Digital Twins in Intelligent Transport Systems: A Systematic Review of Resilience Mechanisms and Smart City Implications

Highlights

What are the main findings?

• Digital twins applied to intelligent transport systems are evolving from a simple logic of virtual replication towards service-oriented cyber–physical layers capable of supporting monitoring, prediction, simulation, optimization, decision support, service continuity, and post-disruption learning.
• The literature remains fragmented due to conceptual confusion between digital model, digital shadow, and digital twin; a lack of large-scale empirical validation; a sectoral concentration of studies; and the still limited consideration of human, organizational, and governance dimensions.

What are the implications of the main findings?

• The contribution of digital twins to the resilience of intelligent transport systems depends heavily on data quality, interoperability, real-time synchronization, model reliability, and the ability of stakeholders to transform digital intelligence into operational responses.
• For smart cities, digital twins should be designed not only as technical modeling tools, but also as integrated decision-support mechanisms linking infrastructure monitoring, mobility management, and resilience governance.

Abstract

The intensification of urbanization, the growth in mobility demand and the multiplication of disruptions expose intelligent transport systems to new operational vulnerabilities, making resilience a central issue for the management of transport networks. By conducting a systematic literature review, in accordance with the PRISMA 2020 and PRISMA-S recommendations, this article examines the extent to which, through which functions, under which conditions, and with what level of evidence digital twins can support the resilience of intelligent transport systems. The literature search covered the 2020–2025 period across Scopus, IEEE Xplore, and TRID and was complemented by backward and forward citation chaining, targeting research and review articles dealing with digital twins in transport, mobility, transport infrastructures, or ITS, as well as their links with resilience mechanisms. After screening and eligibility assessment, 61 studies were included in the final analytical corpus. The results show that digital twins are no longer limited to a simple virtual representation of the physical system but are increasingly established as service-oriented cyber–physical layers capable of supporting functions that may contribute to resilience, including real-time visibility, disruption anticipation, scenario simulation, dynamic optimization, decision support, service continuity, and post-disruption learning. The study also highlights several persistent limitations, notably conceptual instability, the sectoral concentration of studies, the lack of large-scale empirical validation, and the insufficient consideration of organizational, human, and governance dimensions. It concludes that digital twins should not be interpreted as automatically enhancing the resilience of ITS, but rather as having the potential to support resilience when they are associated with reliable data, interoperability, synchronization, model reliability, and the ability of stakeholders to transform digital intelligence into coordinated operational responses.

1. Introduction

In an increasingly turbulent environment, transport systems are becoming increasingly structured and complex to manage, control, and master. This heightened turbulence remains the result of the intensification of urbanization and changes in mobility behaviors. Moreover, projections put forward by the United Nations confirm that, by 2050, this rate will reach 68% [1]. Indeed, this trend will increase pressure on the entire transport and mobility ecosystem and, consequently, will multiply its vulnerability to congestion, incidents, and operational and organizational disruptions. In addition to this increasingly heightened geographical pressure, passenger transport demand is expected to increase by 79% by 2050, while freight demand could almost double. Indeed, the analytical framework requires that transport no longer be understood or classified as an immutable system but rather examined through an interconnected approach and strongly linked to data and real-time adjustment capabilities [1,2].

The International Transport Forum states that transport ecosystems have become more fragile due to the multiplication of disruptions affecting the different mobility networks, namely climate change, geopolitical tensions, and energy security [3]. In the same vein, Ref. [4] emphasizes that the tensions and disruptions experienced affect mobility, exchanges, and economic stability [4]. Indeed, this leads transport operators to the need to improve their capacity for prior preparedness, resistance, and recovery after dysfunction, while preserving an appropriate level of service. This consideration repositions resilience as a strategic and structuring requirement for the operational robustness of intelligent transport ecosystems [5,6].

On the other hand, the use of dynamic virtual modeling of the real physical transport system has become more inevitable than ever. This digital alternative, fed by flows and movement information, aims to monitor, analyze, and simulate its functioning. However, Ref. [7] emphasizes that the analysis of the field is characterized by heterogeneous and disparate conceptualization. In transport management, digital twins operate at the operational level and concern real-time monitoring, scenario simulation, the prospective estimation of critical and new events, and, consequently, decision support. Ref. [8] highlights the perspective of aligned, real-time synchronization between the digital model and observed traffic. In the same vein, Ref. [9] shows that these devices are increasingly deployed in risk projection, operational coordination, and the optimization of responses to evolving changes.

In their synthesis, covering 136 studies published between 2000 and 2024, Ref. [7] reveal that 76% of these studies appeared between 2021 and 2024. Confirming this dynamic scientific trend, these authors emphasize that the conceptualization of digital twins and their implications remain heterogeneous, both in terms of the objects studied and the modalities of contribution. In parallel, other studies have been structured according to a sectoral perspective. Ref. [10], for example, focused on transport equipment by examining roads, bridges, tunnels, and hubs. On the other hand, Refs. [11,12] focused on more specific micro-domains, namely public bus transport. By rapidly expanding and becoming denser, this thematic ordering reveals that the analysis of the field suggests a fragmentation of knowledge across several levels of analysis, several categories of transport, and several levels of technological maturity.

Ref. [7] also confirms this conceptual instability and highlights the analytical insufficiency of the concept and its applications. These authors corroborate this observation through the mistaken assimilation, in a significant number of studies, of digital models or digital shadows with genuine digital twins, even in the absence of requirements for bidirectional linkage and interaction, as well as synchronous connectivity. The same authors also report that 31% of the studies do not clearly distinguish between the operational uses effectively derived from the proposed devices. This diagnosis leads to a biased literature, focused more on modeling than on the analysis of the operational scope produced. Refs. [10,13], for their part, confirm this orientation by emphasizing that current research focuses mainly on infrastructures, architecture, stakeholders, as well as implementation and application challenges.

While attesting to the growing importance of the contributions of digital twins as a field of investigation and an object of research in the analysis of intelligent transport systems, these contributions remain dispersed and insufficiently structured. This absence of a cross-cutting analytical framework offering a comprehensive systematic approach to the relationship linking these digital devices to the resilience of transport systems leads us to formulate the following research question: to what extent, through which functions, under which conditions, and with what level of empirical evidence can digital twins support the resilience of intelligent transport systems? By jointly examining these questions, we will seek to analyze the conceptual foundations mobilized and to propose an integrative reading framework that makes it possible to better understand the role of digital twins in the resilience of intelligent transport systems. In particular, the analysis distinguishes potential resilience-supporting functions, such as monitoring, prediction, simulation, optimization, and decision support, from empirically verified resilience benefits related to disruption anticipation, robustness, absorption, service continuity, response, recovery, adaptation, or post-disruption learning.

In what follows, we begin by presenting the methodological approach adopted, specifying the literature search strategy, the inclusion and exclusion criteria, as well as the selection process for the retained articles. Subsequently, we develop the literature review in three stages. We first clarify the conceptual foundations relating to digital twins, intelligent transport systems, and resilience. We then examine the technological architectures, operational purposes, emerging trends, and imbalances observed in the literature. Finally, we analyze the main functions through which digital twins may support the resilience of intelligent transport systems. The final Section 5 concludes the article by highlighting the main theoretical and operational contributions of the study, its limitations, and future research perspectives.

2. Methodology

2.1. Search Strategy and Keywords

The present study adopts a systematic literature review approach, in accordance with the reporting standards defined by PRISMA 2020 [14], and guided by the recommendations of [15,16]. This methodological choice is justified by the objective of synthesizing a still fragmented field, located at the intersection of digital twins, intelligent transport systems, and resilience. The documentary strategy was also aligned with PRISMA-S [17] to strengthen the transparency, traceability, and reproducibility of the search process.

The bibliographic search was conducted in three complementary databases: Scopus, IEEE Xplore, and TRID. Scopus was selected because of its broad coverage of peer-reviewed publications in transport, engineering, intelligent systems, smart cities, and digital technologies. This choice is in line with comparative studies showing that Scopus constitutes one of the most extensive bibliographic databases for identifying the international scientific literature [18,19]. However, given the interdisciplinary nature of digital twin and intelligent transport systems research, the Scopus search was complemented by searches in IEEE Xplore and TRID. IEEE Xplore was included to capture engineering, computer science, cyber–physical systems, Internet of Things, edge/cloud architectures, artificial intelligence, and intelligent transport systems contributions. TRID was added to improve coverage of transport-oriented studies, particularly those dealing with mobility, transport infrastructure, road networks, transport resilience, and transport policy. This multi-database strategy was used to reduce database selection bias and to improve the comprehensiveness, transparency, and reproducibility of the review, in accordance with systematic review requirements [14,17].

The same conceptual search logic was adapted to the syntax and search fields of each database. In Scopus, the query was applied to titles, abstracts, and keywords. In IEEE Xplore, the search was applied to metadata fields, including document title, abstract, and index terms. In TRID, the query was adapted to the database search interface and focused on title, abstract, subject terms, and descriptors. The final searches across the three databases were initially conducted on 20 March 2026 and updated on 3 July 2026. The review protocol was not prospectively registered. The search strings used in each database, the search date, the filters applied, and the number of records retrieved are reported in

Table A1. Boolean search queries used in the review.

Database	Boolean Search Query
Scopus	TITLE-ABS-KEY (((“digital twin*” OR “digital-twin*” OR “digital shadow*”) AND (transport* OR transportation OR traffic OR mobility OR “intelligent transport system*” OR ITS OR “smart transport*” OR “smart mobility” OR “urban mobility” OR railway OR rail OR road* OR highway* OR bridge* OR port* OR maritime OR “urban logistics”) AND ((resilien* OR robustness OR robust* OR disruption* OR failure* OR incident* OR emergency OR emergenc* OR hazard* OR shock* OR recover* OR recovery OR adapt* OR vulnerability OR “risk management” OR “service continuity” OR “continuity of service” OR “post-disruption”) AND (simulation OR prediction OR forecasting OR optimization OR “decision support” OR “real-time monitoring” OR “scenario analysis” OR “adaptive control” OR rerouting OR reconfiguration)))) AND PUBYEAR > 2019 AND PUBYEAR < 2026
IEEE	(“digital twin” OR “digital twins” OR “digital-twin” OR “digital-twins” OR “digital shadow” OR “digital shadows”) AND (transportation OR traffic OR mobility OR “intelligent transport system” OR “intelligent transport systems” OR ITS OR “smart mobility” OR “urban mobility” OR railway OR rail OR road OR highway OR bridge OR port OR maritime OR “urban logistics”) AND (resilience OR resilient OR robustness OR robust OR disruption OR failure OR incident OR emergency OR recovery OR adaptation OR adaptive OR vulnerability OR “risk management” OR “service continuity” OR “continuity of service”) AND (simulation OR prediction OR forecasting OR optimization OR “decision support” OR “real-time monitoring” OR “scenario analysis” OR “adaptive control” OR rerouting OR reconfiguration)
TRID	(“digital twin” OR “digital twins” OR “digital-twin” OR “digital-twins” OR “digital shadow” OR “digital shadows”) AND (resilience OR resilient OR robustness OR robust OR disruption OR disruptions OR failure OR failures OR incident OR incidents OR emergency OR emergencies OR recovery OR recover OR adaptation OR adaptive OR vulnerability OR vulnerabilities OR “risk management” OR “service continuity” OR “continuity of service”) AND (simulation OR simulations OR prediction OR predictions OR forecasting OR forecast OR optimization OR optimisation OR “decision support” OR “real-time monitoring” OR “real time monitoring” OR “scenario analysis” OR “adaptive control” OR rerouting OR reconfiguration)

.

