studies were excluded for various reasons: published before 2015, lack of framing, not peer reviewed, mathematical models and stochastic models that do not specifically address reverse supply chains of solid waste (RSCSW) within the circular economy, studies about COVID-19, etc. The PRISMA flowchart is presented in Figure 1 to illustrate this process.

After a thorough analysis of the entire text, 22 studies were included in the systematic review. The studies included in this review are summarized in

Table 3. Studies related by eligibility criteria—main characteristics of the included studies.

Ref.	Author(s)	Year	Journal	Study Title	Methodology	Key Contributions
[45]	Meng, Z. et al.	2024	Applied Soft Computing	Preference learning based on adaptive graph neural network for multi-criteria decision support.	Graph neural networks (GNN) and TOPSIS.	Adaptive method with high accuracy and interpretability.
[46]	Kartal, H. et al.	2016	Computers & Industrial Engineering	An integrated decision analytic framework of machine learning with multi-criteria decision making.	Combination of machine learning (SVM, ANN) and MCDM (AHP, VIKOR).	High accuracy in inventory classification with imbalanced data.
[47]	Oyebode, O.J. et al.	2023	Civil and Environmental Engineering	Optimization of Supply Chain Network in Solid Waste Management Using Genetic Algorithm and Fuzzy Logic.	Genetic algorithm and fuzzy logic.	Flexible model for waste management in Lagos, Nigeria.
[48]	Sengupta, D. et al.	2023	Environmental Science and Pollution Research	A sustainable green reverse logistics plan for plastic solid waste management using the TOPSIS method.	TOPSIS and mathematical modeling.	Reduction of costs and carbon emissions.
[49]	Ma, G. et al.	2024	Journal of Environmental Management	Empirical and simulated investigation of the solid waste reverse supply chain.	Agent-based modeling and regression analysis.	Identification of factors influencing waste sorting behavior.
[50]	Al-Thani, N.A. et al.	2022	Heliyon	Integrated TOPSIS-COV approach for selecting a sustainable PET waste management technology.	TOPSIS combined with a coefficient of variation (COV).	Closed-loop recycling as the optimal solution.
[51]	Feitó-Cespón, M. et al.	2020	Expert Systems with Applications	A Fuzzy Inference-based Scenario Building in Two-Stage Optimization Framework for Sustainable Recycling Supply Chain Redesign.	Fuzzy Inference System and epsilon-constraint method.	Robust solutions for redesigning sustainable reverse supply chains under uncertainty.
[52]	Gardas, R. et al.	2024	Decision Analytics Journal	An analysis of critical factors for adopting machine learning in manufacturing supply chains.	DEMATEL and qualitative analysis.	“Technology integration” and “Forecasting” as essential factors.
[53]	Oluleye, B.I. et al.	2023	Sustainable Production and Consumption	Adopting Artificial Intelligence for enhancing the implementation of systemic circularity.	Systematic review and SWOT analysis.	Framework for AI application throughout the building product lifecycle.
[54]	Hu, Y. et al.	2023	Journal of Cleaner Production	Waste tire valorization: Advanced technologies, process simulation, and system optimization.	Review of advanced technologies and sustainability assessment.	Proposal for a unified sustainability assessment system.
