The global maritime transportation sector, encompassing ports, shipping companies, and interconnected global supply chains, is undergoing profound structural shifts. Historically driven by physical asset expansion and traditional economies of scale (Haralambides 2019; Stopford 2009), the industry now operates within an ecosystem defined by extreme market volatility, shifting geopolitical realities, stringent environmental regulations, and sudden systemic shocks (UNCTAD 2023). To maintain resilience, commercial viability, and operational sustainability, the sector has increasingly turned toward digital transformation. Technologies that were once considered experimental, such as Artificial Intelligence (AI), Machine Learning (ML), Digital Twins, Robotic Process Automation (RPA), and advanced geospatial network analytics, inter alia, are now actively reshaping the strategic and operational fabric of maritime logistics (Notteboom et al. 2026).
However, modern maritime digitization is no longer just about adopting isolated software solutions. It requires fundamental restructuring of big data management, predictive diagnostics, and workflow automation. As shipping companies push toward decarbonization, green fuel selection, and operational efficiency, the deployment of AI and ML must confront the harsh realities of maritime environments, such as fragmented connectivity, strict data privacy, and the high carbon footprint associated with heavy computational models (Kalafatelis et al. 2025c, d, a). This has led to a shift from fixed centralized architectures to more flexible decentralized, resource-aware frameworks, such as serverless gossip and federated learning, which allow vessels to collaboratively train predictive models without transmitting massive volumes of raw data over expensive satellite links (Kalafatelis et al. 2026). At the same time, the operational side of maritime logistics is experiencing a major transition toward automated decision-making (Fruth and Teuteberg 2017). By combining these decentralized intelligent systems with Robotic Process Automation (RPA), shipping companies can bridge the gap between real-time data collection and automated back-office execution, creating a highly responsive and resilient supply chain ecosystem.
The necessity for rigorous research into these technologies stems from several critical gaps between theoretical computational capability and empirical maritime realities, including:
1.1 Inadequacy of traditional frameworks
Traditional linear models and conventional time-series forecasting frameworks increasingly fail to capture nonlinear dynamics, structural breaks, and volatile spikes inherent to modern shipping operations, such as fluctuating freight rates, sudden corporate financial distress events, variations in vessel sailing speeds and berthing times, and fluctuations in port throughput, inter alia. Furthermore, conventional predictive maintenance strategies rely on rigid, scheduled overhauls or simplistic threshold alerts that overlook complex thermodynamic variances in marine machinery (Wang et al 2024). For instance, predicting the critical anomalies of primary sub-systems, such as main engine Exhaust Gas Temperature (EGT), requires modeling highly intricate, multi-variable dependencies (Kalafatelis et al. 2025c, d, a).
The unprecedented volatility across the maritime value chain demands a paradigm shift toward advanced non-linear methodologies that can adapt to rapid market shocks and evolving regulatory frameworks, such as the EU Emissions Trading System (ETS) and the International Maritime Organization’s (IMO) emissions regulations. Within this context, understanding carbon emission price dynamics and developing robust hedging strategies becomes vital for risk management in shipping markets (Syriopoulos et al. 2023). To address these non-linear complexities, the sector is moving toward hybrid architectures that merge deep learning with explainable AI. By combining Bidirectional Long Short-Term Memory (BiLSTM) networks, attention mechanisms, and Kolmogorov-Arnold Networks (KAN), modern frameworks can capture long-term temporal dependencies while maintaining the mathematical transparency required by engineers and regulators alike (Kalafatelis et al. 2025c, d, a, ). The penetration of AI in solving these operational inefficiencies is manifested through these hybrid deep learning architectures that process complex thermodynamic variables, effectively shifting the industry from reactive maintenance to highly accurate, explainable anomaly detection.
1.2 Socio-technical and infrastructure bottlenecks
While broad concepts, such as “smart ports”, are discussed extensively, their widespread deployment is constrained by a lack of common definitions and understanding and severe integration barriers (Haralambides et al 2026). As Zeng et al. (2020) demonstrate, the adoption of digital platforms and advanced technologies across the maritime supply chain is frequently bottlenecked by multi-stakeholder misalignment, data-sharing anxieties, resource scarcity, upfront investment requirements, and high integration complexities with legacy systems.
Beyond port infrastructure, these bottlenecks extend to the vessels themselves, where communication constraints and the computational overhead of standard AI models create significant friction. Centralized data processing forces ships to transmit massive raw datasets via costly satellite communication, creating severe bandwidth strain and risking data privacy breaches. To overcome this, recent paradigms leverage decentralized, carbon-aware gossip orchestration protocols (Kalafatelis et al. 2026). By utilizing a serverless control plane to manage localized training, vessels can share intellectual model updates rather than raw data, matching the constrained communication capabilities of a dynamic marine network.