The search strategy was constructed around three complementary lexical blocks. The first block concerned the central concept of digital twin, captured through the expressions “digital twin*” and “digital-twin*”. The second block targeted the field of transport and mobility, incorporating terms such as transport*, transportation, traffic, mobility, intelligent transport system*, ITS, smart transport*, smart mobility, and urban mobility. The third block was designed to capture resilience-relevant contexts and digital twin-enabled functional mechanisms. It was structured around two complementary sub-blocks. The first sub-block covered disruption- and resilience-related terms, including resilien*, robustness, robust*, adapt*, recover*, recovery, disruption*, failure*, incident*, emergency, risk management, vulnerability, and continuity of service. The second sub-block covered technical functions commonly associated with DT-ITS applications, including simulation, prediction, forecasting, optimization, decision support, real-time monitoring, scenario analysis, adaptive control, rerouting, and reconfiguration. These functional terms were considered relevant to the review only when they were explicitly associated with resilience-related contexts, such as disruption anticipation, abnormal operating conditions, recovery, service continuity, adaptive response, or system reconfiguration.

The 2020–2025 period was selected to examine the contemporary phase of operational consolidation of digital twins in intelligent transport systems, rather than to reconstruct the complete historical development of the concept. This choice was informed by the marked concentration of transport-related digital twin research during this period [7] and was defined in accordance with the requirement to report and justify eligibility and search restrictions transparently [14,17]. Foundational studies published before 2020 were retained where relevant for conceptual framing and citation chaining, but were not included in the analytical corpus. The upper limit was set at 2025 because the search update conducted in July 2026 covered only a partial publication year, which could have introduced incomplete-year and indexing bias.

The identification process initially retrieved 3662 records from the two reference bases. The first reference base was built from Scopus, IEEE Xplore, and TRID, and included 3577 records, distributed as follows: 1740 records from Scopus, 1666 from IEEE Xplore, and 171 from TRID. During the identification phase, 2251 records were removed because they did not correspond to the eligible document types, such as conference papers, book chapters, books, and other non-article documents. In addition, 154 duplicate records were removed, leaving 1172 records for screening. During the screening phase, 401 records were excluded as off-topic. Among the remaining 771 records, 400 could not be retained because they were not accessible. This resulted in 371 records being assessed in depth from the first reference base, among which 182 were excluded because they were out of scope, 67 because they covered too wide an area of interest, and 7 on methodological grounds. Together with the 400 inaccessible records, this resulted in a total of 656 excluded reports. Following this assessment, 115 references were retained: 82 references were kept exclusively for contextual and conceptual discussion, while 33 studies were included in the final analytical corpus.

The second reference base initially included 85 records. After the removal of 9 duplicate records, 76 records were assessed according to the same inclusion and exclusion criteria. During this assessment, 4 reports were excluded because they were out of scope, 16 because they covered too wide an area of interest, and 2 on methodological grounds, resulting in a total of 22 excluded reports. Of the 54 references retained, 26 were kept exclusively for contextual and conceptual discussion, while 28 studies were retained for the final analytical corpus. Overall, 169 references were retained across the two reference bases. Of these, 61 studies were included in the final analytical corpus of the review. In addition, 108 references were kept for contextual and conceptual discussion only, but were not treated as part of the analytical corpus. This approach is in line with the recommendations of [17,20] regarding the traceability of search strategies in systematic reviews. This Boolean structure was designed to combine digital twin-related terms, transport-related terms, resilience- and disruption-related terms, and functional terms associated with DT-ITS applications. Duplicate records identified through citation searching corresponded to studies already retrieved through the database search or already retained during the first screening sequence. They were therefore removed before eligibility assessment in order to avoid double counting, and only non-duplicated complementary records were assessed against the same eligibility criteria. In accordance with these parameters, the exact search query in Scopus, IEEE Xplore, and TRID is reported in Table A1. Figure 1 summarizes the complete identification, screening, eligibility, and inclusion process.

2.2. Inclusion and Exclusion Criteria

The eligibility criteria were defined prior to the selection process in order to ensure methodological transparency, limit selection bias, and ensure consistency between the research question, the documentary strategy, and the final corpus, in accordance with the recommendations of [21,22,23].

Although both research articles and review articles were eligible for inclusion, they were not treated as equivalent sources of empirical evidence. Primary studies were used as the main basis for extracting evidence on digital twin architectures, transport modes, disruption contexts, resilience mechanisms, validation methods, and empirical or simulation-based outcomes. Previous review articles were retained only for conceptual positioning, comparison with existing syntheses, identification of research gaps, and support for backward and forward citation chaining. To avoid double counting, evidence reported in previous reviews was not counted as independent empirical evidence when the corresponding original studies were already included in the corpus. In such cases, the primary study was used as the evidence source, while the review article was used only to contextualize the state of the literature or to compare the scope and contribution of the present review with prior syntheses.

The analytical corpus was restricted to research articles and review articles written in English, in order to ensure the scientific homogeneity of the sources. Only peer-reviewed studies were retained, in accordance with the methodological recommendations of [24]. We retained studies that explicitly addressed digital twins, digital shadows, or digital twin-enabled systems in the fields of transport, mobility, intelligent transport systems, or transport infrastructures. To ensure alignment with the research question, studies also had to establish an explicit link with resilience-relevant contexts, mechanisms, or outcomes.

However, we did not treat all digital twin functions as direct evidence of resilience improvement. For analytical coding purposes, the included studies were classified into two categories according to the type of resilience contribution they reported. The first category includes studies reporting potential resilience-supporting functions, where digital twins are used for monitoring, simulation, prediction, optimization, decision support, or real-time data integration, but without direct measurement of resilience outcomes. The second category includes studies reporting empirically verified resilience benefits, where the contribution of digital twins is assessed under disruptive or abnormal conditions through indicators such as robustness, absorption capacity, service continuity, incident response, recovery time, adaptive rerouting, system reconfiguration, or post-disruption learning.

Accordingly, studies were not included solely because they addressed digital twin-enabled simulation, prediction, optimization, forecasting, decision support, real-time monitoring, or real-time data integration. These functions were only considered eligible when the study explicitly connected them to resilience-relevant contexts or outcomes, such as disruption anticipation, failure detection, incident management, emergency response, service continuity, robustness, recovery, adaptive rerouting, system reconfiguration, or post-disruption learning. Conversely, studies dealing exclusively with routine operational efficiency, generic traffic modeling, congestion reduction under normal conditions, or technical optimization without a disruption- or resilience-related framing were excluded from the analytical corpus.

In parallel, we excluded documents that were not peer-reviewed, publications outside the 2020–2025 period or written in a language other than English, studies dealing with digital twins in other sectors without a clear link to transport, as well as works that were too general, descriptive, or insufficiently related to the resilience of transport systems. Studies that did not provide extractable data and duplicate records were also excluded. All records retrieved from Scopus, IEEE Xplore, and TRID were exported and imported into Mendeley for reference management. Deduplication was performed after merging the records from the three databases. First, automated duplicate detection was conducted using DOI, title, first author, publication year, and source title, in accordance with the recommendations of [25]. Second, a manual verification was carried out to identify near-duplicates and database-related overlaps, particularly records with minor variations in titles, abbreviated journal or conference names, inconsistent punctuation, missing DOI information, or differences between online-first and final publication metadata. For each potential duplicate, the most complete bibliographic record was retained, prioritizing the version containing the DOI, abstract, keywords, source title, volume, issue, and page or article number. Records were removed only when they referred to the same publication. References presenting thematic similarity but corresponding to distinct publications were retained for screening. The references identified through citation chaining were subjected to the same eligibility criteria as the records retrieved from the three databases. This methodological precaution makes it possible to avoid an arbitrary integration of complementary studies and ensures the scientific homogeneity of the final corpus.

2.3. Article Selection Process

The article selection process was conducted according to a progressive approach, in accordance with the recommendations of PRISMA 2020, which require the clear documentation of the stages of identification, screening, eligibility, and inclusion of studies [14]. Two reviewers independently examined the titles and abstracts, and then the full texts, against the defined eligibility criteria; disagreements were resolved by consensus. The reasons for exclusion were recorded in order to ensure the transparency of the process and to allow their reporting in the PRISMA diagram.

To reduce subjectivity in the thematic classification, the retained studies were coded using a predefined protocol covering transport domain, digital twin maturity, study type, evidence level, resilience-related function or outcome, and dominant thematic axis. Categories were assigned only when explicitly supported by the title, abstract, keywords, or full text; unclear cases were coded as “unclear” or “not reported”. Two reviewers independently coded the studies and resolved disagreements through full-text rechecking and consensus. No formal inter-reviewer agreement statistic, such as Cohen’s kappa, was calculated. The coding protocol is reported in

Table A6. Thematic coding protocol and category definitions.

Coding Category	Definition	Coding Rule
Transport domain	Main transport field addressed by the study	Coded only when explicitly stated: road, rail, port, maritime, urban mobility, public transport, infrastructure, intermodal, etc.
Study type	Nature of the study design	Review, conceptual framework, simulation, prototype, case study, real-world deployment
DT maturity	Level of digital twin maturity	Digital model, digital shadow, digital twin, advanced/closed-loop DT, unclear
Resilience contribution	Type of contribution to resilience	Potential resilience-supporting function or empirically verified resilience benefit
Functional mechanism	Main DT function addressed	Monitoring, prediction, simulation, optimization, decision support, recovery, learning
Resilience dimension	Resilience aspect addressed	Robustness, redundancy, service continuity, response, recovery, adaptation, reconfiguration
Evidence level	Strength of evidence	Low, moderate, good, strong, based on Table A4
Thematic axis	Dominant analytical theme	Assigned according to the main contribution of the study; multi-label coding allowed only when explicitly justified

.

