Location–Routing Problems with Sustainability and Resilience Concerns: A Systematic Review

Abstract

Background: Location and distribution decisions are key to efficient logistics network design and are often addressed in an integrated manner as Location–Routing Problems (LRPs). Today, sustainability and resilience must be considered when designing competitive networks. This systematic review examines how and at what decision level both concerns are explored in LRPs, highlighting trends and future research challenges. Methods: A search was conducted in the Scopus database on 3 January 2024. Articles not written in English or lacking a sustainability or resilience focus were excluded. The 36 most-cited articles were selected and analyzed descriptively and theoretically, considering their approaches to sustainability and resilience, as well as the decision levels at which these approaches were considered. The studies were also analyzed based on model features and solving approaches. Results: Our findings indicated that social sustainability was the most neglected. The environmental pillar was often focused on minimizing atmospheric pollution from distribution. Regarding resilience, proactive and reactive strategies were employed to minimize disruption costs and risks and maximize network reliability. Conclusions: Research on sustainable and resilient LRPs is growing, but remains fragmented. Future studies should explore the integration of social impacts, uncertainty modeling, and real-world applications. Stronger alignment with decision maker needs and more holistic evaluation frameworks are essential to support resilient and sustainable network design.

1. Introduction

In today’s globalized world, supply chains play an essential role in economic development [1]. The increasing industrialization and globalization of supply chains require more effective management. The design of the supply chain network plays a major role in supply chain performance [2]. The distribution network found at the end of supply chains is particularly important, as it often comprises several small flows of goods towards retailers or end customers. Designing these networks raises two main problems, which are locating facilities and designing the distribution routes that should serve customers [3]. These problems represent two recognized interdependent logistics decisions that should be approached in an integrated manner as LRPs [4].

Concerning supply chain decisions, sustainability has become a critical factor [2]. This growing concern results from the accelerated pace of consumption and production, supported by technological developments, as well as increased pollution and the depletion of natural resources in the interest of economic development. The negative social and environmental effects of rapid industrial growth demand the adoption of sustainable production and consumption practices. Sustainability concerns are, thus, forcing regulatory authorities, consumers, and organizations to reconsider their practices and question their implications for society and the environment [5].

The World Commission on Environment and Development [6] defines sustainable development as “development that meets the needs of the present without compromising the ability of future generations to meet their own needs”. Organizational sustainability comprises the following three components: environment, society, and economic performance [7]. These correspond to the three dimensions of sustainable development, also known as the triple bottom line [5]. According to this idea, at the intersection of social, environmental, and economic performance, organizations can engage in activities that have a positive impact on society and the environment and lead to long-term economic benefits and a competitive advantage for the company [7].

A sustainable supply chain must integrate environmental and social measures into economic practices [8]. Sustainable supply chains play an essential role in promoting companies’ activities towards sustainability [9].

The primary goal of any supply chain remains the proper delivery of products or services to customers. However, due to the uncertainty of today’s competitive environment, supply chains are exposed to different events that can interfere with their operations [10]. Under normal circumstances, they must be able to quickly adapt to common uncertainties, namely fluctuations in demand and supply or even small delays. However, they must also have the capacity to remain unaffected, or less affected, by unexpected disruptions such as natural or man-made disasters, labor strikes, and economic crises [10,11]. These disturbances, even with a low probability of occurrence, can, if they occur, cause serious impacts on companies [12]. Moreover, given the complexity and interdependence of supply chains, the impacts of disturbances can be subject to a ripple effect and be propagated and amplified throughout the different supply chain levels [13].

Ensuring the continuity of operations is crucial, even with the possibility of deteriorating financial results in steady-state operations [14]. Addressing supply chain resilience has, therefore, become a key factor in developing a sustainable supply chain competitive advantage [12]. Ribeiro and Barbosa-Póvoa [12] argue that a resilient supply chain “should be able to prepare, respond and recover from disturbances and afterwards maintain a positive steady state operation in an acceptable cost and time”. A company that responds to a disturbance better than its competitors can effectively improve its market position [15].

Sustainability and resilience have, thus, become two important trends in the supply chain domain, and, therefore, in the logistics domain. In recent years, the incorporation of sustainability concerns into the quantitative modeling of numerous supply chain problems has received increasing attention [8]. The use of quantitative models to support the development and operation of resilient supply chains has become prominent, although not yet thoroughly explored in the literature [12]. The integration of different decision-making levels in models—from strategic to operational—has been widely addressed in the literature due to its positive impact on supply chain performance and competitive advantage. Initially, most models focused primarily on optimizing economic objectives. However, there has been a clear shift toward incorporating environmental and social concerns, leading to analyses of the trade-offs among these dimensions with the aim of fostering the sustainability of logistics systems [16,17]. Similarly, resilience has gained prominence as a critical factor, with several authors advocating for the integration of proactive and reactive strategies into comprehensive models that encompass various decision-making levels [18]. This approach aims to enhance the ability of supply chains to cope with disruptions, thereby ensuring operational continuity.

According to Carissimi et al. [19], the relationship between sustainability and resilience in supply chain management can be framed from four distinct perspectives. First, the two concepts can be regarded as autonomous, each characterized by specific objectives and dedicated implementation strategies. In contrast, they can be seen as synonymous, particularly when both are aligned toward the overarching goal of ensuring the long-term survivability of the supply chain. A third perspective views resilience as a component of sustainability, based on the premise that supply chains unable to cope with disruptions inevitably compromise their sustainable continuity. Conversely, sustainability may also be seen as a component of resilience, as sustainable practices help to reduce social, environmental, and economic risks—thus strengthening the supply chain’s ability to prepare for, respond to, and recover effectively from disruptive events.

This paper presents a systematic review of LRPs that integrate the concepts of sustainability, resilience, or both. The broad definition of sustainability and the ambiguity of the resilience concept mean that the strategies applied can vary between different works, depending on the authors and their objectives. Additionally, LRPs involve decisions that often belong to different planning levels, namely strategic location decisions and operational distribution decisions. Therefore, this review seeks to analyze, in detail, the approaches that authors have applied at each of these decision levels to develop more sustainable and resilient networks. Each paper is also analyzed in terms of the key features of the models and solving methods employed. Finally, some research gaps and promising areas for future development are identified. Following this motivation, this systematic review addresses the following research questions:

• RQ1: When and in which journals were the most relevant studies published regarding the integration of sustainability and resilience into LRPs?
• RQ2: How and at what decision level are environmental and social sustainability addressed in LRPs?
• RQ3: How and at what decision level is resilience addressed in LRPs?
• RQ4: What are the characteristics of LRPs that address sustainability and resilience concerns, in terms of model features and solving approaches?
• RQ5: What challenges remain regarding the integration of sustainability and resilience into LRPs and what directions for future research can be identified?

The remainder of this paper is organized as follows: Section 2 presents relevant prior literature reviews and highlights the contribution of the present paper. In Section 3, the review methodology adopted is described, and in Section 4, the material collected for analysis is defined and delimited. Section 5 includes an assessment of the formal aspects of the material reviewed, i.e., a descriptive analysis that enables answering RQ1. Section 6 and Section 7 contain the theoretical analysis of the material reviewed. RQ2–RQ4 are answered in Section 6 and RQ5 is answered in Section 7. Finally, conclusions are drawn in Section 8.

