Periodic Reporting for period 1 - VERGE (AI-powered eVolution towards opEn and secuRe edGe architEctures)
Reporting period: 2023-01-01 to 2023-12-31
The main goal of VERGE is to provide an integrated approach on how to tackle the challenges of edge computing evolution described around three main pillars: (1)“Edge for AI (Edge4AI)”, namely a flexible, modular and converged edge platform design, unifying the Life Cycle Management (LCM) and closed-loop automation for cloud-native applications, Multi-access Edge Computing (MEC) and network services across a multi-domain edge-cloud continuum, while fully exploiting the capabilities of cutting-edge multi-core and multi-accelerator platforms for ultra-high computational performance. (2) “AI for edge (AI4Edge)” namely an AI-powered portfolio of solutions that will leverage the multitude of information and metrics provided by the VERGE platform monitoring mechanisms to manage and orchestrate the VERGE platform computing and network resources. (3) “Security, privacy and trustworthiness for AI (SPT4AI)”, through a suite of methods and tools that will ensure a) security of the AI-based models against adversarial attacks, b) privacy of data and models, and c) sample-efficient training, safety verification and explainability for model decisions, resulting in improved trust in AI models.
This main goal is achieved through the following specific obejctives:
1.- Design an open and secure edge computing architecture to efficiently support highly demanding XR and IoT-driven applications in terms of computation and connectivity, relying on an edge-cloud continuum that fosters virtualisation of massive heterogeneous edge computing resources across a multi-access RAN.
2.-Design and build the Edge4AI concept that provides the infrastructure, mechanisms and programming tools to enable the development, deployment and orchestration of massively distributed AI processes across heterogeneous computing and memory resources.
3.-Design an AI4Edge framework that encompasses cutting-edge AI solutions for managing and orchestrating the edge computing and the RAN resources towards an optimum performance that satisfies the highly demanding applications requiring extremely low latencies and/or very high-capacity justifying edge processing and computing.
4.-Develop tools that ensure the security, privacy and trustworthiness of the VERGE system.
5. Showcase the VERGE solutions by means of Proof of Concept (PoC) demonstrations.
6. Carry out extensive dissemination, standardisation and exploitation activities.
The project solutions intend to impact on different improvement areas, namely reduced latency, enhanced service availability, enhanced service reliability, efficient resource utilization, energy efficiency, flexibility/scalability, enhanced computation, efficient AI model training, AI safety and AI explainability.
This main goal is achieved through the following specific obejctives:
1.- Design an open and secure edge computing architecture to efficiently support highly demanding XR and IoT-driven applications in terms of computation and connectivity, relying on an edge-cloud continuum that fosters virtualisation of massive heterogeneous edge computing resources across a multi-access RAN.
2.-Design and build the Edge4AI concept that provides the infrastructure, mechanisms and programming tools to enable the development, deployment and orchestration of massively distributed AI processes across heterogeneous computing and memory resources.
3.-Design an AI4Edge framework that encompasses cutting-edge AI solutions for managing and orchestrating the edge computing and the RAN resources towards an optimum performance that satisfies the highly demanding applications requiring extremely low latencies and/or very high-capacity justifying edge processing and computing.
4.-Develop tools that ensure the security, privacy and trustworthiness of the VERGE system.
5. Showcase the VERGE solutions by means of Proof of Concept (PoC) demonstrations.
6. Carry out extensive dissemination, standardisation and exploitation activities.
The project solutions intend to impact on different improvement areas, namely reduced latency, enhanced service availability, enhanced service reliability, efficient resource utilization, energy efficiency, flexibility/scalability, enhanced computation, efficient AI model training, AI safety and AI explainability.
During the first year of the project, a detailed specification of two use cases has been given, namely "XR-driven edge-enabled industrial B5G applications" and "Autonomous tram services for safety and entertainment in a smart city environment". They have been broken down in several user stories with different requirements and constraints, identifying the involved actors and their roles and associated Service Level Requirements that need to be met by the proposed VERGE architectural design. Based on this, an edge computing architecture has been proposed for the creation of an integrated B5G-enabled multi-site edge-cloud compute continuum unifying the three VERGE conceptual pillars into a modular and flexible design.
Aligned with the proposed architecture, the first version of Edge4AI, AI4Edge and SPT4AI pillars has been delivered.