The main screening decisions and reasons for exclusion are summarized in the PRISMA flow diagram, while the final retained corpus is presented in

Table A2. Selected studies included in the systematic review.

No.	Paper	Proposed Thematic Axis	Digital Twin Relevance	Transport/ITS Relevance	Resilience Relevance	Mechanism/Function
1	[77]	Logistics/humanitarian transport	Direct	Direct	Direct	Emergency response/disaster logistics/evacuation
2	[78]	Connected/autonomous vehicles/vehicular networks	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
3	[79]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Recovery/restoration and interdependent infrastructure resilience
4	[80]	Logistics/city logistics/supply chain	Direct	Direct	Direct or implicit	Disruption impact analysis/resilience planning/supply-chain adaptation
5	[81]	Logistics/healthcare waste transport	Direct	Direct	Direct	Robust optimization under disruption
6	[82]	Railway/smart rail	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
7	[83]	Connected/autonomous vehicles/vehicular networks	Direct	Direct	Direct	Safety, collision avoidance and situational awareness
8	[84]	Connected/autonomous vehicles/vehicular networks	Direct	Direct	Direct	Robust optimization under disruption
9	[85]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
10	[86]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
11	[87]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Structural health monitoring/infrastructure robustness
12	[88]	Maritime/ports/waterways	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
13	[89]	Railway/smart rail	Direct	Direct	Direct	Fatigue damage assessment/structural health monitoring
14	[90]	Urban traffic/ITS/smart mobility	Direct	Direct	Direct	Recovery/restoration and interdependent infrastructure resilience
15	[91]	Maritime/ports/waterways	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
16	[92]	Road infrastructure/bridges/pavements	Direct	Direct	Direct or implicit	Multi-scale infrastructure integration/maintenance support
17	[93]	Connected/autonomous vehicles/vehicular networks	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
18	[94]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
19	[95]	Maritime/ports/waterways	Direct	Direct	Direct or implicit	Cybersecurity/trustworthiness of DT-enabled systems
20	[96]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Structural health monitoring/infrastructure robustness
21	[97]	Connected/autonomous vehicles/vehicular networks	Direct	Direct	Direct	Fault-tolerant federated learning/blockchain-enabled resilience
22	[98]	Urban traffic/ITS/smart mobility	Direct	Direct	Direct	Safety, collision avoidance and situational awareness
23	[99]	Urban traffic/ITS/smart mobility	Direct	Direct	Direct or implicit	Cybersecurity/trustworthiness of DT-enabled systems
24	[100]	Connected/autonomous vehicles/vehicular networks	Direct	Direct	Direct	Safety, collision avoidance and situational awareness
25	[101]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Safety, collision avoidance and situational awareness
26	[102]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
27	[103]	Air transport/UAM/UAV	Direct	Direct	Direct or implicit	Predictive maintenance/fault diagnosis/asset management
28	[104]	Railway/smart rail	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
29	[105]	Connected/autonomous vehicles/vehicular networks	Direct	Direct	Direct or implicit	Vehicle vulnerability prevention/safety and security review
30	[106]	Maritime/ports/waterways	Direct	Direct	Direct	Damage control and decision support
31	[63]	Urban traffic/ITS/smart mobility	Direct	Direct	Direct	Safety, collision avoidance and situational awareness
32	[107]	Road infrastructure/bridges/pavements	Direct	Direct	Direct	Predictive maintenance/fault diagnosis/asset management
33	[108]	Urban traffic/ITS/smart mobility	Direct	Direct	Direct	Safety, collision avoidance and situational awareness
34	[8]	Motorway traffic DT/real-time simulation	Direct	Direct	Indirect	Potential monitoring and simulation for dynamic traffic states
35	[9]	AI-enabled DT/safe intelligent transportation	Direct	Direct	Direct	Safety-oriented monitoring/decision support
36	[42]	Transportation DT/traffic safety and mobility review	Direct	Direct	Direct	Safety-oriented mobility DT
37	[109]	Road engineering lifecycle DT	Direct	Direct	Indirect	Lifecycle monitoring and maintenance
38	[73]	Road infrastructure management DT	Direct	Direct	Indirect	Infrastructure management and maintenance
39	[110]	Adaptive traffic signal DT under limited synchronization	Direct	Direct	Indirect	Adaptive control under constrained conditions
40	[111]	Highway asset maintenance DT	Direct	Direct	Indirect	Asset maintenance/information management
41	[112]	Port asset management/BIM-GIS-DT	Direct	Direct	Indirect	Asset management and maintenance
42	[113]	Rail anomaly detection DT	Direct	Direct	Direct	Failure/anomaly detection and monitoring
43	[114]	Road pavement DT	Direct	Direct	Indirect	Pavement condition/lifecycle monitoring
44	[72]	Port multi-equipment scheduling under uncertainty	Direct	Direct	Direct	Adaptive scheduling under uncertainty
45	[65]	Intermodal container terminal rescheduling	Direct	Direct	Direct	Adaptive rescheduling/disruption management potential
46	[115]	Road geometric DT maintenance	Direct	Direct	Indirect	Maintenance and change detection
47	[116]	Railway bogie DT updating	Direct	Direct	Indirect	Model updating for vehicle components
48	[117]	Bridge traffic load DT	Direct	Direct	Indirect	Structural load/mechanical effects
49	[118]	Bridge lifecycle management DT	Direct	Direct	Indirect	Lifecycle management and maintenance
50	[119]	Digital twin trains/railway digitalization	Direct	Direct	Indirect	AI-powered services and digital train functions
51	[120]	AI-assisted DT for smart railways	Direct	Direct	Indirect	Reference architecture for reliable smart railways
52	[121]	Subway tunnel DT review	Direct	Direct	Indirect	Tunnel intelligence, lifecycle and monitoring
53	[70]	Resilient port DSS with digital twinning	Direct	Direct	Direct	Decision support for port resilience
54	[122]	Rail transit structural health monitoring	Direct	Direct	Direct	Damage/fault monitoring and condition assessment
55	[123]	Pedestrian/connected vehicle in-the-loop DT co-simulation	Direct	Direct	Indirect	Co-simulation for safety-oriented testing
56	[124]	Urban rail traction power DT	Direct	Direct	Indirect	Power-system modeling and reliability support
57	[71]	Autonomous driving testing DT	Direct	Direct	Indirect	Scenario testing and validation
58	[48]	Resilience-oriented DT/adaptability	Direct	Direct	Direct	Adaptability/resilience improvement
59	[125]	Airspace management/advanced air mobility	Direct	Direct	Indirect	Operational coordination and management
60	[126]	Smart freeway DT	Direct	Direct	Indirect	Real-time freeway monitoring/simulation/traffic control
61	[127]	Overheight vehicle warning and rerouting	Direct	Direct	Direct	Incident prevention and adaptive rerouting

. The first reference base contributed 33 studies to the final corpus, while the second reference base contributed 28 studies. Overall, the final analytical corpus consisted of 61 studies.

The data were extracted using a predefined analytical grid, in accordance with the methodological recommendations relating to systematic reviews and the structuring of data extraction processes [16,24,26]. This grid included bibliographic information, the field of application, the type of transport system studied, the nature of the digital twin used, the associated technologies, the resilience mechanisms addressed, the methods employed, the main findings, the identified limitations, and the contributions to the understanding of intelligent transport systems. It enabled us to ensure the homogeneity of coding, facilitate comparison between studies, and structure the thematic synthesis of the corpus.

In addition to data extraction, a study-quality and evidence-strength appraisal was conducted in order to avoid treating heterogeneous types of studies as equivalent sources of evidence. Because the retained corpus included conceptual frameworks, review articles, simulation-based studies, case studies, prototypes, and real-world deployment studies, a tailored appraisal framework was used rather than a single clinical risk-of-bias tool. Each retained study was assessed according to five criteria: clarity of the digital twin concept and system boundaries, methodological transparency, strength of empirical or simulation-based validation, explicitness of the resilience-related mechanism or outcome, and consideration of limitations, uncertainty, or transferability. This appraisal was not used as an additional exclusion criterion, but as a weighting mechanism for interpreting the strength of the evidence reported in the synthesis.

In addition to study type and evidence strength, the maturity of the digital twin system was coded in order to avoid treating digital models, digital shadows, and operational digital twins as equivalent technological configurations. The coding distinguished between digital models, digital shadows, digital twins, and advanced or closed-loop digital twins. This classification was based on the degree of data connection with the physical system, the presence of automatic or manual data exchange, the level of synchronization, the existence of bidirectional interaction, and the presence or absence of feedback or control capabilities. When the available information was insufficient to determine the maturity level, the system was coded as unclear and interpreted cautiously in the synthesis. The digital twin maturity classification criteria are reported in

Table A5. Digital twin maturity classification criteria.

Category	Main Criterion	Data Flow	Synchronization	Feedback/Control
Digital model	Static or manually updated digital representation	Manual or absent	No real-time synchronization	No feedback
Digital shadow	Automatic data flow from the physical system to the digital representation	One-way physical-to-digital	Possible near-real-time update	No systematic feedback
Digital twin	Synchronized representation connected to the physical system	Two-way or decision feedback	Continuous or near-real-time synchronization	Decision support or operational feedback
Advanced/closed-loop digital twin	Operational system with adaptive intervention capability	Bidirectional and operationally integrated	Real-time or near-real-time	Automated or semi-automated control, rerouting, reconfiguration
Unclear	Insufficient information in the study	Not clearly reported	Not clearly reported	Not clearly reported

.

The inventory below (Table A2) presents the studies retained for review after the selection and eligibility stages. Each entry indicates the reference of the study as well as the thematic axis proposed for the synthesis, in order to facilitate the analytical structuring of the corpus and to highlight the main fields of application covered by the retained literature, in line with the predefined inclusion and exclusion criteria as well as with the PRISMA flow diagram presented in Figure 1.

Based on Table A2, a multi-thematic classification was conducted to identify the dominant axes of the corpus, as shown in Figure 2. The strongest concentrations concern road infrastructure, highways and traffic; simulation, prediction and decision support; resilience, security, sustainability and energy; ITS, IoV and connected/autonomous systems; and enabling technologies. Figure 2 also summarizes the aggregated relevance profile of each thematic axis in relation to digital twins, transport/ITS, and resilience. Because the classification is multi-thematic, the same study may be assigned to more than one category when this is explicitly justified by its content. The relevance labels derive from the study-level coding reported in Table A2 while the coding protocol is provided in Table A6.

3. Results of the Systematic Review and Thematic Synthesis

3.1. Conceptual Foundations of the Link Between Digital Twins, Intelligent Transport Systems and Resilience

3.1.1. Genesis, Definition and Evolution of the Digital Twin Concept

The genesis of the digital twin concept generally remains associated with the work of [27], within the framework of product lifecycle management, before it was taken up, introduced, and extended in NASA’s technological roadmap from 2010 onwards by [28].