2. Previous Works

This section presents relevant prior literature reviews, highlighting the contribution of the present paper. The following two main categories of contributions can be distinguished: reviews focused on LRPs and reviews focused on sustainability and resilience in the supply chain domain.

Since it was recognized that making location and distribution decisions independently can lead to sub-optimal solutions, the literature on LRP has grown considerably. As can be seen in

Table 1. Characterization of recent literature reviews on LRPs.

Reference	Method	Material Reviewed	Problems Addressed
Nagy and Salhi [20]	Survey and classification	JA	Standard LRP and variants
Lopes et al. [21]	Survey and taxonomy	JA [149]	Standard LRP and variants
Prodhon and Prins [3]	Survey	JA [72]	Standard LRP and variants
Drexl and Schneider [22]	Survey and classification	JA, CP, TR, and PhD_d	Variants of the standard LRP
Schneider & Drexl [23]	Survey	JA, CP, TR, and PhD_d	Standard LRP
Mara et al. [24]	Survey and taxonomy	JA [222]	Standard LRP and variants
Tadaros and Migdalas [25]	Survey and classification	JA [80]	Multi-objective LRP
Hosoda and Irohara [26]	Survey and classification	JA and CP	Variants of the standard LRP
Schiffer et al. [27]	Survey and classification	N/A	Vehicle Routing Problem (VRP) and LRP with intermediate stops
Cuda et al. [28]	Survey	N/A	Two-echelon LRP, two-echelon VRP, and truck and trailer routing problem
Y. Li et al. [29]	Literature review	JA [66]	Green LRP
Arevalo-Ascanio et al. [30]	Literature review	JA [36]	Standard LRP and variants

, this rapid growth in LRP research has led to the appearance of several reviews in a short period of time.

Several reviews have been developed considering the standard LRP and its numerous variants and extensions. The survey by Nagy and Salhi [20] is a seminal contribution to the literature on LRP, and was later updated in the study by Prodhon and Prins [3]. Lopes et al. [21] propose a review and a taxonomy for the LRP literature, giving special emphasis to multi-objective approaches. Drexl and Schneider [22] propose a survey addressing variants and extensions of the standard LRP, while Schneider & Drexl [23] focused only on the standard LRP. More recently, Mara et al. [24] continue the previous studies by Lopes et al. [21] and Prodhon and Prins [3] by proposing a review and a taxonomy for the literature on LRP published from 2014 to 2019. Hosoda and Irohara [26] propose an update of the survey by Drexl and Schneider [22] and categorize the studies published from 2015 to 2020 according to the main characteristics of the problem, applications, and solving approaches. Some authors focus their review studies on more specific variants of the standard LRP, namely the two-echelon LRP [28], the LRP with intermediate stops [27], the multi-objective LRP [25], and the green LRP [29]. Arevalo-Ascanio et al. [30] focus on the strategies and methods used to solve LRPs.

Regarding reviews focused on supply chain network design and optimization, there are some studies that focus on sustainability [31,32,33], resilience [12,34] or both of these topics [1,35]. However, many of these studies adopt a broad supply chain network design perspective, incorporating decisions beyond location and routing—such as inventory, sourcing, and production—which differentiates them from LRP-focused contributions. For this reason, the scope of these studies is much broader than that of the present review.

As can be seen, some reviews related to the research focus of this paper have been published previously. However, there is still a lack of comprehensive analysis specifically focused on how environmental and social sustainability and resilience are integrated into LRPs, especially when considering the planning level at which these decisions are made. Accordingly, this work aims to contribute to this research topic by narrowing the focus to LRPs and offering a decision-level-based analysis of how sustainability and resilience are embedded in these models.

3. Review Methodology

The methodology adopted in this systematic literature review was based on the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA 2020) statement. The methods and results of systematic reviews should be specified in detail, allowing readers to assess the reliability and applicability of the results obtained. Additionally, they should be reproducible or updatable. The PRISMA 2020 statement provides a set of guidelines that promote greater transparency and completeness in the review process [36].

The systematic literature review presented in this paper was conducted according to the following steps [31,37]: material collection, descriptive analysis, category selection, and material evaluation. Firstly, the material collected was defined and delimited. In the descriptive analysis, the formal aspects of the material collected were evaluated to obtain the first insights upon which the theoretical analysis was based. Accordingly, each article was assessed based on its year of publication, number of citations, and the publishing journal. In the last two steps, the material was categorized and analyzed according to the structural dimensions defined. Each article was categorized and analyzed based on its approaches to sustainability and resilience, as well as the decision levels at which these approaches were considered. The studies were also analyzed based on relevant features of the models and solving approaches applied. The results were interpreted to identify research gaps and relevant issues.

Data was collected by the first author using a standardized Excel spreadsheet developed for this systematic review, without the use of automation tools. In cases of ambiguity, the data was discussed with the co-authors until a consensus was reached. Information missing from the original studies was recognized in the manuscript.

This literature review did not involve the development or registration of a formal review protocol, but instead followed an exploratory approach aligned with the objective of analyzing and characterizing the existing literature. Furthermore, a formal critical appraisal of the included sources of evidence was not conducted. Since all selected studies were published in peer-reviewed journals indexed in the Scopus database, a minimum level of scientific rigor and methodological quality was assumed. As such, the objective of identifying and characterizing trends and gaps in the literature was prioritized over assessing the quality of individual studies. Consequently, no quality appraisal results are presented.

4. Material Collection

The search was carried out on 3 January 2024 in the Scopus database. The search term “location-routing” was combined with two large groups of terms, one related to sustainability and the other to resilience. The search terms belonging to each group are shown in

Table 2. Terms employed in the search.

Search Terms Concerning Sustainability				Search Terms Concerning Resilience
sustainab * environment * green pollution emission fuel	waste hazard * recycle * resource consumption energy consumption reverse logistic	closed loop social equity equality balance * fair *	shortage coverage job opportunity * employment safety satisfaction	resilien * reliab * flexib * robust * redundan * agil * collaborat *	disturbance failure unfores * unexpect * disrupt * risk uncertain *	penalty cost failure cost lost sale shortage cost backlog cost damage cost recovery cost

* The asterisk (*) is used as a wildcard character in the search query to capture all possible word endings. For example, “sustainab *” retrieves “sustainable”, “sustainability”, and similar variations.

. The search was performed within the article title, abstract, and keywords, returning 697 records. Prior to screening, 235 records were automatically excluded using filtering tools based on document type and language, retaining only peer-reviewed journal articles written in English. No duplicate records were identified during this process. This resulted in 462 studies selected for manual screening based on title and abstract, each assessed for its focus on sustainability and resilience. Of the 166 studies that remained after this step, the 20 most-cited articles addressing sustainability and the 20 most-cited articles addressing resilience were selected for full-text review. This approach was adopted to ensure the feasibility of the review while maintaining a balanced representation between sustainability and resilience studies. Citation count was used as an indicator of academic relevance and influence, allowing for the identification of the most prominent contributions within each topic.

To ensure the quality and relevance of the selected studies in relation to the research questions and objectives, the following criteria were defined:

• Studies that were not articles written in English were not considered;
• Studies focused on telecommunication networks or other types of networks (e.g., social) were not considered if they were not applied within a logistics or supply chain context;
• Studies that did not address either sustainability or resilience in an LRP model were excluded.