[55]	Xia, H. et al.	2024	Waste Management	Uncertain programming model for designing multi-objective reverse logistics networks.	Multi-objective programming with uncertainties and NSGA-III algorithm.	Robust model to minimize costs and environmental impacts.
[56]	Darzi, M.A.	2024	Journal of Environmental Management	Evaluating e-waste mitigation strategies based on Industry 5.0 enablers.	BWM (Best-Worst Method) and F-VIKOR.	“Take-back practices” as a key strategy for sustainable e-waste management.
[57]	Hashemi, S.E.	2020	Expert Systems with Applications	A fuzzy multi-objective optimization model for sustainable reverse logistics network design.	Fuzzy multi-objective programming and genetic algorithms.	Bee colony algorithm outperformed NSGA-II in solution exploration.
[58]	Ghosh, S. et al.	2023	Applied Soft Computing	Closed-loop multi-objective waste management through vehicle routing in a neutrosophic fuzzy environment.	Neutrosophic fuzzy logic and TOPSIS.	Pareto-optimal solutions for emission reduction and profit maximization.
[59]	Xu, Z. et al.	2017	Waste Management	Global reverse supply chain design for solid waste recycling under uncertainties.	Mixed-integer linear programming with robust optimization.	Robust model to handle uncertainties and carbon emission constraints.
[60]	Lv, J.Y. et al.	2020	Waste Management	Optimization of recyclable MSW recycling network: A Chinese case of Shanghai.	P-median optimization model.	Reduction of operational costs in municipal solid waste management.
[61]	Trochu, J. et al.	2018	Resources, Conservation and Recycling	Reverse logistics network redesign under uncertainty for wood waste in the CRD industry.	Mixed-integer linear programming (MILP) and scenario analysis.	Cost reduction and increased efficiency in wood recycling.
[62]	Huang, L. et al.	2022	Journal of Cleaner Production	Waste material recycling and exchanging decisions for industrial symbiosis network optimization.	Mixed-integer programming and epsilon-constraint method.	Balance between economic, environmental, and social objectives.
[63]	Santander, P. et al.	2020	Resources, Conservation and Recycling	Closed loop supply chain network for local and distributed plastic recycling for 3D printing.	Mixed-integer linear programming (MILP).	Economic and environmental benefits of distributed plastic recycling.
[64]	Gholizadeh, H. et al.	2022	Computers & Industrial Engineering	Modelling uncertainty in sustainable-green integrated reverse logistics network using metaheuristics.	Metaheuristics and fuzzy programming.	Efficient solutions for sustainable reverse logistics networks.
[65]	Chaube, S. et al.	2024	International Journal of Mathematical, Engineering and Management Sciences	An Overview of Multi-Criteria Decision Analysis and the Applications of AHP and TOPSIS Methods.	Review of MCDA methods (AHP and TOPSIS).	Detailed comparison between AHP and TOPSIS for multi-criteria decision-making.
[66]	Vanany et al.	2024	Asia University	A mathematical model for fish reverse supply chain considering social-economic and environmental impacts.	Mixed-integer linear programming (MILP).	Increased profit and reduced carbon emissions.

