MobiSpaces delivers an end-to-end mobility-aware and mobility optimised data governance platform, that concerns the offerings of data acquisition, in-situ processing and all the security- and privacy-related operations.
MobiSpaces envisions a set of toolboxes, suites, and tools that implement the MobiSpaces concept. MobiSpaces identifies the AI-based Data Operations Toolbox, including an additional list of tools, namely the Declarative querying, Decentralized Data Management, and Online Data Aggregator. And the Edge Analytics Suite, including a further list of tools, namely the XAI Prediction Modelling, Edge-driven Federated Learning, and Visual Analytics.
The MobiSpaces platform will be surrounded by a Green & Environmental Dimensioning Workbench for monitoring and advising the processing behavior, ensuring guidelines and legislations towards the zero-carbon footprint and the “do no significant harm” principle. Read on to learn more about the set of toolboxes, suites, and tools MobiSpaces will deliver.
AI-based Data Operations Toolbox
MobiSpaces will empower data management systems and enhance their capabilities by integrating AI-based data operations in a toolbox that can be applied in mobility settings. This toolbox includes a list of tools which are both integrated into the AI-based data operations toolbox, while being also able to perform autonomously.
MobiSpaces follows an approach that defines abstract data operations with clear semantics, which are fundamental for expressing more complex queries.
Decentralized Data Management
MobiSpaces will use a massively distributed and adaptive infrastructure that will be able to make use of large numbers of edge compute nodes to collect, cache, manage, aggregate, analyze, model and present huge amounts of data from a large number of objects and heterogeneous sources
Online Data Aggregator
MobiSpaces will develop a tool that computes and aggregates statistics from multiple, distributed data streams in an online and incremental way. To reduce the memory footprint, appropriate data sketches will be applied on the streaming data, while approximate query processing techniques will be used to compute results with high accuracy.
Edge Analytics Suite
MobiSpaces will deliver a suite of decentralized, edge analytics algorithms, based on techniques for Machine Learning over mobility data that go beyond the state-of-the-art.
XAI Prediction Modelling
MobiSpaces goes one step beyond building robust and accurate mobility models from data, by focusing on model interpretability which is focal in operational environments. A tool equipped with XAI techniques focusing on the explainability of deep neural networks will be delivered.
Edge-driven Federated Learning
MobiSpaces will design and implement a Federated Learning architecture to address the challenges associated with mobility datasets that are massively distributed. Under this setting, data is stored locally at the edge, while a primary model is stored in a centralised location.
This tool aims to develop interactive and scalable methodologies, which can efficiently handle both historical and streaming spatiotemporal data originating from different sources, with varying levels of resolution and quality.