In terms of infrastructure bottlenecks, this specific layer of AI penetration circumvents connectivity barriers through decentralized edge intelligence, enabling individual vessels to perform localized computational training without overloading maritime satellite communication links. Furthermore, AI-driven predictive models are increasingly being employed to anticipate weather-related disruptions and environmental anomalies that directly affect maritime operations. By leveraging machine learning techniques to analyze historical meteorological records, oceanographic conditions, and real-time sensor data, these models can forecast evolving wind patterns, extreme weather events, and navigational risks, enabling shipping operators to proactively adjust routing decisions, optimize fuel consumption, and mitigate supply chain disruptions (Arora et al 2026). Such predictive intelligence not only enhances operational resilience but also supports safer and more sustainable maritime logistics by reducing uncertainty in voyage planning and improving the responsiveness of the overall supply chain. Consequently, there is an urgent need to explore immediate, non-invasive, and cost-effective digital stepping stones, such as software automation and localized algorithms, that can optimize workflows without requiring cost-prohibitive structural overhauls.
1.3 The need for dynamic decision support
Modern port and shipping operations generate vast amounts of real-time data that remain underutilized. To improve operational efficiency and mitigate systemic disruptions, maritime nodes require advanced analytical frameworks that offer real-time monitoring, as well as predictive and optimization capabilities. In this context, frameworks such as digital twinning have emerged as critical decision support systems necessary for maintaining port resilience and optimizing resource allocation under deep uncertainty (Zhou et al. 2021). Machine learning models have been tailor-developed and applied to enhance the predictive abilities of ports, leading to greater efficiency in container handling and reductions in equipment movements and fuel consumption (Xie et al. 2025). Crucially, the success of these analytical systems depends on bridging the gap between back-office manual administration and front-line data execution. Integrating Robotic Process Automation (RPA) alongside Internet of Things (IoT) frameworks allows shipping companies to seamlessly automate the ingestion, verification, and dispatch of operational logs, regulatory compliance data, and shipping documents without needing to rewrite legacy software architectures. Moreover, the strategic penetration of AI effectively bridges macro-level management with micro-level office workflows by pairing automation with cognitive algorithms, easing the socio-technical transition for the maritime workforce.
Confronting these challenges, the objective of this MEL Special Issue is to compile cutting-edge, empirically validated research that moves beyond theoretical speculation to demonstrate the pragmatic application of new technologies across maritime networks. By consolidating nine pioneering studies, this issue aims to provide maritime economists, port authorities, logisticians, and policymakers with actionable methodologies, comparative analytical tools, and empirical evidence to navigate the complexities of digital transformation of the maritime cluster.
To provide a cohesive narrative, the contributions in this MEL Special Issue are structured into three interrelated thematic parts:
-
1.
Port Digitalization, Artificial Intelligence, and Digital Twins
-
2.
Robotic Process Automation (RPA) in Maritime Operations and Shipping Firms
-
3.
Advanced Machine Learning and Analytics for Forecasting and Risk Assessment.
1.3.1 Thematic part I: port digitalization, artificial intelligence, and digital twins
Ports and inland waterways serve as the primary critical infrastructure nodes within global supply chains. The papers in this section explore how digital overlays can enhance infrastructure utilization, optimize operations, and map out future technology pathways.
In “Assessing the Adoption of Artificial Intelligence in the Maritime Transport Industry: An Empirical Exploration in the Port of Rotterdam”, Valentina Fani addresses the real-world operational realities of AI integration. Conducting qualitative empirical interviews within a multi-stakeholder ecosystem, including freight forwarders, IT specialists, port authorities, and carriers at Europe’s largest logistics hub, her study uncovers a nuanced landscape. While AI exhibits significant potential to optimize logistics flows, mitigate human errors, and elevate supply chain coordination, its widespread implementation is bottlenecked by high financial costs, legacy software compatibility, and acute workforce anxieties regarding job redundancies. The author points out that successful AI adoption depends heavily on robust data governance and active organizational upskilling to manage the socio-technical transition. This empirical reality highlights that the true challenge lies not in the algorithmic capabilities of AI, but in navigating the structural human and institutional friction that occurs when traditional shipping workflows collide with modern, data-driven automation frameworks.