, which presents the main characteristics of the studies (year, journal, title, methodology, and main contributions).

The inclusion of the 22 articles strictly followed the PRISMA guidelines [39], ensuring transparency in the selection and analysis of the studies. To facilitate an understanding of the studies’ adherence to the P.I.C.O. criteria,

Table 4. Adherence to PRISMA and P.I.C.O. criteria for the 22 included studies.

Ref	Author(s)	Population	Intervention	Comparison	Outcome
[45]	Meng, Z. et al.	Reverse supply chain of plastics	GNN + TOPSIS	Heuristic	5.78% improvement in ranking metrics
[46]	Kartal, H. et al.	Inventory classification	SVM, ANN + AHP, VIKOR	Traditional	High accuracy in imbalanced data classification
[47]	Oyebode, O.J. et al.	Solid waste management in Lagos	Genetic algorithm + fuzzy logic	Traditional	Flexible model for waste management
[48]	Sengupta, D. et al.	Plastic solid waste management	TOPSIS + mathematical modeling	Traditional	Reduction of costs and carbon emissions
[49]	Ma, G. et al.	Solid waste reverse supply chain	Agent-based modeling + regression	Traditional	Identification of factors influencing waste sorting behavior
[50]	Al-Thani, N.A. et al.	PET waste management	TOPSIS + coefficient of variation (COV)	Traditional	Closed-loop recycling as the optimal solution
[51]	Feitó-Cespón, M. et al.	Sustainable Recycling Supply Chain Redesign	Fuzzy Inference System and epsilon-constraint method	Traditional	Robust solutions for redesigning sustainable RSC under uncertainty
[52]	Gardas, R. et al.	Manufacturing supply chains	DEMATEL + qualitative analysis	Traditional	Identification of critical factors for AI adoption
[53]	Oluleye, B.I. et al.	Building product lifecycle	Systematic review + SWOT analysis	Traditional	Framework for AI application in circularity
[54]	Hu, Y. et al.	Waste tire valorization	Advanced technologies + simulation	Traditional	Proposal for a unified sustainability assessment system
[52]	Xia, H. et al.	Multi-objective reverse logistics	Multi-obj programming + NSGA-III	Traditional	Robust model to minimize costs and environmental impacts
[56]	Darzi, M.A.	E-waste mitigation strategies	BWM + F-VIKOR	Traditional	“Take-back practices” as a key strategy for sustainable e-waste management
[57]	Hashemi, S.E.	Sustainable reverse logistics	Fuzzy multi-objective programming + genetic algorithms	Traditional	Bee colony algorithm outperformed NSGA-II in solution exploration
[58]	Ghosh, S. et al.	Vehicle routing in waste management	Neutrosophic fuzzy logic + TOPSIS	Traditional	Pareto-optimal solutions for emission reduction and profit maximization
[59]	Xu, Z. et al.	Global reverse supply chain	Mixed-int linear programming (MILP) + robust optimization	Traditional	Robust model to handle uncertainties and carbon emission constraints
[60]	Lv, J.Y. et al.	Municipal solid waste recycling	P-median optimization model	Traditional	Reduction of operational costs in Shanghai
[61]	Trochu, J. et al.	Wood waste reverse logistics	MILP + scenario analysis	Traditional	Cost reduction and increased efficiency in wood recycling
[62]	Huang, L. et al.	Industrial symbiosis network	MIP + epsilon-constraint method	Traditional	Balance between economic, environmental, and social objectives
[63]	Santander, P. et al.	Plastic recycling for 3D printing	MILP	Traditional	Economic and environmental benefits of distributed plastic recycling
[64]	Gholizadeh, H. et al.	Sustainable reverse logistics	Metaheuristics + fuzzy programming	Traditional	Efficient solutions for sustainable reverse logistics networks
[65]	Chaube, S. et al.	Multi-criteria decision analysis	Review of AHP and TOPSIS	Traditional	Detailed comparison between AHP and TOPSIS for multi-criteria decision-making
[66]	Vanany, I. et al.	Fish reverse supply chain	MILP	Traditional	Increased profit and reduced carbon emissions

outlines the rationale for including each study. This approach strengthens the validity of the findings, ensures methodological reproducibility, and highlights the evidence presented in each study. However, the quality of the included articles was assessed by using the CASP [44] quality appraisal tools and Covidence [43]. The results indicated that the included studies were of high quality. Closed-access studies were excluded. Additionally, the heterogeneity of the studies was mitigated through qualitative synthesis, allowing for the integration of results from different methodological designs.

4. Results and Discussion

This systematic review of the integration of multi-criteria decision analysis (MCDA) and Artificial Intelligence (AI) in reverse supply chains of solid waste (RSCSW) within the circular economy framework reveals significant advancements in operational efficiency, cost reduction, and material recovery. The findings are structured around four main themes derived from the analysis of the 22 articles reviewed:

• Integration of MCDA and AI in RSCSW Optimization—Examination of how these methodologies enhance decision making and operational efficiency.
• Operational Efficiency, Cost Reduction, and Material Recovery—Analysis of how AI-driven and MCDA-based techniques contribute to economic and environmental performance.
• Framework and Roadmap for Optimizing RSCSW—Development of a structured integration model based on evidence from the reviewed studies.
• Governance and Policy Barriers—Discussion of the regulatory, financial, and technological constraints affecting implementation.

4.1. Integration of MCDA and AI in RSCSW Optimization

The integration of MCDA and AI has emerged as a transformative approach to addressing the complexities of RSCSW. The reviewed studies demonstrate that this integration enhances decision-making processes by combining the strengths of both methodologies:

-

MCDA provides a structured framework for evaluating multiple criteria, such as economic, environmental, and social factors, enabling decision makers to balance trade-offs effectively. Techniques like TOPSIS, AHP, and BWM have been widely applied to prioritize strategies, select optimal locations for recycling centers, and evaluate the sustainability of waste treatment methods [45,49,56].

-

AI introduces advanced computational capabilities, such as machine learning, neural networks, and genetic algorithms, which enable the automation of complex decision-making processes, predictive analytics, and real-time optimization [35,38].