Initially, its conceptualization had no connection with the field of transport, but was part of an industrial approach aimed at monitoring the development trajectory of a product and ensuring correspondence between a physical entity and its virtual representation. In this sense, Ref. [29] emphasizes, through this historical evolution, that digital twins, in their origins, did not emerge as a simple tool for virtual representation but as a mechanism of continuity and informational linkage.

This was represented as the virtual modeling of a real object, system, or process; continuously updated on the basis of field data; and mobilized to monitor its configuration, examine its functioning, assess its interaction, and support decision-making. Moreover, digital twins cannot be reduced, in their conceptualization, to the automation of data flows between the physical object and its virtual counterpart. Refs. [29,30] corroborate this point and emphasize that, beyond duplication, digital twins must ensure evolutionary, synchronized, and operational modeling throughout the life cycle. However, as argued by [31], despite this nuance, the use of this concept remains uneven and varies according to the field and the use case. These latter authors illustrate this heterogeneity, in the field of transport for example, through the still erroneous assimilation of simple digital models or digital shadows to genuine digital twins. For their part, Refs. [32,33] explicitly recall this gradation and confirm that this nuance remains essential to avoid any terminological confusion.

This is why the evolution of the concept has also led to a clearer distinction between digital model, digital shadow, digital twin, and advanced or closed-loop digital twin. In this reading, and in line with [32,33], a digital model refers to a static or manually updated representation of a physical system, without automatic data exchange. A digital shadow refers to a system in which data flow automatically from the physical system to the digital representation, but without a systematic feedback loop from the digital layer to the physical or operational layer. A digital twin refers to a more advanced configuration characterized by continuous or near-real-time synchronization and the capacity to support bidirectional interaction or decision feedback. Advanced or closed-loop digital twins go further by enabling automated or semi-automated feedback, adaptive control, rerouting, reconfiguration, or operational intervention.

This gradation remains essential to avoid terminological confusion, especially in the transport field, where simple digital models or digital shadows are sometimes assimilated to genuine digital twins. It is also important because the resilience implications of a digital model, a digital shadow, and a closed-loop digital twin are not equivalent. While digital models may support offline analysis and scenario exploration, digital shadows mainly improve visibility and monitoring. More mature digital twins can support decision feedback, adaptive response, and, in some cases, operational reconfiguration. Consequently, resilience-related conclusions were interpreted according to the maturity level of the digital system described in each study. This conceptual evolution is illustrated in Figure 3, which distinguishes the main levels of coupling between the physical system and its digital representation.

Moreover, by moving from a logic of representation to a service-oriented approach, the conceptualization of digital twins has become increasingly linked to specific functions and applications. NASA’s technological roadmap supports this approach and emphasizes this type of linkage in order to support decision-making. Indeed, the importance of this reorientation and transformation lies in preparing its extension to intelligent and complex systems [34]. These authors also recall that the uses of digital twins now cover real-time monitoring, planning, optimization, and maintenance. This evolution is decisive, as it directly prepares its extension to complex systems such as infrastructures, urban networks, or intelligent transport systems.

Indeed, the application of digital twins to other sectors, namely smart cities, energy, healthcare, ports, and road networks, attests to this dynamic, which goes hand in hand with increasing sophistication. In the same vein, Refs. [10,35] clearly illustrate this transposition in the field of transport, and more particularly in smart highways and mobility systems. They also argue that this expansion implies a reconfiguration of the concept with other interconnected, decentralized, and more dynamic objects and systems.

3.1.2. Functional Components of Digital Twins in Transport

When addressing the components and dimensions of digital twins in the field of transport, it clearly appears that they cannot be restricted to a simple dual structure, namely a physical object and its digital replica.

In their studies on transport equipment and infrastructure, Ref. [36] thus specifies that digital twins deployed in this field must be based on five main dimensions, namely the physical asset, the virtual model, digital twin data, the updating and communication mechanism, as well as a management system covering the entire life cycle. Indeed, through these dimensions, these authors emphasize that the importance of digital twins also remains based on the quality of the data mobilized, the dynamics of updating, and the integration of associated services. Indeed, as illustrated in Figure 4, this functional structuring can be represented through an articulation between data sources originating from the real world, their integration into digital twins, and their translation into real-time operational actions.

Thus, illustrating a dual interest, this structuring makes it possible, on the one hand, to clarify what digital twins are in the field of transport and, on the other hand, to pave the way for analyzing intelligent transport systems as a privileged application framework for this technology [10]. Thus, ITS constitute a particularly favorable field for the deployment of digital twins, insofar as they combine an increasing availability of multi-source data, a need for continuous observation of system states, and a requirement for near-real-time action. As shown in Figure 5, these three conditions enable digital twins to transform the technical and informational resources of ITS into capabilities for steering, simulation, and decision support.

3.1.3. Intelligent Transport Systems as an Application Framework

Considered as one of the most suitable application frameworks for intelligent transport systems, these intelligent devices aim to improve transport safety, efficiency, and sustainability. Ref. [37] reveals that ITS rely on automation, real-time traffic data analysis, as well as V2V and V2I communications in order to streamline urban mobility and strengthen road safety. In the same vein, Ref. [38] emphasizes that these ITS remain perceived as essential for optimizing flows, mitigating pollutant emissions, and improving the performance of the transport system. This operational orientation directly brings ITS closer to the cyber–physical paradigm and, consequently, to digital twins. In this sense, Ref. [39] specifies that the incorporation of sensors into infrastructures and means of transport ensures continuous data collection and transfer and, consequently, reconfigures the functioning and processes of ITS. These authors emphasize that ITS now constitute an integrated, connected, distributed, and decision-oriented ecosystem, particularly conducive to the implementation of digital twins. Ref. [13] corroborates this observation and emphasizes that digital twins do not emerge or position themselves alongside ITS, but rather become embedded and integrated as an advanced and sophisticated extension of their capabilities for perception, modeling, simulation, and steering. Moreover, they reveal that the architecture of digital twins-ITS is mainly based on four main dimensions, namely the system structure, expected services, stakeholders, and implementation-related constraints.

In the same vein, Ref. [38] indicates, through their study of an interurban road corridor, that digital twins can only be understood as a developed function of ITS, based on instantaneous data, continuously collected and provided by traffic sensors and sensing devices, as well as V2X communications. Indeed, these authors explain that although digital twins strengthen the capacity for observation, monitoring, simulation, and intervention, they do not replace ITS. Furthermore, Refs. [40,41] suggest, through their study of digital mobility, that the application framework of ITS is not limited solely to infrastructures but also integrates synchronized interactions among all actors, connected objects, and mobility environments. Based on a cloud–edge–device configuration, these authors propose a mobility digital twin composed of three blocks, namely the human digital twin, the vehicle digital twin, and the traffic digital twin. Figure 6, adapted from [40], synthesizes this articulation by showing how the cloud–edge–device configuration supports the construction of a mobility digital twin structured around these three interdependent components.

For their part, Ref. [9] develops a digital twin solution assisted by artificial intelligence, with the objective of predicting traffic risk and congestion, detecting anomalies, and supporting decision-making. Indeed, ITS have become a genuine integrated ecosystem for the activation of digital twins. More specifically, through a hierarchical structure, ITS provide digital twins with the possibility to operate and move from the micro scale to the global network. In this sense, Ref. [42] emphasizes that transportation digital twin systems can, through a hierarchical structure, enable digital twins to operate and move from the local asset to the global network. This modeling contributes to improving operational efficiency, passenger experience, and sustainable urban mobility.

3.1.4. Resilience of Intelligent Transport Systems: Foundations and Dimensions

Gaining increasing importance due to the multiplication of disruptions affecting mobility, infrastructures, and transport services, the recourse to resilience is becoming inevitable and increasingly accelerated. This is justified by the combined effect of operational hazards, weather-related disruptions, capacity restrictions, and the increasing sophistication of interactions between modes and networks [5,43]. By studying urban multimodal networks, these authors link resilience to the capacity of the integrated transport system to absorb shocks and restore its functioning.

For their part, the authors of Ref [44], when addressing urban mobility, associate resilience with the capacity to withstand hazards, absorb impacts, preserve a satisfactory level of service, and recover and reconfigure without financial effort. Thus, Ref. [45] emphasizes the way in which the system reacts and adapts when facing dysfunctions and argues that the objective of resilience is not limited to a simple return to the initial state, but rather leads towards a new, more stable functional state. In the same analytical register, Ref. [46] emphasizes that transport resilience depends mainly on its quantification indicators, its response capacities, and its strengthening approaches.

By projecting this vision onto ITS, this cyber–physical architecture, based on sensors, communication networks, processing platforms, control functions, and digital services, ensures significant improvements in terms of safety, traffic flow, and efficiency. However, Ref. [39] recalls that this will generate a dependence of the system on the quality of information flows, interoperability, computational resources, and data protection. Thus, Figure 7 synthesizes the four structuring dimensions of transport system resilience, namely robustness, redundancy, rapid recovery, and adaptability, by linking them to the operational determinants that condition the system’s capacity to maintain, reorganize, or restore its functions in the face of disruptions.

These dimensions, which interact together, should not be analyzed or interpreted as independent and autonomous components. Moreover, Ref. [5] structures the resilience of multimodal networks around three main dimensions, namely network modeling, resilience assessment, and its optimization. This delineation thus provides ITS resilience with an approach that is both digital and informational. For their part, the authors of Ref. [47] corroborate this convergence and emphasize that failure may result from a loss of connectivity, data inconsistency, software malfunction, a cyberattack, or even desynchronization between subsystems. Consequently, it is no longer linked solely to physical risks or damage.

3.1.5. Conceptual Link Between Digital Twin, ITS and Resilience

Understood within a logic of complementarity, the articulation between the digital twin, ITS, and resilience is apprehended as a relationship of operational interdependence and functional complementarity. Ref [13] clarifies this articulation and argues that ITS provides a socio-technical environment in which information flows, steering mechanisms, and digital interactions are exchanged. These authors demonstrate that digital twins are not juxtaposed with ITS as a separate component or autonomous element, but are embedded within them and positioned as an advanced layer of modeling, analysis, and simulation [7,10].

And from this articulation, resilience intervenes as an evaluative perspective and an analytical purpose. Ref. [48] explicitly highlights that the analysis of transport digital twins must mainly concern their contribution to the adaptive and reconfiguration capacity of these systems. Thus, as illustrated in Figure 8, the conceptual link between ITS, digital twins, and resilience can be understood as a progressive chain in which the informational resources of intelligent transport are transformed into operational intelligence and then into the system’s resilient capacity.