Each article was reviewed thoroughly and all studies that did not meet the inclusion criteria or were not related to the research questions and objectives were excluded from the analysis. Figure 1 details the selection of studies for this systematic review; 36 articles were selected for analysis.

Figure 2 illustrates an analysis of the keywords co-occurring in the studies analyzed. The keyword co-occurrence analysis was conducted using VOSviewer (version 1.6.20). The type of analysis selected was co-occurrence, with keywords as the unit of analysis. The full counting method was applied to ensure that each link had the same weight. A minimum threshold of 1 occurrence per keyword was set to allow for a more comprehensive inclusion of relevant terms, resulting in a total of 138 keywords. All items are displayed, including those not connected to others, to provide a complete overview of the keyword landscape. The default association strength method was used for normalization, and links between keywords represent their co-occurrence within the same article.

As opposed to the keyword sustainability, which appears quite prominently, the term resilience is not visible on the map. Regarding resilience-related concepts, the terms risk, disruption, and reliability are more frequently applied by authors. In the context of sustainability, the map suggests that greater importance is given to the environmental pillar, namely through the consideration of fuel consumption. No strong link is noticeable between the keywords related to sustainability and the terms related to resilience. Finally, there are certain keywords indicating that metaheuristic approaches are often considered and uncertainty is incorporated into the models.

For the present study’s purposes, the interpretation of the inclusion criteria was significantly restricted to focusing on studies relevant to the research questions and objectives. As a result, some studies were identified as relevant within the context of designing sustainable and resilient distribution networks, but were excluded in the selection process.

Studies in which sustainability was only referred to in a qualitative and conceptual way were not included in the analysis. LRPs have several possible applications, since the facilities to be located can be of many distinct types. Therefore, some studies only address the social and environmental dimensions of sustainability in their model application. Regarding the environmental sustainability pillar, some studies aim to optimize the selection of locations for charging stations and the routing of electric vehicles. Both problems have been extensively studied in order to develop greener logistics distribution networks and meet the sustainable development goals of companies [39]. Other works have dealt with waste management networks and hazardous materials, with the aim to locate waste treatment and recycling facilities rather than just basic disposal centers. Aligned with this motivation, some studies address residue collection networks to be transformed into energy. In these cases, residues, such as used cooking oil or municipal solid waste, must be collected in generation areas to be transformed into energy, such as biofuel, in conversion centers. With the growing population and consequent increase in waste generation, the logistics of hazardous materials and waste plays a key role in reducing costs and potential risks posed to the environment and society [40,41].

There are studies that address networks with specific configurations, which could bring environmental or social advantages from a sustainability point of view. Closed-loop or reverse supply chain configurations, perceived as applications of circular economy principles at the supply chain level [42], and resource sharing approaches can contribute to a more efficient use of resources. The integration of green vehicles, such as drones and electric vehicles, into distribution fleets can reduce the environmental impact of the distribution process. Articles in which the potential environmental or social benefits provided by these strategies were not explicitly evaluated were not considered in the present review.

Studies dealing with social equity in emergency logistics or networks operating with obnoxious materials were also not included in the analysis. These often involve a range of humanitarian and political issues related to access to basic resources and services, which go beyond the scope of the present review.

The resilience definition considered in the present study was the one presented in Section 1. Only disruptions were analyzed, i.e., events that abruptly interrupt operations and completely block value creation activities. Thus, small variations in supply and demand were not considered and, therefore, studies that only modeled these uncertainties were not included in the analysis. Studies dealing with the resilience of telecommunications networks were also not included in the analysis to avoid broadening the scope of the review beyond logistics applications.

During the screening phase, some relevant studies related to sustainability and resilience were identified, such as the examples mentioned above. These articles were excluded to avoid broadening the scope of the review too much, allowing for a more focused and in-depth analysis of the selected studies.

5. Descriptive Analysis

To better characterize the material analyzed, descriptive dimensions are used to classify the articles. Accordingly, each article is assessed in relation to its year of publication, the number of citations, and the publishing journal. The years of publication of the articles under analysis are displayed in Figure 3. It should be noted that the analysis is carried out separately for studies addressing sustainability and those considering resilience. The graphs also show the number of citations in Scopus. Both sustainability and resilience are recent concepts, albeit with a growing presence in the LRP literature. The number of citations obtained for studies related to sustainability is significantly higher than that obtained for studies concerning resilience. Graph (a) shows a downward trend in citations between 2016 and 2019. Interestingly, there is a peak in citations in Graph (b) in 2019.

The articles analyzed can be found in 24 different journals. The graph in Figure 4 shows the distribution of the articles among the six journals in which more than one article is published. Many studies are published in Operational Research, Industrial Engineering, and Logistics journals. This reinforces the view that LRPs are often used in the integrated optimization of location and distribution logistics decisions. It is worth noting that the Journal of Cleaner Production, which focuses on cleaner production and environmental and sustainability research and practice, is the second journal where most studies are published. This could highlight the significant contribution of logistics and supply chain network design to sustainable production.

It should be remembered that the descriptive analysis carried out in this section focused on a sample of the most-cited studies and, as such, the conclusions should be interpreted with caution.

6. Category Selection

This section contains a structuring content analysis targeting the studies included in the present review. Each article is categorized and analyzed in relation to its approaches to sustainability and resilience and the decision levels at which these approaches are considered. We examine how the environmental and social pillars of sustainability and resilience approaches have been incorporated into location and routing decisions. The section ends with a summary of the relevant features of the models and solving approaches applied.

6.1. Sustainability Considerations

This section analyzes how the different studies consider the environmental and social pillars of sustainability in both location and routing decisions. It is worth noting that all studies consider the economic dimension of sustainability. As shown in

Table 3. Sustainability considerations in the studies analyzed.

Sustainability Dimension	Environment	Society
Location decisions	Environmental impacts of the establishment of facilities: Biuki et al. [43]; Govindan et al. [44]; Khalili Nasr et al. [45]; J. Tang et al. [46]; S. Wang et al. [47] *; Zhou et al. [48] * Environmental impacts of the operation of facilities	Employment opportunities: Biuki et al. [43]; Khalili Nasr et al. [45]; Ouhader and El-Kyal [49]; Zhalechian et al. [50] Economic development: Zhalechian et al. [50]
	Handling: Biuki et al. [43]; Govindan et al. [44] Producing: Biuki et al. [43]; Govindan et al. [44]; J. Tang et al. [46] Purchasing: J. Tang et al. [46] Recovering products: J. Tang et al. [46] Holding inventory: J. Tang et al. [46]
Routing decisions	Emissions resulting from distribution activity: Biuki et al. [43]; B. Zhang et al. [51]; Govindan et al. [44]; Govindan et al. [52] *; J. Tang et al. [46]; Khalili Nasr et al. [45] *; L. Li et al. [53] *; Nataraj et al. [54]; Quintero-Araújo et al. [55]; S. Wang et al. [47] *; Toro et al. [56]; Zhou et al. [48] *; Zhalechian et al. [50]; Ouhader and El-Kyal [49]; Dukkanci et al. [57] *; Koç et al. [58] * Fuel consumption: Govindan et al. [52] *; Govindan et al. [41] *; Khalili Nasr et al. [45] *; Toro et al. [56]; Nataraj et al. [54]; Quintero-Araújo et al. [55]; S. Wang et al. [47] *; Zhalechian et al. [50]; Dukkanci et al. [57] *; Koç et al. [58] *; J. Yang and Sun [59]	Workload balance: Lin and Kwok [60]; Martínez-Salazar et al. [61]; Safari et al. [62]
Other	Biuki et al. [43]; Govindan et al. [52]; J. Tang et al. [46]; Khalili Nasr et al. [45]; Zhalechian et al. [50]	Biuki et al. [43]; Khalili Nasr et al. [45]

* The sustainability consideration is monetized.

