OBJECTIVES

ASSIST-IoT aims at design, implementation, and validation of an open, decentralized reference architecture, associated enablers, services and tools, to assist human-centric applications in multiple verticals. Instances of the architecture will be supported by key enablers, like edge/fog computing, (semi-)autonomy, distributed AI, smart devices, interoperability, Distributed Ledger Technology (DLT) atop a smart network infrastructure, with low latency capabilities, allowing execution of context-aware applications with new interaction interfaces (e.g. AR/VR/MR), etc. The proposed solution will integrate AI-based functions transferring intelligence closer to the edge (data sources), including devices. The reference architecture, and developed enablers, will be validated in three realistic pilots: (i) port automation; (ii) smart safety of workers, and (iii) cohesive vehicle monitoring and diagnostics. The proposed approach is focused on the edge-fog-cloud continuum model. However, to simplify the narrative the word “edge” captures all generic situations, in which data processing takes place in the appropriate location within the IoT ecosystem. This location is as close as possible to sensing/actuating. This means that the “edge node” is the one that was specifically selected, within the continuum, to perform given function(s).

To meet its main goal, ASSIST-IoT will focus on the following Specific, Measurable, Achievable, Realistic and Timely (S.M.A.R.T.) objectives, along with related Key Validation Indicators (KVIs), showing different validation measures regarding objectives completion.

The objectives related to the ASSIST-IoT concept and validation are mentioned below:

Objective 1: Design, implementation and validation of an NGIoT Reference Architecture.

ASSIST-IoT architecture will fully address issues not covered, or deficient covered, by current IoT solutions. The architecture that has been defined for ASSIST-IoT is as follows, composed of different Views and a set of Verticals (addressing properties and cross-cutting concerns):

Device and Edge plane describes the collection of functions that can be logically appointed to physical components of IoT, including, but not limited to, smart devices, sensors and actuators, wearables, edge nodes, as well as network hardware, such as hubs, switches and routers. Note that this plane, like all the others, represents a Functional View. So even though e.g., functions related to self-contained network could be naturally associated with network devices, there is a group of functions that can be identified and separated into functional blocks that belong squarely on the Device and Edge plane. The aforementioned functions include any physical connectivity and interfaces (e.g., Ethernet), low-level security functions (e.g., firewalling). This plane directly interfaces the hardware capable of executing specific functions designed on higher planes.

Smart Network and Control plane manages virtual and wireless aspects of network connectivity. The key functions handled on this plane are encompassed by technologies that deliver software-related and virtualised networks, such as SDN (SD-WAN), NFV, MANO, and anything related to virtualised or self-contained networking. Any direct and logical connection in the communication infrastructure is provided on this plane. The functions on this plane follow the access-network-agnostic approach, in which the network connections are highly flexible. Features, such as dynamic configuration, routing and addressing, and high-level intelligent firewalling help deliver the required flexibility.

Data Management plane handles all functions related to a virtual shared data ecosystem, in which data are acquired, delivered and processed to provide key data-related functions. Those include data interoperability, provenance, fusion and aggregation, but also content-independent functions, such as resilience (e.g., redundancy). Security functions for access grants and trust management also belong to this plane. Moreover, this plane is empowered by semantics and might be supported by judiciously selected DLT.

Application and Services plane crowns the Functional View with end-user and administrative functions and services. It delivers a layer of abstraction that manages functions offered by lower planes. Moreover, it combines them to provide synergistic value for the whole system. Its functions, aided by the Verticals, aim to offer a unified point of access, and provide system-wide intelligence and configuration capabilities. Because of the high level of abstraction, this plane enables the creation of advanced and intelligent applications, including configurable autonomous systems, that benefit from the lower planes, and their interconnection

Objective 2: Definition and implementation of distributed smart networking components.

ASSIST-IoT will identify business drivers, technical requirements and best practices for adoption of smart networks, combining smart connectivity, decentralised edge nodes, AI-based data analytics, efficient distributed computing, and security by design.

The Smart Network and Control plane is in charge of key aspects of the ASSIST-IoT architecture. On the one hand, it is responsible for the connectivity among network elements, aiming at ensuring low latency and resiliency. On the other hand, this plane covers as well the orchestration of virtualised functions, not only for network-related functions (e.g., VNFs for delivering services such as load balancing, firewall, packet inspection, etc.) but also for Next-Generation functions such as data governance, interoperability, privacy, security, and intelligence, among other functionalities.

This plane has been designed following the SDN/NFV paradigm, considering auto-configuration capabilities to provide continuous support for real-time applications. To this end, the plane is composed of four functional blocks, namely (i) smart orchestrator, (ii) SDN controller, (iii) VNFs and (iv) self-contained network. Further information of smart networking components of ASSIST-IoT will be delivered later in deliverables and through this webpage.

Objective 3: Definition and implementation of decentralized security and privacy exploiting DLT.

Security, privacy and data protection by design will be founding assumptions for the ASSIST-IoT to guarantee security, interoperability, confidentiality, integrity and availability. Thus, ASSIST-IoT will introduce a holistic cross-plane solution for security, trust and privacy across NGIoT ecosystems.

An intriguing aspect that ASSIST-.IoT approach will cover is securing decentralised intelligence. Decentralisation and federation (Federated Learning – FL) are interesting key novel concepts in the ASSIST-IoT technological proposition, therefore the architecture is in current research and design  most appropriate ways and include enough provisions to allow the “conceptualization and testing” of DLT-powered Federated Learning in the project.