This perspective, on the one hand, corroborates contributions associating digital twins with real-time monitoring, simulation and scenario-building, as well as decision-making steering, and, on the other hand, aligns with findings presenting transport resilience in terms of service continuity, adaptation and recovery [48]. This approach can also be understood through a causal sequence linking ITS to resilience. Refs. [10,49] reveal that digital twins in transport infrastructures can combine data integration and modeling frameworks in order to strengthen operational steering and support maintenance. Moreover, digital twins and ITS, through the combination of sensors and digital twins, ensure system perception as well as the responsiveness of steering actors and operators.

However, the integration of digital twins into ITS does not guarantee automated resilience. Ref. [10] highlights the emergence of architectural and governance constraints, as well as service and integration challenges. Furthermore, these authors specify that transport infrastructures still suffer from insufficiencies in terms of technical and conceptual heterogeneity in the deployment of digital twins. Indeed, this articulation between digital twins and ITS, understood as a conditional potentiality, remains dependent on data quality, information flows, interoperability, synchronization quality, model maturity, and the capacity of actors to exploit the services produced.

3.2. Technological Structuring, Operational Purposes and Literature Trends on Digital Twins in Transport

3.2.1. Architectures and Technologies Mobilized

Viewed from a technological angle, the Internet of Things, through embedded sensors, road devices, and roadside devices, positions itself as the main resource feeding the digital twin. Ref. [10] indicates that the dependence of transport infrastructures on these technologies comes from collection, through integration, to data synchronization. In addition, Ref. [50] mentions that this integration requires increased synchronization, a requirement without which digital twins cannot function. On the other hand, Ref. [38] mentions a second technological family, namely that of connected communications, in particular V2X, V2I, and IoV. These authors explain that their highway digital twins are associated with fixed sensors and vehicle-to-everything interactions. Furthermore, Ref. [51] attests to this technical observation and emphasizes that the deployments of digital twins in the IoV mobilize sharing and cooperation, as well as coordination.

In addition, Ref. [9] presents a third category and groups it into artificial intelligence, machine learning, and data aggregation processes. Assisted by artificial intelligence, these authors describe these enhanced digital twins in a framework that combines network maps, traffic histories, embedded data as well as video sequences in a centralized cyber–physical platform. Ref. [52] highlights the determining and growing role of artificial intelligence in its support of digital twins in order to predict traffic in 5G and IoV environments, in real time.

In the same vein, Ref. [7] proposes another pivotal technology, namely simulation, and confirms that the latter remains omnipresent in most sub-domains of the corpus. Refs. [8,53] illustrate, respectively, in road and port visualization, that simulation-based integration plays a pivotal role and now constitutes an essential direction. Moreover, Ref. [54] also calls attention to the contribution of high-definition mappings and demonstrates that they can serve as a foundation for reactive and evolutive transport digital twins. For their part, Ref. [35] emphasizes that the issues of informational flows, physical representation, and digital simulation remain intimately interconnected.

3.2.2. Dominant Purposes and Uses of Digital Twins in Transport

In terms of usage interests and purposes, the recourse to digital twins remains justified by several mechanisms, namely monitoring, simulation, prediction, optimization, and decision support. Thus, Ref. [7], by studying transport planning, reveals that digital twins are widely deployed in simulation and that the services derived from them are still underdeveloped. As observed and represented in Figure 9, simulation appears as the most frequent modeling component, while a significant proportion of studies remains insufficiently explicit regarding the services actually derived from digital twins.

In the same vein, Ref. [55] shows that the digital twin platform, in road nodes and by tracking congestion and lane configurations in real time, promotes proactive actions and, consequently, improves safety and limits disruptions and interruptions. In the case of smart highways, it establishes itself as an enhanced observer and supervisor of the system through sensors and V2X communication flows [38]. Indeed, real-time monitoring emerges as a first dominant purpose and as a basic function. It constitutes the entry point for other more advanced uses.

On the other hand, Ref. [7] explicitly reveals that, in addition to updated and synchronized monitoring, scenario-building and its simulations constitute a widely recurrent configuration of representation. Operationally, this functionality paves the way for testing other alternatives and obtaining a global and precise view of all collaborators and different stakeholders before reaching the implementation stage. Ref. [8] illustrates this approach by combining and linking observed data with a simulated environment in order to simulate road traffic dynamics. Consequently, scenario simulation, using digital twins that then serve as an operational laboratory, appears as a second recurrent purpose, making it possible to reduce the uncertainty associated with decisions made on the physical system.

Moreover, Ref. [9] presents a third framework based on artificial intelligence and oriented towards state prediction as well as the early detection of critical situations. These authors highlight that this predictive orientation may concern traffic risk forecasting, anomaly detection, and congestion prediction; it may also serve real-time traffic data prediction [52]. In this regard, this orientation strengthens the strategic scope of digital twins and shifts prediction, considered as a third dominant purpose, from a descriptive role towards an anticipatory role.

For their part, Ref. [50] shows another facet of digital twins by studying public transport and associating it with the optimization of operational efficiency as well as the improvement of passenger experience. At the port level, digital twins effectively contribute to improving resource allocation, the fluidity of operations, and reconfiguration under constraints. In this perspective, optimization is positioned as a fourth structuring and clearly visible purpose. It remains essential, as it directly links digital twins to the management of unstable environments and to performance under disruption.

Appearing less as one function among others than as a structuring purpose of digital twins, Ref. [56] states that their value lies in their capacity to combine and transform heterogeneous data, simulations, and predictive models into actionable information in order to support decision-making. Ref. [9] associates their digital twin platform with decision-making capabilities based on the exploitation of artificial intelligence models as well as the integration of multi-source data. Consequently, digital twins are increasingly presented as an assisted decision-making tool that cuts across all these uses. Decision support no longer merely represents or records the system, but contributes to the formulation of actions, the choice of responses, and the evaluation of alternatives.

Moreover, certain sectoral purposes also appear in a more specific manner. In transport infrastructures, Ref. [57] shows that the dominant uses concern asset monitoring, maintenance, operation, and life-cycle management. In ports, greater emphasis is placed on operational coordination, scheduling, and logistics integration. In urban mobility, the focus is directed towards the integrated representation of actors, behavioral analysis, and flow steering services [58]. This sectoral differentiation shows that the purposes of digital twins are not identical everywhere. They are reconfigured according to the nature of the observed system, the level of instrumentation available, and management priorities. Consequently, the mapping of dominant uses reveals a relatively clear hierarchy. Monitoring, simulation, prediction, optimization, and decision support constitute the five most visible purposes in the corpus. Figure 10 presents a hierarchical synthesis of the main purposes associated with digital twins in the field of transport.

These uses show that transport digital twins are increasingly less mobilized as simple digital replicas. They are progressively becoming an operational platform for perception, experimentation, and action. This evolution directly prepares the following section, which will focus on emerging trends and on the imbalances still present in literature.

3.2.3. Emerging Trends and Imbalances in the Literature

Emerging Trends

The analysis of the literature points to the emergence of several trends and converges from an approach of pure delimitation and definition towards a perspective of operational precision and practical application [7]. These authors confirm this orientation and emphasize that the earliest studies dealing with digital twins aimed to address their conceptualization, positioning, and distinction from classical transport models, and that, over time, research began to move towards more concrete operational uses. Ref. [9] corroborates this transition in transport infrastructures and demonstrates that recent studies increasingly address integration challenges and future directions.

On the other hand, one of the most visible markers of this evolution lies in the increasing use of artificial intelligence, data combination, and predictive analysis. Ref. [9], by proposing a digital twins platform, highlights its capacity to integrate infrastructure maps, traffic histories, as well as other telematics data and video streams, while deploying artificial intelligence for their fusion and for detecting potential anomalies. For their part, Ref. [59], from the same perspective, underlines their capacities to manage flows, improve traffic safety control, and provide assistance for autonomous driving. Consequently, digital twins are increasingly emerging as support for inference and an effective decision-support tool, and are no longer used solely to represent the system.

Moreover, Refs. [60,61] reveal the emergence of a third trend, namely the one linking digital twins to sustainability and resilience. These authors already emphasize that studies on transport and logistics associate digital twins with the improvement of environmental performance and energy optimization. For their part, Ref. [56] explains the decisive role of digital twins in contexts of vulnerability, in risk reduction, the simulation of critical events, and decision support. Moreover, Ref. [62] also reveals a growing orientation of digital twins towards simulation, optimization, and resilient operational steering. Consequently, the field is increasingly moving towards more strategic and decision-oriented purposes, namely performance and sustainability.

Imbalances in the Literature

These developments must be addressed in a comprehensive manner and should not conceal the emergence of several asymmetries and structural limitations. Ref. [7] explicitly shows that a significant number of studies focus more on road infrastructures, corridors, railway systems, and some port environments than on more diffuse mobility services or open multimodal systems. The same authors demonstrate that the most developed digital twins remain deployed within more restricted geographical perimeters and shorter time horizons. As a result, the sectoral distribution of studies emerges as a first persistent imbalance, highlighting faster progress on well-delimited objects than on complex large-scale mobility systems. In spite of that, they also show that several contributions still conflate and consider simple digital models and digital shadows as digital twins, although synchronized and bidirectional interconnection is absent or remains incomplete. This erroneous assimilation, although it weakens the entire comparison of contributions, also obscures and makes difficult any assessment of the maturity level of the devices analyzed and the applications reviewed, hence the emergence of the second imbalance directly affecting the analytical robustness of the literature, namely the conceptual instability of the field.

Furthermore, Ref. [60] insists that, despite the growth of studies on digital twins in transport and logistics, they still remain weakly supported by fully operational applications. Moreover, Ref. [10] corroborates this observation and recalls that several constraints and critical issues, such as scalability, data integration, model reliability, and interoperability, continue to limit the implementation of digital twins in transport infrastructures. This implies the emergence of the third imbalance, namely the gap between theoretical ambition and empirical validation. In other words, the literature contains numerous technological promises, but fewer homogeneous pieces of evidence on robust large-scale deployments.

On the other hand, the authors of Ref. [13], in their studies, give greater attention to organizational and human aspects. They argue that the place of these dimensions, as well as implementation challenges such as coordination mechanisms, data governance, or appropriation, in research on DT-ITS remains limited and superficially addressed. Indeed, the still relatively secondary place given to organizational, governance-related, and human dimensions emerges as a fourth imbalance. This latter remains particularly important in the context of resilience, given the contribution of digital twins to resilience and the dependence of these implementation dimensions.