, the analysis summary illustrates that environmental concerns are more frequently included in routing decisions, while the social pillar is significantly less addressed, regardless of the decision level.

6.1.1. Environmental Considerations in Location Decisions

To address environmental concerns in location decisions, emission factors are applied to quantify the environmental impacts of the establishment and operation of facilities. As can be seen in Table 3, all studies consider the minimization of environmental impacts resulting from the establishment of facilities. Some also consider the externalities of operating such facilities.

Govindan et al. [44] address the environmental pillar of sustainability in the design of a perishable food distribution network. The study aims to optimize costs and pollutant emissions, and, in the specific case of manufacturers, the environmental impacts associated with the establishment of facilities and the production process depend on the technology used.

6.1.2. Environmental Considerations in Routing Decisions

Environmental concerns are addressed in routing decisions by considering the minimization of fuel consumption, equivalent environmental impacts, or even both. The reason for this is that CO₂ emissions can be considered as directly proportional to the fuel consumed by vehicles [63].

Several authors quantify transportation externalities, including greenhouse gas emissions, using emissions factors. Emission factors can be applied to estimate environmental impacts based on the distance traveled, the use of a specific network link, the amount of product transported, and the fuel consumed by vehicles. In turn, fuel consumption by vehicles is calculated in diverse ways, often depending on the distance traveled and the payload of vehicles.

S. Wang et al. [47] address the environmental pillar of sustainability in the design of a perishable food distribution network under a carbon tax policy. The study aims to optimize costs and pollutant emissions, including the carbon emissions generated by refrigeration equipment. The authors apply the approach proposed by Xiao et al. [64], in which fuel consumption per unit distance is expressed as a linear function depending on the vehicle payload. The authors find that, although it has no impact on location decisions or the number of vehicles, considering carbon tax policies can reduce environmental impacts. Even so, the total costs are higher when this concern is addressed. Zhou et al. [48] propose a robust optimization method for an LRP under carbon trading policies, showing that these policies can contribute to reducing environmental impacts.

B. Zhang et al. [51] address the environmental pillar of sustainability in the design of an emergency logistics network under uncertainty. Among other parameters, the authors consider emissions to be uncertain. An uncertain multi-objective programming model is proposed to optimize total costs, travel time, and pollutant emissions.

J. Yang and Sun [59] address an LRP focused on the locations of intra-route battery swap stations and the design of routes for a fleet of electric vehicles with limited battery autonomy. The study stands out for estimating the marginal costs of reducing emissions when electric vehicles are used.

Toro et al. [56] address a bi-objective mixed-integer linear programming model for an LRP that aims to optimize total costs and environmental impacts. The authors introduce a new mathematical model to estimate fuel consumption, which is based on the forces acting on each vehicle during its operation. As noted in other studies, using more facilities and vehicles, traveling shorter routes, and carrying less payload can reduce fuel consumption [48,50,56,57]. In the long run, these fuel savings can offset the higher setup costs [56]. Under a carbon trading policy, companies must consider these setup costs against the cost of carbon emissions [48].

Koç et al. [58] and Dukkanci et al. [57] address LRPs that aim to optimize total costs, including the costs of pollutant emissions and fuel consumed by vehicles. Both studies apply approaches based on a well-known emissions model named the Comprehensive Modal Emission Model. Koç et al. [58] propose an integer programming model, and their results for an urban logistics context indicate that the shortest route is not necessarily the cheapest, the fastest, or the one that generates the least environmental impact. In this sense, locating facilities outside the city center can lead to economic and environmental benefits. Dukkanci et al. [57] present a mixed-integer programming model, which includes deciding the speed on each link of the route so that customers’ time windows are met. Their results indicate that speed has a greater impact on fuel consumption than payload, and, therefore, restricting time windows can increase environmental impact. In the study by Ouhader and El-Kyal [49], another well-known emissions model called the Methane Emissions Estimation Tool is employed.

Nataraj et al. [54] and Quintero-Araújo et al. [55] explore the economic and environmental impacts of hypothetical cases of horizontal collaboration. Although the objective of the LRPs addressed is to minimize costs, both studies compare the environmental impacts of the solutions obtained for different levels of cooperation. The authors apply the approach proposed by Ubeda et al. [65], which estimates fuel consumption according to the payload carried and distance traveled by vehicles and applies a conversion factor to calculate the equivalent amount of CO₂. Nataraj et al. [54] conclude that the higher the levels of cooperation, the greater the reduction in total costs and pollutant emissions. However, Quintero-Araújo et al. [55] argue that higher levels of cooperation may not lead to the greatest environmental benefits. This is because, in these scenarios, the number of depots tends to be smaller and the distances traveled tend to be greater. The authors also conclude that the benefits tend to be more significant in cases where the network nodes are more geographically dispersed.

6.1.3. Social Considerations in Location Decisions

Different studies consider the social dimension of sustainability in location decisions, addressing the social implications of job creation through the maximization of employment opportunities created by the opening of facilities. Ouhader and El-Kyal [49] extend the analyses by Nataraj et al. [54] and Quintero-Araújo et al. [55] to the social pillar of sustainability and conclude that horizontal collaboration can have a negative impact on the employment opportunities generated. The authors also compare the different methods that can be used to assess the total gains from collaboration for different partners. Note that, in this study, employment opportunities are calculated based on the used capacity of facilities.

Biuki et al. [43] address a Location–Routing–Inventory Problem (LRIP) for the design of a perishable food distribution network under demand uncertainty. A fuzzy multi-objective optimization model is proposed that considers the trade-offs between the three sustainability dimensions. The results show that decentralizing facilities can promote increased employment opportunities in less developed areas. At the same time, it can reduce distribution costs and pollutant emissions.

Zhalechian et al. [50] explore the trade-offs between the three sustainability dimensions in an LRIP for the design of a closed-loop supply chain under mixed uncertainty. The authors address the impact of establishing facilities on the economic development of local communities. By prioritizing less developed regions, the authors aim to promote balanced economic development. Of note is the uncertainty of the parameters used to calculate social and environmental impacts. The authors show that for low values of returned products, the remanufacturing process in reverse logistics is no longer economically viable.

It should be noted that the studies by Biuki et al. [43] and Zhalechian et al. [50] show an effort to promote the growth of employment opportunities in regions with higher unemployment rates.

6.1.4. Social Considerations in Routing Decisions

Focusing on social sustainability, Lin and Kwok [60] and Martínez-Salazar et al. [61] examine LRPs that aim to optimize total costs and balance the workload between drivers. All studies addressing social sustainability in routing decisions focus on balancing the workload of vehicle drivers. This balance reflects fairness in work assignment and can affect employee satisfaction [60]. Workload balance is defined based on the distance traveled [61,62], as well as the payload carried and the working hours allocated to vehicles [60].

