Periodic Reporting for period 2 - MobiSpaces (New data spaces for green mobility)
Reporting period: 2024-03-01 to 2025-08-31
A vast amount of real-time data is collected by sensors and devices and processed in data centres using cloud computing, Big Data, and AI. To fully leverage this data within secure digital infrastructures, effective governance is needed to enable sharing, reuse, and interoperability across organizations. MobiSpaces is a Research and Innovation Action that targets optimising data spaces for mobility data, by optimising the complete data path from data acquisition to data processing and data analytics. The aim is to develop a mobility-aware and mobility-optimized platform that addresses the full lifecycle of data, by enabling secure and trustworthy data sharing and exchange for mobility data. Towards this goal, MobiSpaces contributes to the design of future mobility data spaces by (i) offering data governance services for mobility data for data providers, (ii) supporting data consumers via advanced data analysis services, (iii) providing these services in a data spaces compliant way, (iv) in a trusted environment that takes into account the architectural blueprint of data spaces. Key results include the delivery of a data governance framework for mobility data validated in urban and maritime scenarios, a data operations toolbox that comprises multiple and diverse data management solutions for mobility data focusing on declarative querying, an edge analytics suite that encompasses multiple algorithms for edge analytics, federated learning, explainable AI and visual analytics. Notably, in combination with data-related operations, MobiSpaces exploits edge computing technology to process mobility data near the sources of data collection, while also targeting effective resource allocation in multi-cluster environments, making it possible to efficiently deploy dynamic pipelines of mobility-related processing components. The outcome is innovative tools and techniques for building a robust and sustainable ecosystem for the full lifecycle of mobility data — from acquisition to AI-driven analysis —ensuring secure, interoperable, and reusable mobility data that contributes to the vision of European data spaces.
MobiSpaces has delivered a set of software modules that implement innovative algorithms for mobility data management and analytics. The main scientific and technical achievements can be summarized as follows:
• A data governance platform for mobility data, comprising data-related operations (semantic modelling, data transformation, data cleaning, data provenance) as well as security-related operations both for design-time security and for real-time privacy preservation. Also, a mechanism for automatic deployment of data-related pipelines in multi-cluster settings that includes edge clusters is also provided.
• A set of declarative querying systems that support the concept of “SQL on everything”, also tailored for mobility data. This includes systems such as MobilityDB and PyMEOS for spatio-temporal data, NoDA for querying NoSQL stores, LeanXcale for high-throughput transaction processing.
• An end-to-end framework for the design and deployment of AI workflows that includes an AI workflow builder that is able to assign such workflows to a smart resource allocator that tries to find the optimal deployment in a given cluster setting, taking into account availability, trustworthiness and energy consumption metrics.
• An edge analytics suite that contains specialized algorithms for mobility analytics, including edge analytics, federated learning algorithms, explainable AI for spatio-temporal data, and privacy-aware visual analytics.
• Integration of the above technologies with IDS-compliant data connectors, to showcase the applicability of MobiSpaces technology in the context of European data spaces.
• Evaluation and validation in five (5) challenging, commercial use-cases, covering two (2) mobility domains: urban and maritime.
• A data governance platform for mobility data, comprising data-related operations (semantic modelling, data transformation, data cleaning, data provenance) as well as security-related operations both for design-time security and for real-time privacy preservation. Also, a mechanism for automatic deployment of data-related pipelines in multi-cluster settings that includes edge clusters is also provided.
• A set of declarative querying systems that support the concept of “SQL on everything”, also tailored for mobility data. This includes systems such as MobilityDB and PyMEOS for spatio-temporal data, NoDA for querying NoSQL stores, LeanXcale for high-throughput transaction processing.
• An end-to-end framework for the design and deployment of AI workflows that includes an AI workflow builder that is able to assign such workflows to a smart resource allocator that tries to find the optimal deployment in a given cluster setting, taking into account availability, trustworthiness and energy consumption metrics.
• An edge analytics suite that contains specialized algorithms for mobility analytics, including edge analytics, federated learning algorithms, explainable AI for spatio-temporal data, and privacy-aware visual analytics.
• Integration of the above technologies with IDS-compliant data connectors, to showcase the applicability of MobiSpaces technology in the context of European data spaces.
• Evaluation and validation in five (5) challenging, commercial use-cases, covering two (2) mobility domains: urban and maritime.
MobiSpaces has delivered several results that push the research frontier beyond the state of the art in several areas related to data sharing, management and analytics. An indicative list follows:
1. A data pipeline for making mobility data FAIR, exploiting semantic representation, metadata annotation, data interlinking and data provenance.
2. The Security Risk Modeler, a tool for monitoring security at design-time.
3. Open-source libraries for mobility data processing, suitable to run on edge environments, such as PyMEOS, and streaming mobility data.
4. A novel format for cloud-based storage, called TrajParquet, that extends Apache Parquet to become applicable for trajectories.
5. An AutoML approach that enables automatic selection of the best clustering algorithm for a given tabular dataset.
6. A highly efficient data warehousing solution for storing and querying AIS datasets that span several years, thus allowing easy and efficient data analytics on historical vessel positional data.
7. An extremely efficient suite of edge analytics algorithms that outperform existing solutions by several orders of magnitude due to optimized implementation.
8. Techniques for adapting explainable AI methods for the domain of spatial and spatio-temporal data.
9. Several federated learning algorithms for mobility-related applications, including data cleaning, anomaly detection, future location prediction.
1. A data pipeline for making mobility data FAIR, exploiting semantic representation, metadata annotation, data interlinking and data provenance.
2. The Security Risk Modeler, a tool for monitoring security at design-time.
3. Open-source libraries for mobility data processing, suitable to run on edge environments, such as PyMEOS, and streaming mobility data.
4. A novel format for cloud-based storage, called TrajParquet, that extends Apache Parquet to become applicable for trajectories.
5. An AutoML approach that enables automatic selection of the best clustering algorithm for a given tabular dataset.
6. A highly efficient data warehousing solution for storing and querying AIS datasets that span several years, thus allowing easy and efficient data analytics on historical vessel positional data.
7. An extremely efficient suite of edge analytics algorithms that outperform existing solutions by several orders of magnitude due to optimized implementation.
8. Techniques for adapting explainable AI methods for the domain of spatial and spatio-temporal data.
9. Several federated learning algorithms for mobility-related applications, including data cleaning, anomaly detection, future location prediction.