In addition, Ref. [56] suggests linking digital twins to risk management and resilience within an integrated framework. These latter authors reveal that several studies addressing monitoring and simulation from a prospective perspective do not examine, with the same level of depth, the impact of digital twins on adaptation, recovery, or continuity after disruption or dysfunction. Figure 11 synthesizes the main imbalances that still limit the scientific consolidation of the link between digital twins, transport systems, and resilience.

3.3. Digital Twin Functions and Their Contribution to Resilience Phases in Intelligent Transport Systems

For analytical clarity, the functions discussed in this section are not treated as resilience mechanisms in themselves. Rather, they are interpreted as digital twin-enabled functions that may support specific phases or dimensions of resilience when they are explicitly connected to disruption-related contexts or outcomes. Real-time monitoring, prediction, simulation, optimization, decision support, recovery, and learning can contribute differently to preparedness, absorption, response, recovery, adaptation, and transformation.

Table 1. Mapping of digital twin functions onto resilience phases and dimensions.

Digital Twin Function	Main Resilience Phase (s) Supported	Main Resilience Dimension (s)	Resilience Contribution
Real-time monitoring and visibility	Preparedness, absorption, response	Robustness, situational awareness, response capacity	Improves situational awareness, early detection of anomalies, identification of critical points, and rapid understanding of system degradation.
Prediction and anticipation	Preparedness, absorption	Robustness, anticipatory capacity, absorption capacity	Supports risk forecasting, weak-signal detection, disruption anticipation, and estimation of probable system trajectories before or during disturbances.
Scenario simulation and decision support	Preparedness, response	Redundancy, response alternatives, decision robustness	Enables testing of alternative actions, comparison of intervention scenarios, prioritization of decisions, and reduction in uncertainty before operational intervention.
Dynamic optimization and adaptive control	Response, adaptation	Adaptability, operational reconfiguration	Supports resource reallocation, adaptive rerouting, scheduling adjustment, traffic control, and operational reconfiguration under changing conditions.
Service continuity and recovery support	Absorption, response, recovery	Rapid recovery, service continuity, absorption	Helps maintain minimum service levels, contain disruption effects, support restoration pathways, and reduce recovery time.
Post-disruption learning and model recalibration	Recovery, adaptation, transformation	Adaptability, transformation, long-term resilience	Enables learning from previous incidents, recalibration of thresholds and models, improvement of intervention protocols, and long-term system transformation.

summarizes this functional mapping and provides the analytical basis for the thematic synthesis that follows.

A further analytical distinction was made between routine operational performance and disturbance-related resilience improvement. Improvements in traffic flow, congestion reduction, energy efficiency, passenger experience, or resource allocation under normal operating conditions were not interpreted as evidence of resilience improvement in themselves. They were treated as routine performance benefits unless the study explicitly linked them to disruptive conditions, abnormal operating states, incident management, service continuity, robustness, recovery, adaptive rerouting, or system reconfiguration. Accordingly, this review analyzes normal operational optimization separately from resilience-related outcomes associated with preparedness, absorption, response, recovery, adaptation, and transformation.

This additional mapping clarifies how digital twin functions support not only chronological resilience phases, but also core resilience dimensions such as robustness, redundancy, recovery speed, adaptability, service continuity, and long-term resilience. It also shows that the same digital twin function may support more than one resilience phase. For example, real-time monitoring contributes to preparedness when it improves system awareness before disruption, to absorption when it helps detect degradation during disruption, and to response when it supports rapid operational decisions. The following subsections therefore analyze each function according to the resilience phase or dimension it primarily supports.

3.3.1. Real-Time Visibility and Monitoring

In terms of contribution, the literature addressing digital twins in ITS often evokes their impact on the improvement of the visibility of these systems. Ref. [7] dissects this dimension and explicitly demonstrates that digital twins allow, through various informational flows and across several digital environments, to orchestrate a set of transport models in a synchronized manner. This evolving and updated perception remains decisive for the improvement of resilience because it offers the system the possibility to anticipate and adapt its responses and its reactions.

The authors of Ref. [13] deepen their studies about this variable and demonstrate that visibility is not limited to the accumulation of data, but it offers the system the capacity to function, to operate fluctuations and movements as well as to detect tension signals in order to monitor dynamics and evolutions. The same authors argue that the applications related to digital twins in ITS, based on the state of the system, traffic, and the generation of operational knowledge by all stakeholders, closely depend on a functional orientation and a usage-oriented approach.

In addition, Ref. [38] highlights this dimension, in the context of an interurban highway, in a more tangible manner and underlines that digital twins are continuously fed by sensors and V2X communication interactions. Consequently, this allows generating a corridor state and following its progression and supporting operational decisions through a fine perspective of a synchronized state. By articulating simulation and collected data, Ref. [8] corroborates this observation and virtualizes the dynamics of highway traffic. The latter insists on evolutionary synchronization between real traffic and its simulated framework, and underlines that this matching converts data into operational visibility. At a more localized scale, Ref. [55] ensures, through their digital twins platform for intersections, an improvement of visibility, which, in turn, contributes to the strengthening of the operators’ capacity to rapidly identify degraded and critical situations.

In the context of transport infrastructures, Ref. [10] confirms that digital twins are becoming increasingly deployed in order to optimize state monitoring, lifecycle management, predictive maintenance, and functional steering. These authors, by analyzing this dimension at the level of roads, bridges, and tunnels, underline that visibility extends to cover the physical state of assets and their performances, as well as their deteriorations and their operational performances. This confirms that visibility goes beyond the simple monitoring of traffic flows to touch resilience because dysfunctions and disturbances can result both from a traffic incident and from a weakness of the infrastructure.

Similarly, Ref. [40], by defining the mobility digital twin, combines, within an integral framework, the human being, the vehicle, and traffic in a cloud–edge–device environment. These authors extend the visibility generated by digital twins to go beyond, only, infrastructure and informational flows. They argue that monitoring becomes multi-level and confirm that it can encompass behaviors and interactions of all stakeholders.

Indeed, this reinforced visibility facilitates the identification of critical points and offers an earlier detection of dysfunctions. Moreover, it also improves coordination among all stakeholders at the different levels of management by presenting information that is transmissible, more coherent, and more synchronized. Ref. [9] shows, through their AI-assisted digital twin framework, that the combination of multi-source monitoring, anomaly detection, and congestion forecasting guarantees support in terms of intelligent transport management. On their part, Ref. [56] emphasizes the combination of data analysis and monitoring in order to improve risk management and, consequently, they position visibility as a fundamental prerequisite of resilience.

3.3.2. Prediction and Anticipation of Disruptions

Beyond synchronized visibility, digital twins offer projections on probable evolutions before the occurrence of dysfunctions and their repercussions. Ref. [10] explicitly shows that this forecasting capacity is positioned as one of the most significant shifts in the field. Ref. [7] reveals that the study of digital twins applied to transport has moved, in the most recent studies, toward forecasting and decision-support functions to the detriment of simulation previously. This mutation, or even evolution, highlights that observation and synchronized monitoring remain insufficient to improve the level of resilience. Ref. [56] insists on the support of digital twins for anticipatory and proactive steering and consequently for resilience. These latter authors link digital twins with resilience through three main illustrations, namely: the detection of weak signals, the estimation of trajectories as well as the preparation of the reaction before the appearance of dysfunctions and ruptures.

However, this dimension requires an integration and a transformation of data coming from several sources into predictive information capable of offering probable scenarios in the future. Similarly, Ref. [13] emphasizes the importance of DT-ITS functions such as advanced analytics, alert generation, and decision support. Applied contributions clearly illustrate this orientation by showing that digital twins can aggregate heterogeneous data, then use artificial intelligence models to identify abnormal situations before they become fully critical [10]. These authors, through an AI-assisted digital twin framework for intelligent transport, argue that anticipation is not part of an abstract presentation and a speculative vision, but requires an operational capacity for early detection.

In connected traffic systems, Ref. [52] underlines that the contribution of digital twins resides mainly in the combination of modeling and forecasting. In addition, Ref. [63] shows that digital twins contribute to the improvement of the system’s reactivity, and this through the analysis and reorganization of real-time traffic informational flows. Furthermore, in rail transport, anticipation broadens the field of impact and becomes more explicit, and this by linking digital twins to the forecasting of passenger movements as well as to the evaluation of the repercussions of service interruptions and failures. Ref. [64], through their model combining deep learning and simulation, managed to predict the number of passengers in stations as well as to evaluate the impact of service disruptions.

In the port context, the anticipatory approach intervenes through the management of uncertainty and planning. The contribution in this context, a complex port environment, is to improve the adjustment of operations and not to freeze or to rely only on a single alternative [65]. Ref. [60] argues that digital twins serve, increasingly, to simulate dangerous situations and to support a proactive approach to maintenance and safety. Moreover, they affirm that the orientation toward optimization, forecasting and good steering of complex systems and, consequently, this makes anticipation a structuring function. In addition, Ref. [66] evokes an additional illustration of this function and argues that it allows, in the first place, early preparation and a reduction in operational and functional uncertainty. Subsequently, it allows prioritizing intervention priorities, and this through the identification of the active zones or flows most likely to become critical.

Compared with standalone data-driven machine learning models used for mobility prediction, disruption analysis, or vulnerability assessment, DT-based forecasting is not limited to point prediction. Previous work on machine learning techniques and discrete choice models in mode choice analysis shows that data-driven approaches already provide useful baselines for modeling mobility behavior and travel decisions [67]. The added value of DT-based forecasting lies in the coupling of such predictive models with a structured representation of assets, flows, operational states, and scenario-testing capabilities. This coupling can support the assessment of cascading effects, alternative interventions, and feedback to operational control systems [68]. However, the current corpus does not provide sufficient comparative evidence to claim a systematic predictive superiority of DT-based approaches over ML-only baselines. Future research should therefore compare DT-based forecasting with machine learning and discrete-choice baselines under equivalent disruption scenarios and performance indicators.

3.3.3. Scenario Simulation and Decision Support

By examining digital twins in transport planning, Ref. [7] indicates that simulation remains the most mobilized modeling modality and the most decisive mechanism allowing digital twins to strengthen resilience. This functionality, central for resilience, offers a capacity for experimental reasoning to test, virtually, the different alternatives before any concrete intervention. Ref. [56] corroborates this observation and affirms that digital twins allow decision-makers to explore scenarios and evaluate their probable repercussions. Thus, this comparison of action options broadens the contribution of simulation and offers an integral estimation as well as a reduction in uncertainty before any implementation.

Ref. [10] confirms this functionality and explicitly shows that the service layer of the digital twin, transport infrastructures, combines decision-support tools. Indeed, between responses to accidents, urban planning, traffic monitoring, and congestion forecasting, decisions cannot be planned or taken outside digital twins. In addition, applied contributions further point to this mechanism and show that the virtualization of road flows, based on an evolving and continuous synergy, simulates change and clarifies decisions [8]. Ref. [38], by relying on feeding data from sensors and V2X communications, corroborates this observation through their highway digital twins designed with the objective of assisting real-time decision-making, and this in an interurban corridor.

