Periodic Reporting for period 2 - TALON (Autonomous and Self-organized Artificial Intelligent Orchestrator for a Greener Industry 4.0)
Période du rapport: 2024-04-01 au 2025-09-30
Next-generation industrial systems promise to deliver unprecedented excellence not only in terms of performance, but also explainability, trustworthiness and transparency. To achieve this new objectives, state-of-the-art concepts of artificial intelligence (AI), edge-to-cloud (E2C) computing, blockchain, and visualisation need to be de-risked and applied. Motivated by this, TALON aims at sculpturing the road towards the next Industrial revolution by developing a fully-automated AI platform capable of bringing intelligence near the edge in a flexible, adaptable, explainable, energy and data efficient manner.
To achieve these goals, TALON followed a structured approach made up of four phases: I: Foundations and Alignment; II: Design and Development; III: Validation and Demonstration; and IV: Regulation and standardization.
The continuous interaction across such hases, allowed us to design and develop TALON platform consisting of three fundamental pillars: a) an AI orchestrator that coordinates the network and service orchestrators in order to optimise the edge vs cloud relationship, while boosting reusability of datasets, algorithms and models by deciding where each one should be placed; b) a lightweight hierarchical blockchain schemes that introduce new service models and applications under a privacy and security umbrella; and c) new transfer learning and visualization approaches that enhance AI trustworthiness and transparency.
The TALON architecture combines the benefits of AI, edge and cloud networking, as well as blockchain and DTs, optimized by means of a) new key performance indicators that translate the AI benefits into insightful metrics; b) novel theoretical framework for the characterisation of the AI; c) blockchain used to deliver personalised & perpetual protection based on security, privacy and trust mechanisms; d) AI approaches for automatically and co-optimising edge and cloud resources as well as the AI execution nodes; e) semantic AI to reduce the learning latency and enhance reusability; and f) digital twins that visualize the AI outputs and together with human-in-the-loop approaches. All the technological breakthroughs were demonstrated, validated and evaluated by means of proof-of-concept simulation and four real-world pilots.
The continuous interaction across such hases, allowed us to design and develop TALON platform consisting of three fundamental pillars: a) an AI orchestrator that coordinates the network and service orchestrators in order to optimise the edge vs cloud relationship, while boosting reusability of datasets, algorithms and models by deciding where each one should be placed; b) a lightweight hierarchical blockchain schemes that introduce new service models and applications under a privacy and security umbrella; and c) new transfer learning and visualization approaches that enhance AI trustworthiness and transparency.
The TALON architecture combines the benefits of AI, edge and cloud networking, as well as blockchain and DTs, optimized by means of a) new key performance indicators that translate the AI benefits into insightful metrics; b) novel theoretical framework for the characterisation of the AI; c) blockchain used to deliver personalised & perpetual protection based on security, privacy and trust mechanisms; d) AI approaches for automatically and co-optimising edge and cloud resources as well as the AI execution nodes; e) semantic AI to reduce the learning latency and enhance reusability; and f) digital twins that visualize the AI outputs and together with human-in-the-loop approaches. All the technological breakthroughs were demonstrated, validated and evaluated by means of proof-of-concept simulation and four real-world pilots.
Innovations beyond the State-of-the-Art (SoTA) in TALON are summarised as follows:
[R1] An automated AI E2C orchestrator enabling zero-touch deployment and autonomous runtime management of AI workloads.
[R2] Four slices for i) automatic UATV coordination, ii) I5.0 automation and planning, iii) AR/VR for training and maintenance, and iv) HRC, that in turn proved different modular usages and deployment of the TALON orchestrator.
[R3] A Mobility manager and social-aware caching mechanisms were empowered by data-driven policies derived from pods resource utilisation and energy usage only when allocated for specific tasks.
[R4] The Automatic off-loading mechanism as part of the TALON Edge-to-Cloud (E2C) Orchestrator, enables real-time delegation of computational workloads between edge nodes and the cloud.
[R5] AI self-healing, self-recovery, and self-correcting mechanisms to ensure resilient and zero-touch operation across the Edge-to-Cloud (E2C) continuum.