For example, Meng et al. [45] introduced a method based on graph neural networks (GNNs) combined with TOPSIS, achieving a 5.78% improvement in ranking metrics compared to traditional methods. Similarly, Sengupta et al. [48] used TOPSIS to select optimal strategies for plastic waste reverse logistics, demonstrating the effectiveness of MCDA in reducing costs and carbon emissions. Darzi [56] combined BWM and F-VIKOR to prioritize e-waste mitigation strategies, emphasizing the importance of “take-back practices” for sustainable e-waste management.

In terms of a critical evaluation and divergent perspectives, the following aspects can be highlighted:

-

Computational constraints: LeCun [38] highlights that deep neural networks require significant computational resources and large datasets for training, which can be a barrier in resource-constrained environments. This limitation is particularly relevant in developing countries, where digital infrastructure may be inadequate.

-

Algorithmic sensitivity: Culot et al. [35] note that genetic algorithms are highly sensitive to initialization parameters, which can lead to premature convergence or suboptimal solutions. This raises concerns about the robustness of AI-driven decision making in real-world waste management applications.

-

Lack of standardization: Unlike traditional MCDA methods with well-defined evaluation criteria, AI-driven approaches lack standardized frameworks for waste management applications [14]. This divergence highlights the need for more empirical validation and cross-sectoral benchmarking.

4.2. Operational Efficiency, Cost Reduction, and Material Recovery

The integration of MCDA and AI has led to significant improvements in operational efficiency, cost reduction, and material recovery in RSCSW:

-

Operational efficiency: The application of AI-based models has enabled the automation of decision-making processes, resulting in greater efficiency. For example, Meng et al. [45] demonstrated that the use of graph neural networks (GNNs) combined with TOPSIS improved ranking metrics by 5.78%, highlighting AI’s potential to enhance decision making in complex scenarios. Xia et al. [55] applied multi-objective programming with the NSGA-III algorithm to optimize reverse logistics networks, demonstrating the ability to robustly minimize costs and environmental impacts.

-

Cost reduction: The use of Mixed-Integer Linear Programming (MILP) and genetic algorithms has proven effective in minimizing operational costs. Lv et al. [60] proposed a p-median optimization model to determine the optimal locations for recycling centers in Shanghai, achieving a total cost reduction of CNY 144 million. Similarly, Xu et al. [59] developed a robust MILP model to handle uncertainties and carbon emission constraints, reducing costs in global reverse supply chains.

-

Material recovery: The integration of MCDA and AI has facilitated more efficient material recovery processes. Huang et al. [62] proposed an industrial symbiosis model that optimized the recycling and exchange of waste materials, achieving 90% material recovery efficiency. Santander et al. [63] explored the economic and environmental benefits of distributed plastic recycling for 3D printing, demonstrating the potential for material recovery in localized supply chains.

By performing an analytical review and contrasting points, it is possible to find the following:

-

Scalability concerns: While AI-based methods optimize logistics, Kirchherr et al. [14] argue that the circular economy transition faces scalability challenges due to infrastructure gaps and regulatory fragmentation. This suggests that technological improvements alone are insufficient without broader systemic changes.

-

Regulatory inconsistencies: Xu et al. [59] demonstrated significant cost reductions using MILP models; however, regulatory inconsistencies often delay the implementation of optimization models, as pointed out by OECD [28]. This highlights the importance of policy alignment alongside technological advancements.

-

Social and ethical implications: While AI improves decision making, it may also displace low-skilled labor in waste management sectors, raising ethical concerns about job losses and the digital divide [61].

4.3. Framework and Roadmap for Optimizing RSCSW

Drawing from General Systems Theory (GST) by von Bertalanffy [67], the proposed framework conceptualizes RSCSW as a dynamic system with interconnected inputs, processes, outputs, and feedback. This holistic approach integrates MCDA and AI with governance mechanisms, namely inputs: identification and analysis of RSCSW decision challenges; processes: application of MCDA and AI to optimize governance and operations; and outputs: quantifiable results for stakeholder evaluation.

Table 5. Integrated analytical framework for RSCSW governance, MCDA, and AI.