Lin and Kwok [60] explore the possibility of assigning several routes to one vehicle, i.e., the concept of simultaneous assignment. Their results show that in problems where routes are constrained by capacity rather than working time, simultaneous assignment can reduce workload imbalance, although total costs may be slightly higher than those with the standard sequential approach.

6.1.5. Other Sustainability Considerations

J. Tang et al. [46] introduce the concept of Consumer Environmental Behavior (CEB) in the design of a sustainable supply chain network. Initially, a bi-objective mixed-integer programming model is proposed for an LRIP that considers the trade-off between total costs and pollutant emissions. Next, a revenue function that incorporates the CEB concept is optimized. The results show that the growth in revenue does not keep up with the growth in the level of CEB, since the marginal cost of implementing strategies to reduce pollutant emissions increases in line with the degree of CO₂ emissions.

Zhalechian et al. [50] consider minimizing the environmental impact of energy wasted when vehicles wait to receive services at facilities, using a queuing system to calculate this waiting time.

Several studies [43,45,52] evaluate the sustainability performance of supply chain network suppliers and incorporate their evaluation results into the LRP model. Govindan et al. [52] and Khalili Nasr et al. [45] address LRIPs for designing closed-loop supply chains under demand uncertainty. Govindan et al. [52] propose a fuzzy bi-objective model that seeks to optimize shortages and total costs, including the costs of carbon emissions. Khalili Nasr et al. [45] present a fuzzy multi-objective model that seeks to optimize total costs, negative environmental impacts, the employment opportunities generated, lost sales, and procurement value from sustainable suppliers.

The Preference Ranking Organisation METHod for Enrichment Evaluation technique [43] and the fuzzy Best–Worst Method [45], presented by Guo and Zhao [66], are used to evaluate the sustainability performance of suppliers. To evaluate and select circular suppliers, Govindan et al. [52] use a decision support system based on a Fuzzy DEcision MAking Trial and Evaluation Laboratory method and Fuzzy Analysis Network Process. The sustainability criteria used in these multi-criteria decision-making tools are shown in

Table 4. Evaluation criteria for sustainability performance of the suppliers.

Reference	Biuki et al. [43]	Khalili Nasr et al. [45]	Govindan et al. [52]
Economic criteria	Performance history Market shares Production capacity Operating expenses	Quality Reputation On-time delivery Flexibility Technology capability Service and after sales service	Quality Quality control system Previous customers’ satisfaction Quality of after sales service On-time delivery On-time and efficient production Time management Delivery time
Environmental criteria	Resource consumption Pollution production Renewable and non-renewable energy consumption Waste management	Circular Utilizing eco-friendly and recyclable raw materials Using recyclable materials in packaging products Design of products to reuse Green Environmental management systems Managing air pollution resulting from production products Hazardous waste management Environmental certifications Applying proper and clean technologies Green R&D and innovation	Circular Air pollution Environmental standards Eco-friendly raw materials Eco-design Eco-friendly packaging Eco-friendly transportation Clean technology
Social criteria	Labor working conditions Human rights Number of employees Customer satisfaction	Creating job opportunities Information disclosure Occupational health and safety systems The rights of stockholders The interests and rights of employees

.

Considering the sustainability scores obtained, Biuki et al. [43] select only the most sustainable options as potential suppliers in the supply chain network. Similarly, Govindan et al. [52] only consider as qualified suppliers those who have a final score higher than a minimum value defined according to expert opinion. Khalili Nasr et al. [45] use the sustainability score obtained by each supplier as a coefficient in the objective function that aims to maximize the procurement value from sustainable suppliers.

6.2. Resilience Considerations

This section analyzes how the different studies consider resilience aspects in both location and routing decisions.

Table 5. Resilience considerations in the studies analyzed.

Location Decisions	Routing Decisions	Other
Facility capacity disruption: Ahmadi-Javid and Seddighi [67]; Zokaee et al. [68] Facility availability disruption: Elluru et al. [18]; Ghaderi and Burdett [40]; Ukkusuri and Yushimito [69]; Xie et al. [70]; Y. Zhang et al. [71]	Link disruption: Ahmadi et al. [72]; Elluru et al. [18] Vehicle capacity disruption: Quintero-Araujo et al. [73]; S. Zhang et al. [39]; Zokaee et al. [68] Vehicle availability disruption: Ahmadi-Javid and Seddighi [67]; Govindan et al. [41]; Zokaee et al. [68]; L. Li et al. [53] Route reliability: Safari et al. [62]; Veysmoradi et al. [74]; H. Wang et al. [75]; Ukkusuri and Yushimito [69]; Vahdani, Veysmoradi, Noori, et al. [76]; Vahdani, Veysmoradi, Shekari [77]	Khanchehzarrin et al. [78]

presents a summary of this analysis. Interestingly, these aspects are more frequently included in routing decisions.

6.2.1. Resilience Considerations in Location Decisions

To consider resilience in location decisions, authors address disruptions to facilities that can compromise their capacity or total availability. Ahmadi-Javid and Seddighi [67] address an LRP in which facilities and vehicles are subject to random disruptions. The authors consider the production capacity of a potential facility to be uncertain. In cases where the realization of an open facility’s capacity is less than the sum of the allocated customer demands, mitigation strategies must be used to deal with unmet demand. To handle the stochastic objective of minimizing costs, including disruption costs, a risk measure is applied to scalarize this random cost. The expectation, Conditional Value at Risk (CvaR), and worst-case risk measures are used to measure the risk under moderate, cautious, and pessimistic risk measurement policies, respectively. Zokaee et al. [68] address the design of a disaster relief network. The authors consider three types of disruption, namely disruptions to the capacity of facilities, vehicles, and network links. The Failure Mode and Effects Analysis method and the CvaR measure are used to cope with these disruptions. A scenario-based mixed-integer linear programming model is proposed to minimize costs, including disruption costs.

Ukkusuri and Yushimito [69] address an LRP for the pre-positioning of supplies in a post-disaster relief network, without accounting for the capacity of facilities and vehicles. An integer programming model that maximizes the reliability of reaching a demand point is proposed. The reliability of each facility is calculated according to its probability of failure, which the authors considered to be known a priori.

Elluru et al. [18] explore proactive and reactive approaches to designing resilient distribution networks. The proactive version aims to minimize costs, considering the risk factor associated with each facility. The reactive version, which extends the solution generated by the preventive model, considers cost minimization, including disruption costs, in the face of disruptive scenarios. The authors assume that if a facility is disrupted, other facilities can expand their capacity up to a defined limit, with associated cost implications.