[R6] An AI models repository that enables reusability.
[R7] Optimized few-shot learning models that reduce the data exchanging overhead.
[R8] Federated learning deployments were optimized, focused on minimizing information exchange while maximizing model utility across both distributed and centralized scenarios.
[R9] Optimum joint AI and data placement policies were developed.
[R10] The overall AI energy footprint was significantly reduced through the implementation of smart pricing policies.
[R11] Decentralized and hierarchical blockchain-based mechanisms were implemented to accommodate the requirements of TALON pilots and further optimise the mechanisms of the decentralised ledger.
[R12] Novel KPIs were designed to measure the performance improvements of AI-enabled systems in alignment with business objectives derived from end user requirements.
[R13] A data-driven real-time AI optimizer was finalized and validated during piloting, significantly enhancing E2C deployment, management, and resource reuse.
[R14] Simulation mechanisms that enable fast learning were implemented by means of an AI-based synthetic data generation module.
[R15] An Explainable AI (XAI) framework has been designed and implemented to provide semantics, enabling human understanding of the employed AI models.
[R16] An innovative theoretical framework that quantifies the performance of AI algorithms was delivered.
[R17] Extensive benchmarking and execution of multiple AI pipelines tailored for specific learning tasks were conducted to quantify and optimize AI algorithms operations.
[R18] LIME-compatible local explanation schemes were seamlessly integrated into the TALON Dashboard, fostering transparent, locally-interpretable decision-making within all four industrial pilots.
[R19] Counterfactual-based explainability mechanisms were fully implemented and operationalized.
[R20] A digital model was defined to fully align with the final data schema of the Nakamura2 CNC machine in Use case 2.
[R1] An automated AI E2C orchestrator enabling zero-touch deployment and autonomous runtime management of AI workloads.
[R2] Four slices for i) automatic UATV coordination, ii) I5.0 automation and planning, iii) AR/VR for training and maintenance, and iv) HRC, that in turn proved different modular usages and deployment of the TALON orchestrator.
[R3] A Mobility manager and social-aware caching mechanisms were empowered by data-driven policies derived from pods resource utilisation and energy usage only when allocated for specific tasks.
[R4] The Automatic off-loading mechanism as part of the TALON Edge-to-Cloud (E2C) Orchestrator, enables real-time delegation of computational workloads between edge nodes and the cloud.
[R5] AI self-healing, self-recovery, and self-correcting mechanisms to ensure resilient and zero-touch operation across the Edge-to-Cloud (E2C) continuum.
[R6] An AI models repository that enables reusability.
[R7] Optimized few-shot learning models that reduce the data exchanging overhead.
[R8] Federated learning deployments were optimized, focused on minimizing information exchange while maximizing model utility across both distributed and centralized scenarios.
[R9] Optimum joint AI and data placement policies were developed.
[R10] The overall AI energy footprint was significantly reduced through the implementation of smart pricing policies.
[R11] Decentralized and hierarchical blockchain-based mechanisms were implemented to accommodate the requirements of TALON pilots and further optimise the mechanisms of the decentralised ledger.
[R12] Novel KPIs were designed to measure the performance improvements of AI-enabled systems in alignment with business objectives derived from end user requirements.
[R13] A data-driven real-time AI optimizer was finalized and validated during piloting, significantly enhancing E2C deployment, management, and resource reuse.
[R14] Simulation mechanisms that enable fast learning were implemented by means of an AI-based synthetic data generation module.
[R15] An Explainable AI (XAI) framework has been designed and implemented to provide semantics, enabling human understanding of the employed AI models.
[R16] An innovative theoretical framework that quantifies the performance of AI algorithms was delivered.
[R17] Extensive benchmarking and execution of multiple AI pipelines tailored for specific learning tasks were conducted to quantify and optimize AI algorithms operations.
[R18] LIME-compatible local explanation schemes were seamlessly integrated into the TALON Dashboard, fostering transparent, locally-interpretable decision-making within all four industrial pilots.
[R19] Counterfactual-based explainability mechanisms were fully implemented and operationalized.
[R20] A digital model was defined to fully align with the final data schema of the Nakamura2 CNC machine in Use case 2.