RSCSW Chain	Decision Strategy (MCDA + AI)	Appropriate Governance Strategy	Observed Outcomes
Plastic Waste	TOPSIS + Neural Networks (GNN) for selecting recycling sites and route optimization.	Blockchain + Consortia: Traceability of waste flow and collaboration among municipalities.	Increased material recovery efficiency [62].
Electronic Waste	BWM + F-VIKOR for prioritization of mitigation strategies (e.g., take-back practices).	Harmonized Public Policies: International regulations [28].	Reduction in carbon emissions and increased circularity [56].
Wood Waste (CRD)	Fuzzy Logic + Genetic Algorithms for optimizing reverse logistics under uncertainty.	Fiscal Incentives: Tax benefits for companies adopting wood recycling practices [61].	Improved recycling efficiency [61].
Tire Waste	NSGA-II/III: Multi-objective optimization (cost vs. environmental impact).	Public-Private Partnerships (PPP): Funding for advanced recycling technologies.	Reduction in the use of virgin raw materials [54].
Organic Waste	AHP + Machine Learning (SVM): Waste generation forecasting and composting optimization.	Education and Awareness: Community programs promoting organic waste separation.	Increased organic compost production [49].
Construction Waste	MILP (Mixed-Integer Linear Programming): Optimization of collection and recycling networks.	Local Regulation: Standards to reduce waste in construction sites [28].	Reduced operational costs [59].
Industrial Waste	Epsilon-Constraint + AI: Balancing economic, environmental, and social objectives.	Industrial Symbiosis: Exchange of waste among industries for reuse.	Increased waste reuse efficiency [62].

provides a detailed overview of optimized decision strategies across various solid waste categories, organized into an integrated analytical framework. Additionally, Figure 2 and

Table 6. Five-phase roadmap for implementing MCDA + AI in RSCSW governance.

Phase	Strategic Action	Tools/Methods	Responsible Entities	Success Indicators	References
1. Initial Diagnosis	Mapping RSCSW chains and identifying logistical and regulatory bottlenecks.	SWOT Analysis, Stakeholder Interviews.	Governments, companies, researchers.	Diagnostic reports and gap analysis.	Oluleye et al. [53]: SWOT applied to lifecycle of building products; Gardas et al. [52]: Identifies critical factors via qualitative/DEMATEL analysis.
2. Definition of Governance Model	Selecting regulatory approach and defining decision criteria.	Governance models (Coase, Williamson).	Regulatory agencies, sectoral organizations.	Policy documents and regulatory frameworks.	Oluleye et al. [53]: Proposes framework for AI governance; Gardas et al. [52]: Factors for AI adoption in manufacturing governance.
3. Implementation of MCDA + AI	Applying MCDA for strategy prioritization and AI for automation.	TOPSIS, AHP, Neural Networks, Genetic Algorithms.	AI specialists, strategic decision-makers.	Reduction in operational costs, increased efficiency.	Meng et al. [45]: GNN + TOPSIS in plastic waste chain; Kartal et al. [46]: AI + AHP/VIKOR; Hashemi [57]: MCDA + genetic algorithms.
4. Infrastructure Development	Implementing digital platforms for tracking and optimization.	Blockchain, Big Data, IoT.	Tech companies, municipalities.	Technology adoption rate and data integration levels.	Hu et al. [54]: Simulation for tire recycling; Ma et al. [49]: Agent-based modeling; Ghosh et al. [58]: AI in routing and emission control.
5. Continuous Monitoring and Adjustments	Ongoing assessment of impacts and strategy refinements.	Performance Indicators (KPIs), audits.	Academics, regulators, managers.	Continuous improvement in environmental and economic indicators.	Darzi [56]: KPIs for e-waste practices; Xu et al. [60]: Robust optimization with emissions control; Gholizadeh et al. [64]: Fuzzy + metaheuristics.

show the five-phase roadmap for the practical implementation of this framework, aligning governance tools, technologies, and actors.