Y. Zhang et al. [71], Xie et al. [70], and Ghaderi and Burdett [40] address LRPs in which facilities are subject to probabilistic disruptions independent of each other. Y. Zhang et al. [71] propose a mixed-integer programming model for the two-stage stochastic problem, while Xie et al. [70] address an integer programming model. The minimization of costs and the expected disruption costs among all possible disruption scenarios are considered in both studies. In the study by Y. Zhang et al. [71], when a facility fails, demand nodes originally allocated to it are reinserted into the routes of other operational facilities that have the capacity to serve them, or penalties are incurred. The alternative with the lower cost is selected. Xie et al. [70] assumes that the disruption probability of facilities is identical and known a priori. Each demand node is assigned a certain number of backup facilities. If all the assigned backup facilities are disrupted simultaneously, a penalty cost is incurred for the service loss. To simplify the problem, the authors assume that demand nodes are satisfied in fixed sequences, regardless of the facility from which vehicle is dispatched. Ghaderi and Burdett [40] consider a bimodal network transporting different types of hazardous materials. A two-stage programming model that combines costs and exposure risks into single objective functions is proposed. In the first stage, decisions are made regarding the location of transfer yards; in the second stage, the routing decisions associated with each disruption scenario are determined. It should be noted that if all transfer facilities fail, the product can be transported from its origin to its destination if there is a suitable route. The results show that increasing the probability of disruption can lead to an increase in fixed and transportation costs, since more facilities can be open and longer routes can be used to improve the reliability of the transportation network [40,70,71].

6.2.2. Resilience Considerations in Routing Decisions

To approach resilience in routing decisions, some authors have considered different disruption scenarios in which different links in the distribution networks can be compromised [18,72]. Ahmadi et al. [72] address the design of a disaster relief network. The authors consider partial failures of the transportation network, standard relief time constraints, and the multiple usage of vehicles for serving more than one route. A two-stage stochastic programming model with random travel times is presented, considering the minimization of total distribution time and penalty costs.

Other studies focus on route reliability, which is generally interpreted as the probability of successfully satisfying the demand nodes that are part of a route. It often corresponds to the product of the reliability of the route links, which are known a priori and are considered independently of each other. Safari et al. [62] address an LRP that optimizes total costs, the workload balance between drivers, and minimum route reliability. H. Wang et al. [75], Vahdani, Veysmoradi, Shekari, et al. [77], and Veysmoradi et al. [74] address open LRPs with split deliveries for the design of post-disaster relief networks. Nonlinear programming models are proposed with the aim of minimizing costs and the maximum route travel time and maximizing the minimum route reliability. In the study by Veysmoradi et al. [74], non-capacitated facilities and a bimodal fleet of vehicles are considered, and the reliability of each link differs for ground and air transportation networks. Robust optimization and fuzzy multi-objective programming are combined to deal with the uncertainty in costs and the amount of relief available. The model presented in the study by Vahdani, Veysmoradi, and Shekari, et al. [77] should be distinguished, as it aims to simultaneously route emergency vehicles and repair roads to increase their reliability. In other words, the reliability of each link is updated in line with repair operations. Vahdani, Veysmoradi, and Noori, et al. [76] address a two-stage LRIP for the design of a post-disaster relief network. In the first stage, location allocation and inventory decisions are made to minimize costs; in the second stage, routing decisions are made to minimize vehicle travel cost and time and maximize route reliability. An adapted version of the model that considers split delivery is also presented. A robust optimization approach is applied to deal with uncertain model parameters, including the capacity of facilities and travel time and link reliability. It should be noted that in this study, reliability is calculated according to the sum of the reliability of each link traversed, rather than the product. The results indicate that increasing the minimum route reliability can lead to an increase in the number of facilities, costs, and maximum route traveling time [69,75].

Another frequent strategy is to consider vehicle disruptions, either in terms of availability, capacity, or both. Ahmadi-Javid and Seddighi [67] consider a periodic routing system, where the number of times each vehicle can travel the respective route is modeled as a discrete random variable with finite support. Govindan et al. [41] address an LRP for waste management under uncertainty. A bi-objective mixed-integer linear programming model is proposed to optimize population exposure and total costs, including the cost of fuel consumed by vehicles. Since vehicles are subject to a certain probability of failure, the model is formulated so that the most reliable vehicles are allocated to the collection of infectious waste. L. Li et al. [53] address a location–routing–scheduling problem considering the uncertainty of the driving autonomy of electric vehicles belonging to a mixed bus fleet. The authors introduce a reliability-based two-stage stochastic model with the aim of minimizing network costs, including the costs of vehicle emissions estimated through conversion factors. In the first stage, a scheme of regular services operating on fixed lines and schedules is designed to cover the scheduled trips within a certain reliability range level. In the second stage, ad hoc services provided by third parties or spare buses are determined to deal with energy shortages. The results suggest cost savings when the effects of driving range uncertainty are considered.

S. Zhang et al. [39] address an LRP focused on the locations of intra-route battery swap stations without capacity constraints and the design of routes for a fleet of electric vehicles based on stochastic demands. The authors assume that the actual routes may differ from those defined a priori for the following two reasons: the actual demand of a given node may exceed the vehicle payload, or its autonomy may not be sufficient. The classic resource policy and the preventive restocking policy are extended to consider the influences of battery and vehicle capacity simultaneously. Quintero-Araujo et al. [73] also address an LRP with stochastic demands that are only known when demand nodes are visited by vehicles. A stochastic programming model is proposed, with the aim of minimizing the total expected costs, including those of corrective actions required when route failure occurs due to unexpectedly high demands. Preventive and reactive strategies are considered for potential route failures. According to the former, after serving each demand node, if the expected value of non-served demand in a route exceeds the vehicle payload, then the vehicle completes a “detour” through the facility before visiting the next demand node. The reactive strategy assumes that when the vehicle reaches a node whose demand exceeds its current available payload, it completes a round trip to the facility for a reload. The least expensive option is considered. The authors also consider reliability indexes and safety stock levels, both estimated through simulation.

The literature shows that although designing resilient networks is more expensive, accounting for network disturbances can significantly reduce costs if disruptions occur [18,67,70,71]. Ahmadi-Javid and Seddighi [67] also suggest that, from an economic point of view, it is worse to ignore facility disruption than vehicle disruption. Zokaee et al. [68] show that the effect of disruptions to network links is significantly greater than that of disruptions to vehicles and facilities.

6.2.3. Other Resilience Considerations

Khanchehzarrin et al. [78] address a multi-objective bi-level model considering multiple suppliers for the design of a disaster relief network. The first level minimizes costs and distribution time, and the second level minimizes supply risks and maximizes the efficacy of public donations. The authors suggest that the existence of multiple suppliers can reduce the risk of shortages by avoiding dependence on a single supplier type.

6.3. Modeling and Solving Approaches

This section provides a more detailed analysis of the main characteristics of the mathematical models and solving approaches applied in the studies analyzed. As can be seen in Figure 5, multi-objective approaches are more frequent in studies dealing with sustainability. In fact, different objectives are often used to deal with the conflict between different sustainability dimensions [79]. In single-objective models, there is a clear effort to monetize environmental impacts, namely fuel consumption and CO₂ emissions. These monetized externalities are then aggregated with the economic objective of minimizing costs.

Studies on resilience often focus on assessing the performance of networks in the face of disruptive scenarios and do not use specific resilience metrics. For this reason, single-objective approaches are more common. Economic performance is the most prioritized, with the aim of minimizing the impact of disruptions on costs. Network reliability is also commonly considered in models.

Considering different objectives is not the only way to bring models closer to real-world problems. Heterogeneous fleets, multiple echelons, multiple periods, and time windows can be considered to develop models that are more representative of real-world applications. Figure 6 shows that only some of the studies consider these characteristics in their mathematical models. It should be noted that multi-echelon models are far more addressed in studies related to sustainability. Similarly, Figure 7 also shows that most studies ignore the existence of uncertainties.

