Trustworthy Data- and ML-Ops for Intelligent Transportation Systems and Logistics

Antonio Emanuele Cinà, Giovanni Scodeller, Cecilia Caterina Pasquale, Silvia Siri, Davide Anguita, Fabio Roli, Simona Sacone, Luca OnetoarXiv 2026 · DOI · arXiv

TL;DR

This paper provides a comprehensive roadmap for integrating DataOps and MLOps with Trustworthy AI (TAI) principles specifically for the high-stakes world of Intelligent Transportation Systems and Logistics (ITS&L). It moves beyond simple model accuracy to establish how we can build AI-driven transport systems that are not only efficient but also robust, fair, private, and sustainable.

What Problem Does This Paper Address?

Intelligent Transportation Systems and Logistics (ITS&L)—think autonomous vehicles, smart traffic lights, and global supply chain managers—are drowning in data. This data follows the "3V" model of Big Data: massive Volume, high Velocity (real-time requirements), and extreme Variety (GPS logs, video feeds, sensor data, and social media).

While AI and Machine Learning (ML) offer solutions to optimize these systems, the industry lacks a standardized way to manage the "plumbing" of these models. Furthermore, because these systems are safety-critical and affect public life, we can't just care about performance. We need "Trustworthy AI," yet principles like fairness, privacy, and robustness are often treated as afterthoughts rather than being baked into the system's operational lifecycle.

Methods

The authors propose a unified conceptual framework that layers three critical paradigms:

  1. DataOps: Managing the data lifecycle (ingestion to feedback) to ensure high-quality, reliable data streams.
  2. MLOps: Automating the machine learning lifecycle (training to deployment) to handle continuous updates and model shifts.
  3. Trustworthy AI (TAI): A governance layer that monitors systems for five key dimensions: robustness, explainability, fairness, privacy, and sustainability.

The paper maps these paradigms onto a standard DevOps-style iterative loop, as shown in the integrated view below:

Figure 1: DataOps and MLOps in ITS&L. Heterogeneous data sources are governed through the DataOps lifecycle and subsequently support the MLOps lifecycle for model develop- ment, deployment, monitoring, and updating. The resulting AI- enabled services operate across the main ITS&L application domains, while TAI provides a governance layer grounded in robustness, explainability, fairness, privacy, and sustainability.

The authors also extend this discussion to emerging trends like LLMOps (for Large Language Models) and AgentOps (for autonomous agentic systems), which are beginning to appear in transport applications like airport chatbots and autonomous fleet coordinators.

Datasets

As a comprehensive review, the paper does not introduce a new dataset but analyzes 153 case studies that utilize various benchmarks, including:

  • Perception & Driving: The SHIFT dataset (robustness), German Traffic Sign Recognition Benchmark (GTSRB), and the Udacity autonomous driving dataset.
  • Traffic Flow: The UA-DETRAC dataset for real-time vehicle detection.
  • Social & Emergency: The Philadelphia Crime Incidents dataset (used to test fairness in emergency resource allocation).
  • Real-world City Data: Large-scale deployments in Dublin, Munich, Adelaide, and Bangkok.

Key Insights and Ideas

  • The Synergy is Mandatory: You cannot achieve Trustworthy AI without solid DataOps and MLOps. A model cannot be "fair" if the data pipeline that feeds it is a biased black box, and it cannot be "robust" if there isn't an automated MLOps pipeline to catch performance drifts in real-time.
  • The "Trustworthiness Gap": The authors' meta-review (summarized in Figure 4) reveals that while DataOps is becoming common in transport, MLOps and holistic TAI are lagging. Most research focuses on just one pillar (like privacy) while ignoring others (like fairness).
  • Sustainability as a Technical Metric: In ITS&L, sustainability isn't just an ethical goal; it's a technical constraint. Models running on edge devices (like a delivery drone or a smart camera) must be lightweight and energy-efficient to be viable.
  • From Models to Agents: We are moving from simple predictive models to "agentic" systems that take actions. This introduces "AgentOps," where we must monitor not just the output of a model, but the entire execution trajectory of an autonomous agent to prevent cascading errors.

Key Results

The paper provides a detailed comparison of tools and a summary of the current state of the field:

  • Maturity Levels: DataOps is relatively mature (82 case studies identified), but MLOps is still in its infancy in this sector (only 19 cases).
  • TAI Imbalance: Privacy and Sustainability are the most researched TAI dimensions (17 cases each), while Fairness is the most neglected (only 5 cases).
  • Tooling: The authors provide a "cheat sheet" comparing open-source tools (like Apache Spark and Kubeflow) against proprietary platforms (like Google VertexAI and AWS SageMaker), specifically evaluating their suitability for transport-specific needs like on-premise deployment and real-time scalability.

Limitations and Future Work

The authors identify several "open problems" that the research community needs to tackle:

  • Lack of Unified Frameworks: There is currently no single software framework that integrates DataOps, MLOps, and TAI into one cohesive workflow for engineers.
  • LLMOps in Transport: While LLMs are being adopted, there is almost no public documentation on how to manage their lifecycle (LLMOps) in a safety-critical transport context.
  • Non-Holistic TAI: We need methods that optimize for all five pillars of trust simultaneously, rather than trading off fairness for accuracy or privacy for robustness.
  • Standardized Verification: The field needs better standardized tools to verify trustworthiness during the pipeline, rather than as a final check before deployment.

Summary updated

Abstract

The rapid evolution of Intelligent Transportation Systems and Logistics (ITS\&L) has become a cornerstone of the modern social economy, relying heavily on the integration of Data, Artificial Intelligence (AI), and, more specifically, Machine Learning (ML). This paper provides a comprehensive review of Trustworthy Data and Machine Learning Operations (DataOps and MLOps) in the ITS\&L domain, underscoring their importance in improving efficiency, reliability, and decision-making precision within transportation and logistics services. We begin by identifying gaps in current literature, offering clear context for our contribution. Subsequently, we explore the complexities of DataOps and MLOps, discussing their necessity, key components, available tools, practical insights, and case studies relevant to ITS\&L. Additionally, we address the critical issue of Trustworthiness in AI applications, examining methods and tools designed to strengthen confidence in AI systems - especially in real-world ITS\&L scenarios. The paper concludes with a discussion of persisting challenges and future prospects in this rapidly advancing field, aiming to serve as a vital resource for researchers, industry practitioners, and policy makers. Overall, this work not only establishes a foundational understanding of DataOps and MLOps in ITS\&L but also charts a path for further research and innovation in developing more efficient, sustainable, and trustworthy intelligent transportation and logistics systems.
BibTeX
@misc{cina2026trustworthy,
  author = {Antonio Emanuele Cinà and Giovanni Scodeller and Cecilia Caterina Pasquale and Silvia Siri and Davide Anguita and Fabio Roli and Simona Sacone and Luca Oneto},
  title = {{Trustworthy Data- and ML-Ops for Intelligent Transportation Systems and Logistics}},
  year = {2026},
  eprint = {2610.01282},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.01282},
}
Figure