optimal orchestration
aerOS will enable distributed computing and storage, by orchestrating services on heterogeneous (hardware) nodes running various operating systems.
The aerOS vision
Transition to software-driven components and systems requires efficient integration of a variety of new vertically agnostic technologies and services. It requires data strategy, trusted data exchange frameworks, frugal AI on the (far) edge, and containerisation and virtualisation across complex value chains. Besides, collective decentralised decision-making, and system / network federation must be exploited to govern data exchange within models, applications and services across the IoT edge-cloud continuum. The meta operating system has to (among others) provide flexible orchestration mechanisms for combining and efficiently utilising the heterogeneous Infrastructure Elements comprising the IoT edge-cloud continuum. In particular, it should support exposed standardised services APIs, hardware abstraction, cross-domain resource orchestration, and explainable/decentralised frugal AI. AerOS focuses on the distributed data management to make user-side applications more intelligent and proactive, and to provide foundation for hyper-distributed applications and services, closer to data sources and event-generating processes without sacrificing aggregated data analysis and insights. Additionally, the concept of services as a “unifying abstraction”, across resources (i.e., any physical or virtual IoT edge-cloud continuum resource, from device to far-edge, edge or cloud), is envisioned.
Objectives
Definition and implementation of distributed AI components with explainability
Deliver dynamically deployable, containerised solutions, offering AI/ML support for applications with components executable on (near to) edge devices with the support of distributed model training, e.g., split/federated learning, where only model updates are communicated, fostering data privacy.
Definition and implementation of decentralised security, privacy and trust
Introduce a holistic cross-layer solution for cybersecurity, while supporting federated and distributed data governance, by utilising technologies for mutual trust assurance, leverage containerisation and open APIs for data quality, security, privacy and trust assessments
Specification and implementation of a Data Autonomy strategy for the IoT edge-cloud continuum
will support needs of data producers/consumers towards collection, processing, storage, and distribution of actionable data with mechanisms, such as: (i) data ownership, control and governance enhanced with (cryptographic) non-repudiation and traceability assurance; (ii) timely delivery of data to (ML) processors via just-in-time predictive projections; (iii) semantic and syntactic data interoperability mediation to ease integration and widen possible range of cooperating entities, services, and stakeholders;
Intelligent realisation of smart network functions
aerOS will develop latency-controlled distributed networking components and will also support industrial IoT communication, by considering private 5G NEF and SEAL exposed APIs, TSN integration, mesh networks and time-triggered real-time-capable IoT endpoints.
Global ecosystem creation, maximisation of impact and Open Call conduction
To present/promote results within publications, conferences, website, social media, and workshops. Further, several exhibitions/showcases will take place, including small demonstrations, to engage stakeholders and potential clients from diverse verticals, mobilising key actors. Moreover, business models (aligned with use cases technical capabilities, ecosystem, and functionalities) that include market and sustainability strategy, supported by preliminary value proposition and pricing policy, will be proposed
Definition, deployment, and evaluation of real-life use cases
Five industry-driven heterogeneous use cases are envisioned to demonstrate the value of aerOS in IoT edge-cloud continuum orchestration and highlight the value of data sharing and governance in connection with data spaces. Validation will involve real-world data, with consideration to interoperability, volume, variety, and rate
aerOS Reference Architecture
aerOS architecture will be modular, to fully exploit encapsulated functionalities, with ease of including new modules. In order to be o be scalable, decentralised, hierarchical and adaptable to different topologies comprises of virtualised modules, executed on top of any operating system (e.g., Linux-based, Android, ROS, etc.) of an Infrastructure Element (IE) of the IoT edge-cloud continuum, e.g., a smart device, IoT gateway, edge node or network component. The continuum may (and will) consist of different edge layers (e.g., tiny/far/near), with possibility of multiple distributed groups. For each edge level, there will be an IE acting as a (replicable) master, which will control execution of tasks and services of managed nodes. Any aerOS node communicates with an aerOS master to which it registers (e.g., services that it can provide) and updates the status (e.g., resource utilisation) of its associated IE (e.g., a far edge node with associated sensors) and available services. Each aerOS IE deployment will consist of the following key modules: (i) services and API; (ii) virtualisation, abstraction and container runtime; (iii) core aerOS modules; (iv) supporting aerOS features; (v) orchestration; (vi) security, privacy and trust; and (vii) management framework.
Grant Agreement
This project has received funding from Horizon Europe, the EU’s key funding programme for research and innovation, under grant agreement No 101069732