The key features of the Edge4AI are: (1) Support for programming models and frameworks from the embedded, High Performance Computing (HPC) and AI domains. These leverage parallelism and reconfiguration capabilities of heterogeneous accelerated architectures (GPUs, FPGAs, etc.) to support different levels of distributed computation. (2) A hierarchical service orchestration, management and control layer that supports different levels of granularity, i.e. across multiple edge sites, within an edge cluster, and at task level. (3) Support for close-loop automation, through the data access layer, which handles the collection of relevant metrics, and the cognitive framework, which provides the open APIs and services to facilitate the LCM of AI/ML models.
In AI4Edge, different cutting-edge AI techniques have been developed, including: (1) Edge resource management solutions for dynamically and/or proactively optimizing computational resource allocation (e.g. CPU rightsizing and autoscaling). (2) Advanced learning solutions for heterogeneous edge environments, including federated and collaborative learning, and split learning to offloading part of complex deep neural networks from the UEs to the edge. (3) RAN management solutions for edge-enabled relay management, network slicing, multi-tier RAN optimization, dynamic functional split and micro-orchestration of RAN functions. (4) E2E AI orchestration for collision-free decision-making.
In SPT4AI, the studied methodologies include: (1) Security and privacy issues through threat analysis and mitigation measures, and tackling vulnerabilities arising e.g. in distributed MIMO. (2) Resilience and robustness of AI models in decentralized edge deployments against attacks that compromise the quality of training data. (3) Solutions for the generalization, formal verification, safe deployment and training, and explainability of AI models.
Aligned with the proposed architecture, the first version of Edge4AI, AI4Edge and SPT4AI pillars has been delivered.
The key features of the Edge4AI are: (1) Support for programming models and frameworks from the embedded, High Performance Computing (HPC) and AI domains. These leverage parallelism and reconfiguration capabilities of heterogeneous accelerated architectures (GPUs, FPGAs, etc.) to support different levels of distributed computation. (2) A hierarchical service orchestration, management and control layer that supports different levels of granularity, i.e. across multiple edge sites, within an edge cluster, and at task level. (3) Support for close-loop automation, through the data access layer, which handles the collection of relevant metrics, and the cognitive framework, which provides the open APIs and services to facilitate the LCM of AI/ML models.
In AI4Edge, different cutting-edge AI techniques have been developed, including: (1) Edge resource management solutions for dynamically and/or proactively optimizing computational resource allocation (e.g. CPU rightsizing and autoscaling). (2) Advanced learning solutions for heterogeneous edge environments, including federated and collaborative learning, and split learning to offloading part of complex deep neural networks from the UEs to the edge. (3) RAN management solutions for edge-enabled relay management, network slicing, multi-tier RAN optimization, dynamic functional split and micro-orchestration of RAN functions. (4) E2E AI orchestration for collision-free decision-making.
In SPT4AI, the studied methodologies include: (1) Security and privacy issues through threat analysis and mitigation measures, and tackling vulnerabilities arising e.g. in distributed MIMO. (2) Resilience and robustness of AI models in decentralized edge deployments against attacks that compromise the quality of training data. (3) Solutions for the generalization, formal verification, safe deployment and training, and explainability of AI models.
VERGE consortium has identified an initial list of innovations that the project as a whole expects to develop and that constitute the major areas where the project will contribute beyond the state of the art. These are the following:
- Advanced orchestration of AI-enabled smart city services in distributed edge environments.
- Dynamic computation splitting for real-time services.
- Dynamic computation offloading and split learning between edge and XR devices.
- AI-driven network slicing for XR and for IoT-enabled autonomous tram services.
- Relays with edge computing capabilities for supporting XR services.
- Smart micro-orchestration of disaggregated RAN elements over FPGA platforms.
- Security of AI driven tasks in distributed MIMO (cell-free Massive MIMO).
- Trustworthiness for AI4EDGE framework using security/privacy by design
- AI model generalization ensuring both accuracy and safety
- Causal discovery and hazard prediction using structural causal models.
- Secure edge intelligence empowered by advanced learning solutions.
- Advanced orchestration of AI-enabled smart city services in distributed edge environments.
- Dynamic computation splitting for real-time services.
- Dynamic computation offloading and split learning between edge and XR devices.
- AI-driven network slicing for XR and for IoT-enabled autonomous tram services.
- Relays with edge computing capabilities for supporting XR services.
- Smart micro-orchestration of disaggregated RAN elements over FPGA platforms.
- Security of AI driven tasks in distributed MIMO (cell-free Massive MIMO).
- Trustworthiness for AI4EDGE framework using security/privacy by design
- AI model generalization ensuring both accuracy and safety
- Causal discovery and hazard prediction using structural causal models.
- Secure edge intelligence empowered by advanced learning solutions.