Although these characteristics approximate the models to more realistic scenarios, they tend to increase the complexity of the problems, demanding more input data and more efficient solution approaches. The standard LRP, named the Capacitated LRP, is an NP-hard problem, meaning that large instances are difficult to solve using exact approaches [21]. Even so, Figure 8 shows that many studies apply exact approaches, namely Elluru et al. [18] and Toro et al. [56].

The introduction of environmental and social aspects into supply chain problems stems from real-life concerns. Their modeling usually requires a description of a specific context and depends on a particular case [32]. The same can be argued for resilience concerns. Although most of the studies are based on a specific application context, only 42% of the studies analyzed consider empirical data from case studies.

Table 6. Case studies discussed in the studies analyzed.

Reference	Case Study	Country
Zhalechian et al. [50]	Supply chain of LCD and LED TVs	Iran
Govindan et al. [52]	Automotive parts industry	Iran
S. Wang et al. [47]	Third-party cold chain logistics enterprise	Not specified
Lin and Kwok [60]	Local telecommunication service company—bill delivery	Hong Kong
Khalili Nasr et al. [45]	Garment industry	Iran
J. Tang et al. [46]	Petrochemical industry	China
Govindan et al. [41]	Tehran municipality network	Iran
Ahmadi et al. [72]	Earthquake scenario in San Francisco	United States
H. Wang et al. [75]	Great Sichuan Earthquake	China
Ukkusuri and Yushimito [69]	Sioux Falls network	United States
Xie et al. [70]	Rail network of a major US railway company	United States
L. Li et al. [53]	Bus network of Kowloon/New Territories operated by the Hong Kong New World First Bus Company	Hong Kong
Zokaee et al. [68]	Earthquake crisis in Kermanshah province	Iran
Veysmoradi et al. [74]	Earthquake of the southern region of Arasbaran	Iran
Khanchehzarrin et al. [78]	Flood in suburb of Sari	Iran

shows the case studies considered in more detail, as well as the country where each case study is applied.

7. Future Research Directions

In this section, gaps in the literature are identified and avenues for future research are outlined and discussed. The identified research opportunities are focused on the integration of sustainability and resilience into LRPs.

As mentioned in the previous section, the inclusion of more realistic assumptions in the models represents a gap in the studies analyzed. Several authors [39,48,59,69] suggest the inclusion of customer time windows as a challenge for future studies. In most real-world applications, distribution activity is subject to deadlines and schedules that must be met. Furthermore, an analysis of the impact that time windows can have on sustainability pillars and resilience has not been fully explored in LRPs. Only the work by Dukkanci et al. [57] analyzes the effects of time windows on the amount of emissions released.

Another characteristic to consider may be the heterogeneity of vehicles and their drivers. In many real-world applications, considering homogeneous fleets is not a realistic assumption. Accounting not only for differences in vehicle characteristics, but also for differences between drivers, such as distinct behavior patterns, may represent an approach to tackling more realistic scenarios. Furthermore, the effects of heterogeneous fleets on sustainability pillars and resilience have not been thoroughly studied in LRPs.

One approach that has been attracting the interest of researchers and industrial practitioners is the concept of crowd logistics. Crowd logistics refers to the outsourcing of logistics services to a group of players, including everyday people, whose coordination is supported by a technical platform [80]. Although the primary objective of this strategy is to obtain economic benefits for all stakeholders [80], it can lead to more sustainable and flexible systems [81]. An opportunity for future work could be to explore the implications that crowd logistics can have on the design of sustainable and resilient networks.

Addressing uncertainty prevails as a research challenge in LRPs, which is frequently identified in several of the papers analyzed [40,43,47,48,51,54,62,68,75,76,78]. Models could not only consider uncertainty in demand, but also on the supply side [43,48], in travel times [54,73], and in travel costs [70,77]. In the specific case of resilience modeling, it is also imperative to capture the uncertainties associated with network reliability [71,77] and the unpredictability of disturbances and their impacts. In sustainability modeling, the incorporation of uncertainty into the assessment of environmental and social impacts is often suggested as an opportunity for future work [31,32], and remains an area of research with potential for development in LRPs.

Considering the economic dimension of sustainability, almost all the studies analyzed consider the typical objective of minimizing costs. According to Barbosa-Póvoa and Silva et al. [31], problems involving investments should consider project assessment indicators, such as Net Present Value (NPV). Since LRPs can involve decisions ranging from the operational planning level to the strategic level, it may be appropriate to consider this type of metric. Only the work by Toro et al. [56] considers the NPV metric for economic assessment.

Economic assessment should also be further explored to include a wider view of the benefits of implementing more sustainable practices and strategies and how these could be economically measured [31]. Multi-objective approaches are often applied in LRPs to evaluate distribution networks according to the different dimensions of sustainability. However, the trade-off between the costs and benefits of implementing sustainability approaches is not always thoroughly analyzed. Networks with configurations based on the circular economy concept are often advocated as having the potential to be more sustainable [42]. Nevertheless, an analysis of the additional costs associated with reverse flows versus the environmental and social benefits resulting from the implementation of this type of network has not been thoroughly explored in LRPs. Similarly, the integration of green vehicles into distribution fleets is often advocated as a way of reducing the environmental impact of distribution activity and dependence on fossil fuel vehicles [39]. Nonetheless, only one study [59] estimates the marginal cost of reducing emissions when electric vehicles are used.

Several authors suggest the development of models focused on economic optimization, in which environmental and social factors are monetized or considered as constraints, as an alternative to approach sustainability holistically [1,31,32]. Assessing these environmental and social effects in monetary terms that are understandable to decision makers can facilitate the decision process in planning and operating more sustainable distribution networks [82]. In the studies analyzed, carbon emissions and fuel consumption are the only monetized environmental factors. Balanced pricing mechanisms for externalities such as carbon emissions can be a successful approach to reducing the negative environmental effects of networks. Deciding how to encourage individuals, companies, and governments to reduce emissions through proper pricing is a challenging research topic. Therefore, to achieve an appropriate balance between economic and environmental benefits, there is a need to explore these pricing mechanisms in more detail [47,83]. Moreover, social impacts are not monetized in the studies considered, constituting an unexplored research challenge in LRPs.

Regarding the environmental pillar, air pollution is the most frequently applied sustainability criterion in LRPs. Air pollution is often assessed in terms of harmful emissions resulting from the establishment and operation of facilities, as well as the operation of distribution fleets. As continuously suggested [31,32], broader analyses and diversified indicators are needed to comprehensively assess the environmental externalities of networks. To overcome this gap, future research could consider water and energy consumption, the impacts of land use change [79], and waste generation. The Life Cycle Assessment (LCA) methodology could also be used to assess the sum of environmental impacts throughout the entire life cycle of a product [32], presenting a research potential not yet adequately explored in LRPs.

Customer awareness is one of the main reasons why companies are paying more attention to the environmental and social implications of their activities [32]. Apart from the study by J. Tang et al. [46], the impacts of CEB are not thoroughly explored in LRPs. Exploring factors such as the consumer demand for low-carbon products, customer willingness to pay a higher price for these products, the risk appetite of decision makers, the expectations of market development, and the effects of government intervention through policies and legislation to reduce carbon emissions offers an opportunity for future research [46].