In the port and terminal context, simulation is often associated with decisions. Explicitly, Ref. [69] develops a digital twin-based decision support approach in AGV planning. This virtual model, by reflecting real operations, mobilizes mathematical programming and resolution algorithms in order to suggest, test, and validate alternatives and planning decisions. The contribution of this function appears with even more clarity when [69] underlines that these choices and these steering and planning decisions remain adaptable to the conditions and specificities of each terminal.

Ref. [70] supports this approach and argues that this approach is also observable in the context of replanning and port resilience. These authors developed a decision support system based on a resilience analysis based on digital twinning. Their model presents digital twins as an instrument of arbitration, or even as a decision tool, between several intervention trajectories. Furthermore, the literature also signals that the local reconfiguration of steering mechanisms and control systems can be significantly supported by simulation.

In the same vein, Ref. [55] confirms that the incorporation of adaptive simulations in intersection digital twins allows consolidating operational efficiency. These latter authors observe a significant improvement, at the level of operational performance, compared to static or preprogrammed and predefined controls. In addition, with their DTTF-Sim system, the authors of Ref. [71] introduce the temporal dimension and reveal that a digital twin-based simulation system can generate and contribute to effective traffic situations over an extended horizon for continuous tests.

3.3.4. Operational Adaptation and Dynamic Optimization

Going beyond the simple observation of the network, digital twins constitute a fourth significant dimension of the operational adaptation of the system in operation. This contribution refers to the system’s capacity to reorganize its resources, to adapt its operating devices as well as to restructure and reconfigure its processes, and this within a dynamic and evolving framework [10].

In our context, this capacity for adaptation and adjustment often takes the configuration of a dynamic and evolving improvement. Ref. [13] corroborates this observation and underlines that contributions examining digital twins in intelligent transport systems attribute a structuring position to the functionalities and structures capable of converting data and models into operational actions and steering measures. Ref. [72] particularly illustrates this mechanism by justifying the use of a digital twin-driven proactive–reactive scheduling framework in order to cope with recurrent hazards in the integrated scheduling of port resources. Their study reveals that digital twins could support and strengthen continuous reconfiguration, especially in an unstable and multidimensional operational context. Furthermore, these authors explicitly show that digital twins can support hybrid rescheduling in an intermodal terminal in order to improve coordination and performance.

An intelligent traffic system, within an artificial intelligence-assisted digital twin framework, was proposed by [9]. This framework is based on the combination of infrastructure maps, traffic histories, telematics data, and video flows within the same platform, then mobilizes AI models for anomaly detection, risk prediction, and decision-making. Consequently, optimization becomes intimately linked to field reality and intervenes as a reactive, evolving function activated from a system state that is continuously updated.

In road networks, operational adaptation further broadens to concern the steering of the infrastructure itself. Ref. [73] recalls that no reference model has been, until now, unanimously accepted and adopted. However, these authors underline that recent contributions seek to structure the essential operational levels for integrated and evolving steering. This adaptation, in terms of resilience, contributes to the reduction in the system’s rigidity and to the mitigation of the cumulative repercussions triggered in cascade. Refs. [10,72] confirm that this perspective makes decisions and remains reactive, flexible, and adjustable to changes in context, optimizing resource allocation and capacity distribution.

3.3.5. Recovery, Service Continuity and Learning

Another mechanism is introduced within the framework of strengthening the resilience of intelligent transport systems and concerns system recovery, service continuity, and, especially, capitalization from experienced disturbances and observed shocks. Ref. [74] underlines that, in addition to the capacity to resist or adapt during the shock, the capacity to adjust and reconfigure itself to an appropriate functional threshold after having undergone the dysfunction constitutes an essential mechanism of resilience. Ref. [75] maintains this vision of functional continuity of digital twins and underlines that the preservation of a satisfactory functioning of the service, after having undergone interruptions, also depends on the system’s faculty to contain and master the repercussions and to project a pathway back to functioning.

Applied contributions indicate that this dynamic of restoration and recovery appears especially when digital twins are articulated with replanning and reconfiguration devices. Ref. [65], by studying intermodal hubs, explicitly shows that coordination between trucks and rail cranes for several container flows can be significantly improved. These authors support this observation through the support offered by digital twins to rescheduling enhanced by a hybrid strategy. At the level of ports, from the angle of operational resilience, recovery is not the exception. Ref. [70] develops a decision support system, based on an evaluation of the consequences of critical incidents, which allows examining pre-established restoration responses, and this comes from a decision assistant in the form of a digital twin.

In public transport networks, service maintenance remains addressed through the strengthening and continuous evolution of operations [13]. Thus, even if the contributions addressing recovery and post-crisis reconfiguration in an explicit manner still remain relatively limited, some applied studies reveal that digital twins could contribute to containing interruptions and dysfunctions as well as to limiting their repercussions.

In addition, the relationship with capitalization and continuous learning intervenes from the moment when digital twins are mobilized as an analytical memory and a learning base. Ref. [10] indicates, in transport infrastructures, that digital twins offer informational continuity which potentially allows improving and strengthening models, recalibrating and readjusting alert thresholds and levels as well as refining action and intervention protocols. These same authors defend this observation through the capacity of digital twins to preserve data on previous incidents, on the applied interventions, on the achieved and measured performances as well as on the dynamics of resumption and recovery. Consequently, these digital twins increasingly take charge of the whole path of the life cycle, and this comes from the state monitoring of maintenance and operational steering.

3.3.6. Comparative Synthesis of Evidence, Boundary Conditions and Strength of Support

Taken together, the retained studies show a broad agreement on the idea that digital twins can support resilience when they improve system visibility, anticipation, scenario exploration, decision support, operational adaptation, and post-disruption learning. However, the strength of this agreement varies across mechanisms. The strongest convergence concerns real-time visibility and monitoring, which appear as foundational functions across road, rail, port, infrastructure, and urban mobility studies. Prediction, simulation, and optimization are also frequently reported, but their contribution to resilience is more conditional because many studies demonstrate model-based potential rather than empirically measured resilience outcomes.

The evidence is therefore not homogeneous. Conceptual and review-based studies mainly support the identification of architectures, functions, terminology issues, and research gaps. Simulation-based studies provide stronger support for potential anticipation, testing, and optimization capabilities, but often remain limited by assumptions, simplified disruption scenarios, or controlled experimental settings. Case studies, prototypes, and real-world deployments provide stronger evidence when they assess operational outcomes such as incident response, service continuity, recovery, adaptive rerouting, or system reconfiguration. As a result, the review interprets monitoring, prediction, simulation, optimization, and decision support as resilience-enabling functions unless the study explicitly demonstrates resilience outcomes under disruptive or abnormal operating conditions.

Several inconsistencies also emerge across the corpus. First, studies do not always use the term digital twin with the same level of technological maturity, which creates differences between digital models, digital shadows, synchronized digital twins, and closed-loop digital twins. Second, the transport domains are unevenly represented. Evidence is more developed for infrastructures, road corridors, ports, rail systems, and traffic control environments than for open multimodal mobility systems or large-scale urban networks. Third, the type of disruption considered varies considerably across studies, ranging from congestion, incidents, failures, and scheduling uncertainty to infrastructure degradation, cyber–physical risks, or service interruptions. These differences limit direct comparability across studies.

The contribution of digital twins to resilience is therefore subject to several boundary conditions. It depends on data quality, real-time or near-real-time synchronization, interoperability, model reliability, the maturity of the digital twin architecture, and the capacity of operators to translate digital outputs into coordinated action. In this respect, the review does not interpret digital twins as automatically producing resilience. Rather, they are understood as conditional socio-technical devices whose resilience value depends on the combination of technological maturity, validation context, operational integration, and governance capacity. Applying the predefined coding protocol to the analytical corpus, only 10 studies (16.4%) reported directly verified resilience-related outcomes. Most studies either validated enabling functions without measuring direct resilience outcomes or provided conceptual, architectural, or review-level evidence. The evidence base was dominated by simulation, optimization, and computational studies, while the single real-world operational deployment did not assess resilience performance during an actual disruption. These results are summarized in Figure 12 and

Table 2. Evidence categories in the analytical corpus (n = 61).

Evidence Category	Number	Share
Direct resilience-outcome evidence	10	16.4%
Validated enabling-function evidence	36	59.0%
Conceptual/review evidence	15	24.6%
Total	61	100%

.

The analytical corpus is listed in Table A2, while the coding rules and category definitions are provided in Table A6.

3.3.7. Stratified Interpretation by Transport Mode, System Scale and Disruption Type

Because the retained studies cover heterogeneous transport domains, their findings cannot be generalized uniformly across all intelligent transport systems. A stratified interpretation was therefore used to distinguish the contribution of digital twins according to transport mode, system scale, and disruption type. The detailed stratified matrix is reported in

Table A7. Stratified interpretation of digital twin contributions by transport mode, system scale, and disruption type.

Transport Domain	Typical System Scale	Main Disruption or Risk Context	Main Digital Twin Functions	Main Resilience Contribution	Boundary of Generalization
Road corridors and urban traffic	Intersection, corridor, road network	Congestion, incidents, abnormal traffic states, signal failures	Real-time monitoring, prediction, adaptive control, rerouting	Preparedness, response, adaptation	Transferability depends on sensor coverage, V2X availability, traffic data quality, and local control architecture.
Railways and metro systems	Asset, station, line, network	Passenger flow disruption, rolling-stock anomalies, infrastructure degradation, service interruption	Condition monitoring, passenger flow prediction, simulation, maintenance support	Preparedness, absorption, recovery	Findings are more transferable to closed or semi-closed systems with structured operations and reliable asset data.
Ports and intermodal terminals	Terminal, port system, logistics node	Equipment failure, scheduling uncertainty, congestion, operational disruption	Simulation, scheduling optimization, decision support, resource reallocation	Response, recovery, adaptation	Generalization depends on terminal layout, operational rules, equipment configuration, and multi-actor coordination.
Bridges and transport infrastructures	Asset, infrastructure network, lifecycle system	Structural degradation, maintenance needs, extreme loads, infrastructure failure	Asset monitoring, lifecycle modeling, predictive maintenance	Preparedness, absorption, recovery	Evidence is mainly asset-specific and cannot be directly generalized to full mobility-system resilience.
Urban mobility and public transport	Multimodal network, city scale	Demand fluctuation, service interruption, modal disruption, urban mobility imbalance	Mobility simulation, real-time visibility, decision support, scenario analysis	Preparedness, adaptation, transformation	Transferability is limited by governance capacity, interoperability, data integration, and the openness of urban mobility systems.

