|Vision||The Industrial Internet of Things is currently the segment of IoT that supports the use-cases with the most proven ROI (REF). It is anticipated that IIoT will evolve to support the above-listed industrial challenges. This evolution will leverage innovations in other digital technologies that are very closely related to IoT, such as Artificial Intelligence (AI), Augmented Reality (AR), Artificial Vision (AV) and others. These advances are acknowledged, yet our agenda is primarily focused on Industrial Internet of Things (IIoT). Specifically, the following outlines how IIoT should evolve to support the challenges for Sustainable Manufacturing. Research & Innovation (R&I) directions on realising and validation the following evolutions are expected to be further discussed within AIOTI and with other relevant stakeholders|
Research & Innovation Directions for IIoT to support Sustainable Manufacturing
1. Massive Decentralisation of the IIoT Paradigm over Large Numbers of Smart Objects
State-of-the-art edge/cloud IoT architectures (e.g., based on the Industrial Internet Reference Architecture (IIRA)) rely on centralised cloud coordination and make very limited provisions for decentralised interactions between federated semi-autonomous edge systems (i.e., Smart Objects (SOs), Cyber Physical Systems (CPS)). To address the presented industrial challenges (e.g., scaling-up of cobots and CPS systems) new architectures are required to support applications where SOs and CPS systems will coordinate their interactions and data exchanges autonomously, close to the field, and in a decentralised way, without relying on a centralized middleware infrastructure. Such interactions raise also new interoperability challenges: federated edge systems should be able to discover themselves and to perceive the world based on common semantics to be able to (inter)act autonomously. These challenges must be addressed outside centralised edge/cloud architectures where common data and services standards (e.g., the Next Generation Service Interface-Linked Data (NGSI-LD) used by FIWARE platforms) can be enforced. The implementation of new massively decentralised architectures is likely to combine conventional cloud/edge middleware concepts with new decentralised middleware approaches such as blockchains for orchestrating decentralised interactions across different CPS systems.
2. Blending of IIoT into emerging 5G/6G networks
Many of the presented industrial challenges require very fast network performance at the edge. This is for example the case with remote manufacturing support applications that leverage AR functionalities. 4G communications have limitations when it comes to supporting real-time, low-latency interactions between federated edge systems (e.g., a swarm of drones in a plant), especially for applications that process or exchange rich media (e.g., AR). One of the challenges of the IIoT is to take advantage of 5G communications to ensure high-speed access to IIoT resources. 5G communications can help managing network QoS (Quality of Service) (e.g., guaranteeing bandwidth to IoT applications) and selecting the objects that shall collaborate as part of the decentralised control applications at scale. Nevertheless, state-of-the-art 5G functions, i.e., Virtualized Network Functions (VNF) and their orchestration are not tailored to the needs of heterogeneous IIoT environments that operate at the edge of the network. There is a need for network slicing approaches that address collections of IoT devices at the edge. Furthermore, to support industrial processes and manufacturing applications at scale (e.g., hundreds of cobots in a plant) some 6G features – e.g., simultaneous wireless connectivity that is order of magnitude higher than 5G; ultra-long-range communication with very low latency – will be required. Overall, a combination of 5G/6G networks will be required to support the next generation of IIoT applications that will provide decentralised control over large numbers of Cyber Physical Production Systems (CPPS).
3. Blending AI into the IIoT – Towards Safe, Trusted and Effective Artificial Intelligence of Industrial Things
To achieve flexibility, intelligence and autonomy at scale, an effective integration of leading-edge AI technologies with IIoT is required. This integration with result in a trusted and safe AI within industrial devices. Specifically, the main aspects of this integration will include:
• Data4AI Distributed Spaces
Artificial Intelligence effectiveness strongly depends on the adequacy and quality of input data. CPSs should improve their capability of storing and processing in real time basic Data4AI functions such as data quality, aggregation and filtering. Industrial Data Spaces are often defined as complex ecosystems where Data Engineering artefacts such as FAIR datasets, standard data models and industrial ontologies are entangled with Hw/Sw Engineering artefacts implemented in Industrial Data Platforms, data management, processing, analysis, sharing. In an IIoT context, Data4AI spaces need to be deployed and operated at the far edge of an industrial system, which implies a highly distributed and intelligent data storage system as well as an open and interoperable industrial data platform.
• Embedding intelligence in the Things
CPS systems should exhibit their own intelligence independently from AI deployment in the cloud/edge. This embedded intelligence will take different forms:
o Embedded intelligence in Smart Objects with semi-autonomous behaviour like industrial robots. SOs will be equipped with advanced forms of AI such as deep learning and reinforcement learning. Its operation will be empowered by in-network analytics and intelligence (e.g., as part of 5G/6G networks).
o Embedded intelligence in CPU constrained devices like machine parts and tools. In this direction embedded forms AI like TinyML will be exploited.
• Trusted and Acceptable Intelligence:
All types of IIoT should incorporate AI that is trusted, responsible and ethical, fully in-line with the mandates of the High Level Expert Group (HLEG) on AI . In this direction, IIoT applications and devices should incorporate transparent and explainable AI models, i.e., Explainable Artificial Intelligence (XAI). Industrial things should be able to illustrate their behaviours to manufacturing workers and other human operators (e.g., why they classified a product as defected or why they recommend maintenance for a piece of machinery).
• Distributed AI for Industrial Things
Federated edge systems (e.g., SOs, CPS systems) have their own computational capabilities, which offers opportunities for real-time AI at the edge of the network (i.e., real-time decisions at the shopfloor). Moreover, they enable novel forms of distributed AI that provide increased privacy and coverage of greater areas such as Federated Learning. However, there are still challenges to distributed AI at the edge such as the need to reason over Semantic Knowledge Graphs (SKGs) of federated systems towards implementing advanced intelligence and coordination between them (e.g., cognitive planning of a cohort of industrial robots or cobots).
4. Symbiotic Human-Machine Collaboration Supporting Tactile Internet Paradigms based on IIoT Systems and Applications
The real time exchange of information between humans and smart objects will blur the lines between the physical and virtual worlds. It will enable remote real-time physical interactions between objects/machines and humans, i.e., eliminating time and space barriers in Human-Machine Interactions. Coupled with technologies like Augmented Reality (AR), next generation IoT applications will facilitate remote human-centric, immersive, and ergonomic interactions between smart objects and humans (e.g., it will be possible to remotely control machines and equipment as part of maintenance and repair processes, based on remote, AR-based immersive experiences). This is a foundation for supporting the remote maintenance and remote manufacturing support challenges.
5. Business Models for Next Generation IoT Systems
There is a need for exploring and validating business models for the next generation of decentralised industrial systems, embedding not just IoT/AI automation and control technologies, but also advanced capabilities for interoperability and human interaction. The latter introduce changes in the business roles of the various stakeholders and provide monetisation opportunities for end-users that are willing to share their data with IoT systems. The research shall define the roles of various business actors (e.g., OEMs, network operators, IoT services developers, IoT services providers, end-users) in the emerging IIoT landscape, along with the business interactions between them. New incentives for Data and Knowledge sharing need to be put in place, both at the Organisational and Individual level, with the support of advanced technologies which could enforce and monitor the respect of data sharing contracts in an overall European Legal and Ethical framework. In the former case, data sharing organisations need to be given full control on the access and the usage of their data, at any level including the far edge (Data Sovereignty). In the latter case, knowledge sharing individuals should be re-assured about IPR principles observation and privacy preservation (GDPR) at the IIoT workplace. In both cases, an incentive and compensation system should be implemented.