With regards to the actual implementation, the DLT-based FL techniques will be implemented within the architecture for representative scenarios. This approach could enhance the privacy of data exchanged among the edge nodes when they execute AI functions to extract knowledge from contextual and streaming data within the ASSIST-IoT architecture. More specifically, ASSIST-IoT architecture will foster the use of DLT-related components to exchange the local, on-device models (or model gradients) in a decentralised way avoiding single point of failures acting as a component to manage AI contextual information in an immutable form, and avoiding as well alteration to the data

Objective 4: Definition and implementation of smart distributed AI enablers.

Develop an AI-based framework, by means of modular features (both hardware and software), dynamically deployable to heterogeneous IoT nodes (e.g. devices) in the edge-cloud continuum, combined with smart networking to instantiate applications explored within pilots.

Federated learning (FL) is an approach to train ML models that do not require sharing datasets with a central entity. In FL, a model is trained collaboratively among multiple parties, which keep with themselves their training dataset, but they collaboratively participate in a shared FL process. The notion of parties might refer to entities as different as data centres of an enterprise in different countries, mobile phones, cars, or different companies and organisations.

In ASSIST-IoT, the FL framework has been divided in 5 technical components:

    • FL Orchestrator:The responsible for coordinating the overall Federated Learning process. Hence, it will be in charge of specifying details of FL workflow(s)/pipeline(s), such as job scheduling, FL life-cycle management, or defining stopping criteria for the training.
    • FL Training collector: The FL training process involves that several independent parties commonly collaborate in order to provide an enhanced ML model. In this process, the different local updates suggestions shall be aggregated accordingly. This duty within ASSIST-IoT will be tackled by the FL Training Collector, which will also be in charge of delivering back the updated model.
    • FL Repository: A set of different databases, including initial ML algorithms, already trained ML models suitable for specific data sets and formats, averaging approaches, and auxiliary repositories for other additional functionalities that may be needed, and are not specifically identified yet.
    • FL Local Operations: One of key goals of FL is to assure protection of privacy of data, owned by individual stakeholders. Therefore, data is expected to be trained and used only locally. Therefore, an embedded enabler within each FL involved party/device of the FL systems is needed. It has been defined as FL Local Operations enabler.
    • FL Privacy will guarantee that different parties are not able to derive insights about each other’s training data.

Objective 5: Definition and implementation of human-centric tools and interfaces.

Enablers for implementation of semi-autonomous services, traditionally falling under human responsibility/ liability will be developed. This will assist augmentation of workforce intelligence, via: visualization; decision support; or control of work environment, by tightening collaboration of cyber-physical systems, human beings, and virtual services.

ASSIST-IoT will develop enough software commodities (enablers) for providing immersive experience to practitioners of ASSIST-IoT in general, and for the Worker’s safety pilot in particular as a first action line. The enabler will receive data from the Edge data broker or from the LTS enabler and will transform this data into a suitable format for visualisation capabilities over head-mounted MR displays (in principle, it is foreseen for MS Holo-Lens MR goggles). Information will then be displayed to the user, according to their authorisation/access rights, supporting user interaction with the virtual content and view customisation.

Objective 6: Definition, deployment and evaluation of real-life pilots.

Three industry-driven pilots will demonstrate the value of the architecture, as an innovation tool. Pilots will be validated using heterogeneous data sources in real-life settings and based on actual deployments. ASSIST-IoT outcomes will be validated from both technical and business perspective. More information on the challenges, technical and operational description of the pilots can be found at: https://assist-iot.eu/use-cases/

Objective 7: Establishment of an innovative cooperation and business framework.

ASSIST-IoT will define and validate credible, scalable business models, which will ensure wide and sustainable use of deployed solutions. Business modelling will be fully aligned with technical capabilities of results,  ecosystem, and functionalities of deployed pilots.

Objective 8: Impact creation: Showcasing ASSIST-IoT and Disrupting the current market.

Besides standard dissemination activities, e.g. presenting and promoting the approach and its results, at conferences, website, exhibitions, social media, publications and workshops, ASSIST-IoT will perform several showcases, including small demonstrations, to widely present main outcomes and to show concrete advantages of using ASSIST-IoT architecture and enabling technologies, to stakeholders and potential clients, targeting diverse verticals, mobilising key actors, including the security area.

Pilots formalization, use case analysis and requirements elicitation

Within tasks 3.2 and 3.3 the  ASSIST-IoT  pilots (container  terminal  automation, construction safety and  vehicle  condition  diagnostics and  monitoring) and corresponding pilots’ objectives, have been thoroughly analysed and formalised. To this end, industrial and academic leaders of each pilot collaborated with all the consortium partners to express, in the form of business scenarios, the business needs and solutions to business problems that have to  be  addressed.  Based  on  the  common  understanding  of  the  functionality  of  the  systems  under development, their  functionality  was  documented in the  form  of  use  cases,  which  describe  the  interactions  between  actors  (persons,  devices  or  digital  entities),  the  assumptions  and  the  expected  outcomes  after  the execution of a flow of actions. Further functional and non-functional requirements were identified based on the expertise and experience of all partners. This initial version of the manual includes 19 use cases, 35 functional and 39 non-functional requirements (74 in total; 7 of them common to all pilots).

The analysis of each business domain resulted in more precise definition and refinement of the project’s scope. It also led to better understanding and agreement between partners on how business needs fit existing IoT architectures. Furthermore, it will help define what process improvements, organisational changes and policy developments will be required. The project’s key value indicators and the pilot’s key performance indicators were further decomposed, by defining acceptance criteria to the identified use cases and requirements in order to be able to measure the outcome of the pilots and provide input for the technical evaluation of the project results. The results of these activities have been initially summarised in WP3 deliverables and are briefly outlined in the challenges, technical and operational description of the pilots which can be found at: https://assist-iot.eu/use-cases/.