The social dimension is the least studied in LRPs, so more attention should be paid to social aspects in future studies [41,44,52]. Exploring new methods to quantify social impacts is continually highlighted as a gap and remains a challenging avenue for future research. Employment opportunities and workload balance correspond to the metrics that are most frequently applied. The extended Social-LCA methodology can be applied to assess the social impacts of networks [32], but is not implemented in any of the studies analyzed.

Today, decision makers are increasingly interested in how companies can align the requirements of social sustainability with those of competitiveness [84]. The socially sustainable practices of a supply chain can have a positive impact on the operational performance of a company [85]. Similarly, a culture of social sustainability in companies can also be positively correlated with their financial success [84]. The effect of social sustainability on the performance of distribution networks has not been properly explored in LRPs. Exploring the impact of ISO 26000 on social responsibility appears as a promising avenue for future research [32].

Barbosa-Póvoa and Silva et al. [31] highlight the assessment of the environmental and social impacts of closed-loop networks as a potential avenue for future research. Khalili Nasr et al. [45] minimize undesired environmental impacts and maximize employment opportunities and sustainable purchases from suppliers in a closed-loop supply chain. Similarly, Zhalechian et al. [50] consider minimizing the environmental externalities and maximizing the social impacts of employment opportunities and economic development in a closed-loop network. However, MahmoumGonbadi et al. [42] stress that environmental objectives should explicitly consider circularity measures, namely the depletion of virgin resource stocks and the avoidance of virgin raw material usage. In addition, the authors argue that while recycling and remanufacturing activities can generate new job opportunities, less dependence on the extraction of raw materials can negatively affect the performance of more traditional industries and have controversial effects on local communities. These and other social impacts have not been thoroughly assessed in LRPs.

Developing models that incorporate reverse flows and address resilience concerns is suggested by different authors [12,34] as an opportunity for further research. Reverse flows can mitigate the impact of disruptions occurring in the upstream part of the supply chain and, consequently, improve the flexibility and resilience of the network [34]. None of the studies analyzed consider the impacts of disruptions in closed-loop networks.

Examining how collaboration, whether vertical or horizontal, can influence the resilience and sustainability of distribution networks offers another avenue for future research. Among the studies analyzed, only three [49,54,55] analyze the environmental impacts resulting from different horizontal collaboration scenarios. Ouhader & El-Kyal [49] go further and assess the social impacts. A collaborative supply chain involves a group of independent companies working together to plan and execute supply chain operations more successfully than when acting in isolation. Although collaboration is based on a shared goal, it is a self-interested process in which partners will only participate if individual benefits are achieved [86]. Each company is generally more interested in its own gains and losses [49]. Therefore, future studies should focus on a more individual assessment of collaboration effects, rather than just providing holistic analysis. Only one study [49] compares the different methods that can be used to allocate the total gains to different partners.

Regarding resilience, new metrics should be explored in future studies. Maharjan & Kato [87] identify a lack of explicit consideration of resilience as an objective function, highlighting the absence of unanimous resilience metrics in the literature on the design of supply chain networks. The studies reviewed consider metrics such as minimizing risks and costs related to disruptions and maximizing network reliability. Indicators such as node criticality, flow complexity, and node complexity can be used to model resilience as an objective function [87]. Furthermore, in the studies analyzed, it is often assumed that the probability of failure of network links and facilities is independent. This assumption may not always be realistic, for example, in natural disaster scenarios. Therefore, one avenue for future research could be the development of novel approaches to calculating these probabilities, which consider correlated and site-dependent disruption probabilities [69,70].

The assessment of important phenomena such as the propagation of disturbances and risk pooling using more complex networks with multiple echelons represents a promising research topic [34] which is not explored in any of the LRPs reviewed.

In the context of resilience, Katsaliaki et al. [13] stress the importance of developing new principles and dynamic supply chain models. In this context, digital solutions such as the Internet of Things and Machine Learning can be used to develop real-time reconfigurable models based on ongoing disturbances and knowledge gained from past disruptive events [13]. This represents an area of research with room for further development in the LRP literature.

The impact of integrating sustainability and resilience into supply chain network design decisions, as well as the trade-off between the two concerns, is still an emerging research topic [31,34,87]. Only three of the studies reviewed consider aspects of sustainability and resilience simultaneously. Resilience is considered with environmental [41,53] and social [62] concerns, but not with both sustainability dimensions at the same time. The relationship between sustainability and resilience has not been properly explored in LRPs and is a promising opportunity for future development.

From the point of view of practical application, the treatment of real cases should be further explored in LRPs. A substantial number of the studies reviewed deal with artificial instances and randomly generated data. The application of models in more realistic scenarios can substantially increase the utility and applicability of research [87].

Addressing the research challenges identified above can increase the complexity of the models. Since LRPs are NP-hard, as already mentioned, the development of more effective and efficient solving approaches remains a compelling avenue for future research [62]. Of note is the need to develop new heuristics [41,49] or metaheuristics [41,44,50,52] and more efficient exact methods [43,45,50] that improve the quality of the lower and upper bounds obtained for large-scale instances while reducing computational effort.

It is important to note that the research opportunities outlined in this section should be interpreted in light of the main limitations of this review. As the analysis was restricted to the most-cited studies indexed in Scopus and written in English, there is a risk of both temporal and publication bias, which may lead to an underrepresentation of more recent but potentially influential research. Nevertheless, it should be emphasized that all abstracts of the retrieved records were reviewed during the screening phase. This procedure allowed the authors to gain a broader understanding of the literature landscape and better contextualize the selected studies. Future systematic reviews could adopt broader inclusion criteria, such as applying citation–time normalization metrics, using multiple databases, and including gray literature to enhance comprehensiveness.

8. Conclusions

This systematic review covered the literature on LRPs that incorporates sustainability, resilience, or both. Sustainability and resilience are two important concepts in supply chain network design, particularly in location and distribution decisions. In this paper, the most-cited studies were subjected to a descriptive and theoretical analysis to determine how and at what decision level both concerns were explored in LRPs, highlighting trends and challenges for future research.

A total of 697 records were initially obtained from Scopus, and only those that complied with the established inclusion criteria were considered. The final sample was formed by the most-cited articles written in English, which considered sustainability, resilience, or both.

The analysis of the selected studies revealed that the social pillar of sustainability continues to be the most neglected. In addition, environmental sustainability and resilience were more frequently considered in routing decisions. Although the studies differed in terms of model features and solving approaches, the results suggested that authors have often focused on the same strategies to increase the sustainability of networks. In the environmental dimension, minimizing atmospheric pollution resulting from distribution activity was the main objective. In the social dimension, maximizing employment opportunities and balancing workload were emphasized. Regarding resilience, the models considered both proactive and reactive strategies, aiming to minimize disruption costs and risks while maximizing network reliability. It should be recognized, however, that this study focused only on the most-cited studies, which may introduce temporal bias by favoring older publications. Consequently, recent but potentially influential works with fewer citations may not have been included in the analysis.

This paper is highly relevant in the current context and represents a useful contribution for both researchers and decision makers. It is the first literature review focused on the integration of sustainability and resilience into LRPs. It provides practitioners and academia with an overview of existing strategies and promising future avenues for designing sustainable and resilient logistics networks.