Shifting focus to vital but often under-researched corridors, Ernest Czermański, Aneta Oniszczuk-Jastrząbek, Jakub Jankiewicz, Tammo Märtens, Björn Krämer, and Ebrahim Ehsanfar present “A Digital Twin Approach to Enhancing the Reliability of Inland Waterway Transport via Water-Level Forecasting and Port Call Coordination”. Drawing upon the European Union’s CRISTAL project, the authors develop a Digital Twin (DT) framework tailored to inland navigation, explicitly testing it within the Po River context. By fusing real-time monitoring sensor data with historical records and machine learning, their DT architecture acts as a predictive decision support system. It accurately forecasts water levels and coordinates port calls, providing early warnings that directly mitigate navigability risks and maintenance costs while advancing the environmental sustainability of inland waterway transport (IWT). By dynamically binding hydraulic modeling with machine learning, this research demonstrates how digital twins move the industry away from reactive scheduling toward proactive, predictive operations, mirroring the broader maritime push toward real-time decision-support networks (Cao et al 2025).
To understand where port technology is heading in the long term, Sung Won Cho, Jeongyoon Hong, Han Jin Lee, and Wonhee Lee contribute “Technological Forecasting of Port Digitalization Using Patent Analysis”. Recognizing the deficit in quantitative evaluations of port tech trends, the authors implement an advanced dual-methodological approach utilizing a vast dataset of port-related patents. They deploy the Latent Dirichlet Allocation (LDA) algorithm for topic modeling, successfully identifying four dominant pillars of port digitalization: intelligent vessel condition monitoring, shipping wireless communication networks, port operations optimization, and container inspection/monitoring.
This highlights the versatility of text-mining and topic-modeling approaches like LDA, which have also been successfully deployed in port studies, to analyze corporate communication, sustainability strategies, and ESG disclosures (Tsatsaronis et al. 2025), inter alia. The study, subsequently, maps these findings using Generative Topographic Mapping (GTM) to visually identify “technological vacuums,” offering a quantitative roadmap for future corporate and public R&D investments. In fact, the core pillars identified via patent analysis, specifically intelligent condition monitoring and wireless communication networks, align directly with the industry’s focus on decentralized edge intelligence and advanced non-linear engine diagnostics. This structural alignment confirms that patenting trends are explicitly catching up to the operational needs of modern smart shipping fleets.
1.3.2 Thematic part II: robotic process automation (RPA) in maritime operations and shipping firms
While full autonomous shipping (MASS) remains a long-term goal, maritime organizations require immediate, non-invasive digital solutions to improve efficiency. This theme evaluates Robotic Process Automation (RPA) –software robots that replicate human desktop actions– across back-office and shipboard environments. Importantly, recent advancements in industrial software engineering have shown that RPA is evolving beyond simple, isolated desktop scripts. By embedding RPA within broader Cyber-Physical Systems (CPS) and Internet of Things (IoT) architectures, shipping firms can establish highly automated, end-to-end data pipelines that seamlessly bridge the gap between physical vessel sensors and back-office enterprise resource planning (ERP) platforms (Kalafatelis et al. 2025c, d, a).
Umut Çelen Arıcan and Leyla Tavacıoğlu offer a comprehensive corporate perspective in “Maritime Container Transportation Efficiency Through Robotic Process Automation: A Digital Transformation Perspective”. The authors present a large-scale, pre-post empirical field implementation over a one-month period within a major global container shipping line. Their study rigorously tracks the deployment of RPA bots across three core back-office workflows: Booking, Invoicing, and Inter-System Data Transfers. By analyzing system logs and operational data, the research demonstrates a dramatic reduction in cycle times, an elimination of data-entry errors, and high execution stability. The study proves that well-governed RPA acts as a powerful catalyst for workflow continuity and structural efficiency without requiring an overhaul of existing IT systems. This empirical validation shows how RPA functions as a flexible “digital glue”, resolving the multi-stakeholder data-sharing anxieties and legacy integration bottlenecks that frequently stall maritime digital initiatives.
Expanding this paradigm from land-based offices to the vessel itself, Serkan Karakas contributes “Exploring the Key Enablers of Robotic Process Automation in the Maritime Industry”. The author acknowledges that while full vessel autonomy faces massive regulatory, legal, and social barriers, shipboard RPA offers a minimally invasive alternative to automate repetitive maritime compliance and reporting tasks directly on commercial vessels. Adopting the Technology-Organisation-Environment (TOE) framework and utilizing the fuzzy DEMATEL methodology, based on insights from senior maritime experts, the study maps the critical success factors behind shipboard software robots. The findings reveal that software automation can drastically enhance ship-to-shore coordination and environmental regulatory compliance, serving as a viable mid-term solution, while full hardware automation still remains unfeasible. By linking these shipboard enablers to automated data ingestion, the research highlights how RPA can work alongside decentralized edge intelligence to process critical operational logs. This intersection of non-invasive software robots and real-time data handling provides shipping firms with an immediate, scalable pathway toward fully automated, compliant, and carbon-aware fleet management.