4.4. Justification for Including a Framework and Roadmap in a Systematic Review

A framework in a PRISMA-based systematic review provides structured synthesis. Tranfield, Denyer and Smart [41] emphasize frameworks for conceptual synthesis in management research. Snyder [68] shows how findings can shape frameworks guiding future research. Paul and Criado [69] discuss bridging theory and practice via structured frameworks. Boell and Cecez-Kecmanovic [70] highlight frameworks’ role in knowledge accumulation and practical applicability. The integration of MCDA and AI with robust governance mechanisms enhances transparency, traceability, and operational optimization in reverse supply chains for solid waste. For instance, in the case of plastic waste, the integration of TOPSIS with neural networks (GNNs) has led to significant improvements in material recovery efficiency [62]. Meanwhile, blockchain-based governance mechanisms have facilitated greater traceability and collaboration among stakeholders.

In terms of discussion around this framework, the following can be pointed out:

-

Bridging research and practice: Unlike traditional theoretical models, this framework offers a structured, actionable approach, addressing gaps highlighted by Kirchherr et al. [14].

-

Governance integration: While many AI frameworks focus on efficiency, this proposal integrates governance mechanisms such as blockchain-based traceability and policy-driven incentives, ensuring a holistic and practical approach [19]. The effectiveness of governance also depends on reducing transaction costs [24], a recurring challenge in reverse supply chains. According to Coase [23], well-defined regulatory structures can help minimize these costs and ensure greater operational efficiency. In this context, industrial symbiosis and public–private partnerships (PPPs) emerge as key solutions to financially support the implementation of advanced tracking and optimization technologies.

Given the complexity of RSCSW governance, an implementation roadmap is proposed detailing the necessary steps for the practical adoption of this framework. Figure 2 summarizes the five-phase roadmap for implementing MCDA + AI in RSCSW governance, supporting the framework proposed in Table 6.

To facilitate the adoption of the proposed framework, a structured roadmap is proposed in five phases. Table 6 aligns each roadmap phase with empirical studies from the review corpus, ensuring methodological consistency and verifiability.

The justification for the roadmap is that it complements the conceptual framework by providing a practical step-by-step implementation strategy; enhances the study’s applicability across different regulatory and operational contexts; and integrates theoretical governance models with technological tools and operational strategies, facilitating adoption by public and private stakeholders.

Although structured, this roadmap assumes high levels of technological readiness, which may not be feasible in all contexts. Gardas et al. [52] highlight that infrastructure disparities could delay implementation in certain regions. Regarding stakeholder engagement, while blockchain and AI ensure transparency, stakeholder resistance to digital transformation remains a major concern [28]. Ensuring broad participation will be essential for successful adoption.

4.5. Governance and Policy Barriers

Despite the promising advancements, several governance and policy barriers hinder the widespread adoption of MCDA and AI in RSCSW:

-

Regulatory fragmentation: The lack of cohesive regulatory frameworks across regions and sectors creates challenges in implementing standardized practices for waste management. Kirchherr et al. [14] identified regulatory inconsistencies as a major barrier to the circular economy, which aligns with the findings of Trochu et al. [62], who highlighted the need for accurate information and policy adjustments to improve reverse logistics networks.

-

Technological challenges: The integration of AI and MCDA requires significant technological infrastructure, including data collection systems, computational resources, and skilled personnel. Tseng et al. [19] emphasized the need for technological innovations to support the integration of MCDA and AI in RSCSW, which is further supported by Gardas et al. [52], who identified “Technology Integration” as a critical factor for adopting AI in supply chains.

-

Collaboration among stakeholders: Effective governance of RSCSW requires collaboration among diverse stakeholders, including governments, private sectors, and communities. Tseng et al. [19] highlight the role of Information and Communication Technologies (ICT), such as blockchain and the Internet of Things (IoT), in improving governance by providing greater visibility over material and waste flows.

-

The adoption of AI and MCDA in waste management may have significant social and ethical implications, such as the displacement of low-skilled workers. Public policies and training initiatives must be implemented to ensure a fair and inclusive transition to more advanced technologies.

Thus, it is concluded that fragmented policy frameworks do exist, and many studies advocate for AI-driven governance, but Kirchherr et al. [14] emphasize that without standardized policies, these technologies will remain underutilized. Moreover, there are also financial barriers; unlike high-income economies that lead AI implementation, emerging economies face limited public and private funding for waste management technologies, a concern echoed by Trochu et al. [61].