.

Overall, asset-based studies, such as bridges and infrastructure management, mainly support preparedness, absorption, and maintenance-oriented recovery. Traffic and road-network studies more frequently address monitoring, response, adaptive control, and rerouting. Port and intermodal terminal studies provide stronger evidence on scheduling, operational reconfiguration, and recovery under uncertainty. Urban mobility and public transport studies offer broader perspectives on adaptation and transformation, but their generalization remains more limited because of higher system openness, institutional complexity, and data interoperability constraints. Consequently, the findings of this review should be understood as conditional and stratified rather than universally transferable across all transport modes and scales.

3.4. Enabling and Constraining Conditions for DT-ITS Resilience

The contribution of digital twins to ITS resilience depends on a set of enabling and constraining conditions. Based on the reviewed corpus, these conditions can be grouped into three interrelated dimensions: technical, organizational, and institutional.

First, technical conditions include data quality, temporal granularity, sensor coverage, interoperability, synchronization, model validity, computational scalability, cybersecurity, and the capacity to represent cascading failures. Cybersecurity also appears as a critical technical condition. Because DT-ITS architectures connect sensors, vehicles, infrastructures, communication layers, data platforms, simulation models, and decision-support tools, they may expand the cyber–physical attack surface of intelligent transport systems. Resilience assessment should therefore consider not only physical disruptions, but also data manipulation, communication failures, software malfunction, model desynchronization, and cascading failures across connected subsystems. Without these conditions, digital twins may remain descriptive representations rather than operational instruments for resilience-oriented decision-making.

Second, organizational conditions refer to the capacity of transport operators, infrastructure managers, emergency services, and mobility authorities to interpret digital outputs and transform them into coordinated action. These conditions include operator expertise, emergency protocols, decision rights, data-sharing arrangements, inter-agency coordination, and the integration of digital twin outputs into operational routines.

Third, institutional and governance conditions concern the broader smart city environment in which DT-ITS are deployed. These include data governance, privacy rules, regulatory interoperability, procurement capacity, accountability arrangements, and alignment with urban resilience strategies. These conditions are particularly important in smart cities, where multimodal mobility systems involve multiple actors, heterogeneous infrastructures, and overlapping responsibilities.

Operationally, these enabling and constraining conditions can be validated through simulation-informed and graph-based resilience assessments. For example, Ref. [76] shows how traffic-simulation models can be used to assess road-network vulnerability under flood-induced closure scenarios by mobilizing indicators such as average control delay, travel time, volume-to-capacity ratios, and vulnerability indices. In the context of DT-ITS, similar approaches can help transform qualitative conditions, such as data quality, interoperability, synchronization, model reliability, cybersecurity, and operational coordination, into measurable thresholds. Traffic-simulation models, road-closure scenarios, weighted graph representations, vulnerability indices, throughput-retention measures, volume-to-capacity ratios, and recovery-time indicators can therefore help assess whether a DT-ITS can maintain service, reconfigure flows, prioritize interventions, or accelerate recovery under actual physical disruptions.

4. Discussion

4.1. Evidence Limitations, Theoretical Contributions and Practical Implications

The findings of this review should be interpreted with caution because the retained studies do not provide the same level of evidence. While case studies and real-world deployments offer stronger support when they assess disruption-related outcomes, many contributions remain conceptual, architectural, simulation-based, or prototype-oriented. Therefore, this review distinguishes between digital twin functions that may support resilience and empirically verified resilience benefits.

An additional limitation concerns the exclusion of conference papers and other non-journal publications. This criterion improved the homogeneity of the reviewed sources. However, it may have led to the underrepresentation of emerging methods, prototype systems, and early-stage engineering applications. Such contributions are often first disseminated through high-quality conferences in intelligent transportation, computer science, and the Internet of Vehicles. Consequently, the review may offer a more mature but less comprehensive representation of rapidly evolving technological developments. Future reviews could address this limitation by conducting a targeted assessment of high-quality conference proceedings.

The restriction to the 2020–2025 period also limits the historical and temporal coverage of the review. Although foundational pre-2020 studies were considered for conceptual framing, they were not systematically included in the analytical corpus. Similarly, studies published during 2026 were excluded to avoid partial-year and indexing bias. The findings should therefore be interpreted as a synthesis of the contemporary consolidation phase of DT-ITS research rather than as a complete historical account of the field.

This review contributes to the literature by clarifying the relationship between digital twins, intelligent transport systems, and resilience. First, it distinguishes digital twin functions from resilience mechanisms and maps these functions onto resilience phases such as preparedness, absorption, response, recovery, adaptation, and transformation. Second, it differentiates between potential resilience-supporting functions and empirically verified resilience benefits. Third, it highlights the importance of digital twin maturity by distinguishing digital models, digital shadows, synchronized digital twins, and advanced or closed-loop digital twins. Finally, the review proposes a stratified interpretation of evidence according to transport mode, system scale, and disruption type, thereby limiting excessive generalization across heterogeneous transport contexts.

For practitioners, infrastructure managers, mobility operators, and public authorities, the results suggest that digital twins can support resilience-oriented decision-making only when several implementation conditions are met. These include high-quality and timely data, interoperability between systems, reliable models, continuous or near-real-time synchronization, cybersecurity, and the ability of actors to transform digital outputs into coordinated operational responses. Digital twins should therefore be designed not only as modeling or optimization tools, but also as decision-support infrastructures embedded in governance, coordination, and emergency-response processes.

4.2. Smart City Implications, Urban Governance, and Future Research Priorities

The smart city implications of DT-ITS are not limited to the technological deployment of sensors, platforms, or simulation tools. In dense urban environments, digital twins operate within a broader governance architecture linking traffic control centers, infrastructure managers, public transport operators, emergency services, data platforms, and local authorities. Their resilience value depends on the ability of this urban ecosystem to combine multimodal data, test disruption scenarios, prioritize interventions, and coordinate responses across institutions.

The corpus also suggests that prediction and optimization functions do not operate in the same way in sensor-rich urban environments and in interurban or asset-based systems. In smart cities, higher data density can improve situational awareness, prediction, and scenario exploration. However, it also increases interoperability requirements, privacy constraints, cybersecurity exposure, and the need for cross-institutional coordination. By contrast, interurban corridors or asset-based infrastructures often provide clearer operational boundaries and more controllable feedback loops, but their findings are less directly transferable to open multimodal urban mobility systems. Consequently, DT-ITS should be understood not only as technical modeling tools, but also as smart city governance infrastructures. Their contribution to resilience depends on the capacity of urban actors to transform digital intelligence into coordinated operational action, emergency response, mobility regulation, and long-term planning.

Future research should move beyond conceptual and simulation-based demonstrations by assessing digital twins in real disruptive or abnormal operating conditions. More empirical studies are needed to measure direct resilience indicators, such as robustness, absorption capacity, service continuity, recovery time, adaptive rerouting performance, system reconfiguration, and post-disruption learning. Comparative studies across transport modes, urban contexts, infrastructure types, and levels of digital twin maturity would also help clarify the transferability of findings. Finally, future work should pay greater attention to organizational appropriation, governance, cybersecurity, interoperability, and the human role in transforming digital intelligence into resilience-oriented action. Future research should also explore cybersecurity-by-design and failure-propagation modeling in DT-ITS, including the use of machine learning surrogate models, graph-based cascade simulations, and scenario-based stress testing to evaluate how disruptions spread across cyber–physical transport architectures. More specifically, the corpus reveals three scaling barriers. The first concerns the transition from asset-level or corridor-level digital twins to multimodal urban networks. The second concerns transferability across transport modes, infrastructure types, and governance contexts. The third concerns the lack of comparable resilience performance indicators across studies. Future validation should therefore rely on measurable indicators such as network-level throughput retention during disruption, service-level degradation, recovery-time distributions, adaptive rerouting performance, and post-incident restoration trajectories.

5. Conclusions

Through a systematic literature review, conducted according to the recommendations of PRISMA 2020 and PRISMA-S, this study aimed to analyze to what extent, through which functions, under which conditions, and with what level of evidence digital twins can support the resilience of intelligent transport systems. The investigation allowed us to structure a dispersed and still fragmented field, positioned between intelligent transport systems, digital twin technologies, and the resilience of these systems.

The findings reveal that digital twins evolve and reconfigure themselves towards integrated digital devices and connected technological layers, service-oriented, capable of supporting functions that may contribute to resilience, including real-time visibility, disruption prediction, scenario simulation, dynamic optimization, decision support, service continuity, and post-disruption learning. No longer restricted to a function of virtual representation of infrastructures, flows, or transport assets, their contribution to resilience thus rests on their potential to progressively convert heterogeneous data into operational intelligence, then into adaptive responses to incidents, congestion, service disruptions, or infrastructural vulnerabilities.

The scientific contribution of this investigation lies in the functional clarification of digital twins as intermediate devices between the informational resources of ITS and the resilience capacities of transport systems. Moreover, the study presents an articulated and integrative reading confirming that digital twins do not systematically produce resilience, but make it possible when they are combined with reliable data, robust models, continuous synchronization and decision-making mechanisms that are effectively mobilizable by the concerned actors.

At the practical and institutional level, the results highlight the interest of digital twins for public authorities, infrastructure managers, mobility operators and decision-makers engaged in the development of smart cities. These devices can support a more proactive governance of urban mobility, provided that they are integrated into interoperable, secure and action-oriented architectures. They must therefore be designed not only as technical modeling tools, but also as instruments for coordination, anticipation and steering collective responses to disruptions.

In terms of limitations, the field remains marked by conceptual instability between digital model, digital shadow, and digital twin, uneven sectoral coverage, limited large-scale empirical validation, and insufficient consideration of human, organizational, and institutional dimensions. The conclusions should also be interpreted in light of the heterogeneous quality and evidentiary strength of the retained studies, as well as the absence of a formal inter-reviewer agreement statistic such as Cohen’s kappa, although this limitation was mitigated through predefined coding rules, independent coding, full-text rechecking, and consensus-based resolution of disagreements.

Overall, the main lesson of this review is not that digital twins constitute a universal solution to ITS fragility, but that their resilience value emerges from the alignment between technological maturity, disruption-oriented validation, operational integration, and governance capacity. The field therefore needs to move from demonstrating digital twin feasibility toward measuring when, where, and under which conditions these systems actually preserve service, absorb shocks, support recovery, and enable adaptation.

Digital twins therefore do not automatically resolve the fragility of intelligent transport systems. They may become levers of resilience when digital intelligence is effectively transformed into rapid, coordinated, and sustainable decisions in the service of smart cities.