1.3.3 Thematic part III: advanced machine learning and analytics for forecasting and risk assessment
The final theme centers on predictive intelligence and network science. The papers here harness large datasets and advanced machine learning algorithms to forecast volatile markets, predict corporate insolvency, and map structural network shifts.
In “Machine Learning in Freight Rate Forecasting: A Systematic Literature Review,” Fabian Kjeldsberg, Ziaul Haque Munim, and Hans-Joachim Schramm consolidate a decade of research on market volatility. Synthesizing 28 seminal articles published between 2012 and 2024, the authors systematically catalog 17 target (dependent) and 59 input (independent) variables utilized in global freight rate forecasting models. Their review reveals that while various Neural Network (NN) architectures dominate the academic landscape, specialized hybrid and ensemble models consistently outperform standalone variations due to their superior ability to handle high-fluctuation, non-linear financial data. This paper serves as an essential methodological reference point for researchers designing predictive maritime models. In fact, this comprehensive mapping emphasizes why the industry is steadily abandoning rigid, traditional linear frameworks in favor of highly adaptable, multi-layered computational frameworks that can withstand sudden structural breaks. Beyond freight rates, such machine learning algorithms, including Support Vector Machines (SVM), have proven highly effective in addressing non-linearities in asset markets, particularly in forecasting ship prices (Syriopoulos et al. 2021).
Drilling down into a highly specific and volatile energy market, Junseong Kim and Daisuke Watanabe present “Long-Term Prediction of Liquefied Natural Gas Spot Freight Rates”. Given the intensifying global reliance on LNG, and the extreme price fluctuations of spot charter rates, the authors construct an artificial intelligence framework to execute long-term monthly predictions, extending up to 16 weeks. Fusing critical predictive market indicators –such as global LNG prices, inventory volumes, vessel sailing speeds, port call indices, and charter rates– they conduct a rigorous comparative evaluation between Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models. The empirical findings reveal that the GRU model achieves approximately 70% lower Mean Squared Error (MSE) than the LSTM model, demonstrating superior algorithmic stability and accuracy across extended forecasting horizons. This dramatic variance in error rates proves that selecting the right deep learning architecture is critical when processing non-linear maritime data over prolonged horizons, where standard networks often suffer from memory decay and compounding prediction errors. Furthermore, such market fluctuations are inextricably linked to broader maritime asset dynamics, necessitating a deeper understanding of vessel pricing volatility and investor behavioral patterns under structural shifts (Syriopoulos and Bakos 2019; Tsatsaronis et al. 2026).
Addressing macroeconomic and corporate security, Minsu Kwon, Saeyeon Roh, and Minchul Sohn author “Advanced Machine Learning Approaches for Bankruptcy Prediction in the Shipping Industry: A Comparative Analysis Across Time Horizons”. Operating on a highly imbalanced dataset from the Korean shipping industry, the authors utilize advanced ensemble machine learning techniques (including LightGBM, CatBoost, and XGBoost) alongside SMOTE oversampling to predict corporate distress across one-, three-, and five-year horizons. Their ensemble models vastly outperform traditional linear models and standard deep neural networks. Crucially, by integrating SHAP (Shapley Additive exPlanations) values, they discover that short-term bankruptcy risk is heavily driven by immediate internal cash-flow and liquidity factors, whereas long-term insolvency is highly sensitive to external maritime indicators, such as sustained freight rate drops. By introducing this layer of explainability, the study answers a critical call within maritime operations: moving away from “black-box” models and providing clear, actionable transparency that financial institutions and regulators can rely on during economic downturns (Gavalas and Syriopoulos 2014).
Concluding the issue with a macro-network perspective, Aythami Santana-Padrón, Casiano Manrique-de-Lara-Peñate, and Lourdes Trujillo present “Using Association Rules to Map Vessel Behaviour Patterns and Port Relationships”. Fusing data science with maritime economics, the authors develop a novel Adapted Association Rules Mining (AARM) framework to analyze historical container vessel AIS movement data, spanning 2016 to 2024. Mapping nearly 50,000 distinct port connections, they unveil a hidden structural architecture within global shipping networks, categorizing bilateral port dynamics into three distinct strategic archetypes: a dominant flexible, natural state (~ 80%), and two specialized, asymmetric states showing intense dependencies on either origin or destination ports (~ 20%). Furthermore, their temporal analysis captures the precise structural reconfigurations triggered by the COVID-19 pandemic, showing how major maritime hubs permanently reorganized their strategic relationships following global disruption. This structural plasticity highlights that maritime networks are dynamic, evolving organisms. Understanding these behavioral archetypes enables ports and supply chain managers to transition from reactive crisis management to proactive, resilient network orchestration.