5. Conclusions

The integration of multi-criteria decision analysis (MCDA) and Artificial Intelligence (AI) represents a paradigm shift in the optimization of reverse supply chains of solid waste (RSCSW) within the circular economy framework. Compared to traditional practices, this integration delivers substantial improvements in operational efficiency, cost reduction, and material recovery. Furthermore, it enhances governance and sustainability by promoting transparency, traceability, and environmental stewardship.

However, significant challenges remain, including regulatory fragmentation, technological barriers, and the need for stakeholder collaboration. Future research should focus on developing comprehensive frameworks that combine MCDA, AI, and emerging technologies like blockchain and IoT to further enhance the efficiency and sustainability of RSCSW. Policymakers and industry leaders must also prioritize the development of cohesive regulatory frameworks and invest in technological infrastructure to support the widespread adoption of these advanced decision-making tools.

In terms of the theoretical contribution and framework differentiation, the proposed framework advances existing theory by integrating MCDA and AI into a holistic governance model for RSCSW. Unlike traditional approaches, which often focus on isolated aspects of waste management, this framework offers a systemic view that balances economic, environmental, and social criteria [17]. Moreover, the incorporation of emerging technologies, such as blockchain and IoT, fosters greater transparency and operational efficiency, aligning with the principles of the circular economy [14]. This innovative approach not only optimizes waste management but also contributes to the creation of more resilient and sustainable supply chains.

Although this study demonstrates significant advancements in RSCSW governance through the integration of MCDA and AI, several limitations must be acknowledged to ensure a balanced interpretation of the findings: (1) The reviewed literature may be subject to publication bias, as studies with positive results are more likely to achieve academic dissemination [40]. This could lead to an overestimation of the benefits of integrating MCDA and AI in waste management. To mitigate this bias, future reviews should consider including gray literature and unpublished studies. (2) The heterogeneity of the methods employed in the analyzed articles hinders direct comparisons between different approaches [41]. This heterogeneity can be attributed to regional, sectoral, and methodological differences. For instance, studies conducted in developed countries often benefit from advanced technological infrastructure, while those in developing countries face challenges such as limited resources and inadequate infrastructure. Additionally, the diversity of sectors addressed (e.g., plastic waste, electronic waste, organic waste) and the variety of MCDA and AI techniques used contribute to the variability of the results. (3) The certainty of the evidence presented in this review should be interpreted with caution, as many of the included studies have methodological limitations, such as small sample sizes, lack of randomization, and reliance on case studies from specific regions. These limitations may affect the generalizability of the findings, particularly in contexts with limited technological infrastructure [35]. In terms of challenges, the implementation of AI for governance faces significant technical and operational challenges, such as the need for advanced computational infrastructure and high-quality data. These challenges may limit the applicability of MCDA and AI in emerging countries, where such resources are often scarce.

These limitations highlight the need for future research focused on the empirical validation of these models in different socio-economic contexts, as well as the development of standardized frameworks to address heterogeneity and improve the robustness of findings. Thus, future research should focus on validating the proposed framework across diverse sectors (e.g., construction, agriculture, urban waste) and geographical regions, incorporating socio-economic, institutional, and regulatory variables. Beyond theoretical validation, future work should also explore real-world applications in public policy planning, corporate waste strategies, and intermunicipal cooperation. Pilot implementations can assess the framework’s capacity to guide strategic decisions, prioritize interventions, and support compliance within RSCSW governance. Frameworks such as those suggested by Geissdoerfer et al. [15] can serve as foundational references for sectoral adaptation. Furthermore, the integration of emerging technologies, including blockchain, the Internet of Things (IoT), and digital twins, holds great potential to improve the transparency and efficiency of RSCSW processes. As highlighted by Tseng et al. [19], advancements like big data and blockchain can significantly enhance traceability and governance, further strengthening these systems.

The development of fiscal incentives to support the adoption of AI and MCDA within RSCSW represents another crucial area of focus, as emphasized by Kirchherr et al. [14]. Policymakers must prioritize harmonizing international regulations to mitigate fragmentation and foster cohesive governance frameworks, a need also identified by Wilson et al. [12]. Additionally, the practical applications of these methodologies can be explored through detailed case studies, which would illustrate their impact on decision-making processes in waste management.