THIS PROJECT HAS RECEIVED FUNDING FROM THE EUROPEAN UNION’S HORIZON 2020

RESEARCH AND INNOVATION PROGRAMME UNDER GRANT AGREEMENT № 857191

IoTwins ProjectDistributed Digital Twins for Industrial SMEs: a Big Data Platform

W o r k s h o p w i t h I C T 1 1 p r o j e c t s - H P C , B i g D a t a , I o T a n d A I f u t u r e i n d u s t r y - d r i v e n c o l l a b o r a t i v e s t r a t e g i c t o p i c s ( p a r t 2 )

@ B D V A , 0 3 / 0 7 / 2 0 2 0 – F o l l o w - u p w o r k s h o p

Paolo Bellavista

D e p t . C o m p u t e r S c i e n c e a n d E n g i n e e r i n g ( D I S I ) , U n i v e r s i t y o f B o l o g n a

2

The Project.

3

4

HYBRID and DISTRIBUTED Digital Twins Concept in IoTwins

5

Distributed Training and Control in IoTwins

Future Industry-driven Collaborative Strategic Topics.

6

7

Future industry-driven collaborative strategic topics.

Questions:

Data: avoidance of moving data, data curation and anonymizing, cross-IT infrastructure

management

HPC/Cloud infrastructure to edge: data movement, data sharing and orchestration, use of

blockchain technology, cybersecurity, importance for industry (digital Twin context)

Stimuli to discussion:

Distributed digital twinsDistributed digital twinsDistributed digital twinsDistributed digital twins

No data migration for better ownership, latency reduction, better sustainability (not only economic…)

Distributed cloud continuum infrastructureDistributed cloud continuum infrastructureDistributed cloud continuum infrastructureDistributed cloud continuum infrastructure

Distributed orchestration in open and portable solutions

HPC for hybrid digital twins (complex simulations, …)

HPC for resource-greedy learning phases in distributed, federated, reinforcement, … learning

8

Future industry-driven collaborative strategic topics.

Questions:

Workflows: approaches for mastering the complexity and orchestration. Would a reference

architecture of workflows be helpful?

AI/ML, training: need for automated ML, distributed training, explainability of the decision-

making process

Stimuli to discussion:

Standardized workflow architecture and orchestrationorchestrationorchestrationorchestration could be useful, in particular with

innovative distributed challenges in mind

Distributed cloud continuumDistributed cloud continuumDistributed cloud continuumDistributed cloud continuum

In many application scenarios (smart cities, …, but also I4.0 with data ownership

requirements), need for distributed training distributed training distributed training distributed training and better explainability

Distributed, federated, reinforcement, … learning in the distributed cloud continuum

9

Distributed Cloud Continuum and Small Data Ecosystem Building.

Making the cloud continuum an industrial realityMaking the cloud continuum an industrial realityMaking the cloud continuum an industrial realityMaking the cloud continuum an industrial reality

Interoperability and common APIs

Distributed and portable orchestration

Generating trust around the idea of an EU-based cloud continuum, in particular in some specific vertical domains

Extracting value also from “small data” Extracting value also from “small data” Extracting value also from “small data” Extracting value also from “small data” (D. Estrin, Cornell)

by building and promoting the emergence of communities, ecosystems, … fueled by fueled by fueled by fueled by

companies in the manufacturing domaincompanies in the manufacturing domaincompanies in the manufacturing domaincompanies in the manufacturing domain

10

Distributed Cloud Continum and Small Data Ecosystem Building.

Future challenges are future opportunities!

An EUAn EUAn EUAn EU----based cloud continuum, e.g., for the manufacturing industrybased cloud continuum, e.g., for the manufacturing industrybased cloud continuum, e.g., for the manufacturing industrybased cloud continuum, e.g., for the manufacturing industry

Interoperability and common APIs

Distributed and portable orchestration

Support for quality requirements, such as latency, reliability, scalability, …Support for quality requirements, such as latency, reliability, scalability, …Support for quality requirements, such as latency, reliability, scalability, …Support for quality requirements, such as latency, reliability, scalability, …

Integration with resource slicing, 5G/6G, Time Sensitive Networking, … Integration with resource slicing, 5G/6G, Time Sensitive Networking, … Integration with resource slicing, 5G/6G, Time Sensitive Networking, … Integration with resource slicing, 5G/6G, Time Sensitive Networking, …

Generating trust around the idea of an EU-based cloud continuum, in particular in some specific vertical domains

Extracting value also from “small data” Extracting value also from “small data” Extracting value also from “small data” Extracting value also from “small data”

Specialization national/EU districts and the emergence of communities, ecosystems, … which

allow also SMEs to reach “the critical mass” for their specific sub-domain

11

From IoTwins project perspective, next 2-4 years?

Questions:

How does each field impact your project? Specifically, describe how to incorporate HPC

processing in your use cases.

What are your projects’ plans to provide solutions to these challenges?

Stimuli to discussion:

HPC for hybrid and distributed digital twins

Examples: simulations of machine tool spindles and closure manufacturing machines

HPC for training

Examples: wind turbine predictive maintenance, holistic supercomputer facility management

Modular platform infrastructure to reduce SME barriers to access these KETs, scalability also

towards simpler and more limited solutions

CloudCloudCloudCloud---- and distributed cloudand distributed cloudand distributed cloudand distributed cloud----oriented perspectiveoriented perspectiveoriented perspectiveoriented perspective

12

From IoTwins project perspective, next 2-4 years?

Prioritize the four fields in terms of complexity and importance for R&I calls in Europe and pls.

explain your decision

Distributed cloud continuumDistributed cloud continuumDistributed cloud continuumDistributed cloud continuum

Distributed machine learning over distributed cloudDistributed machine learning over distributed cloudDistributed machine learning over distributed cloudDistributed machine learning over distributed cloud

Sustainable ecosystems for small data communitiesSustainable ecosystems for small data communitiesSustainable ecosystems for small data communitiesSustainable ecosystems for small data communities

QoS guarantee or control for the I4.0 domainQoS guarantee or control for the I4.0 domainQoS guarantee or control for the I4.0 domainQoS guarantee or control for the I4.0 domain

What could be specific contributions of your project partners or other institutions in Europe in

each of these areas?

See the previous slides…

13

Key topics, state of the art, and current limitations (1).

Big Data analytics and AI techniquesBig Data analytics and AI techniquesBig Data analytics and AI techniquesBig Data analytics and AI techniques have an unprecedented chance to bring EU

manufacturing companies (product companies, but not only…) into the world of services and

digital business

The Big Data impact and evolution could be extraordinarily amplified in manufacturing if

coupled with proper cloud continuum solutionsproper cloud continuum solutionsproper cloud continuum solutionsproper cloud continuum solutions

to reduce latencylatencylatencylatency

to support prompt/reliable distributed controlprompt/reliable distributed controlprompt/reliable distributed controlprompt/reliable distributed control

to improve scalabilityscalabilityscalabilityscalability

to improve sustainabilitysustainabilitysustainabilitysustainability

to enable better privacy and raw data ownershipprivacy and raw data ownershipprivacy and raw data ownershipprivacy and raw data ownership

…

Need for more distributed and more explainable AI techniquesmore distributed and more explainable AI techniquesmore distributed and more explainable AI techniquesmore distributed and more explainable AI techniques, first of all for distributed

learning and distributed classification/anomaly detection/control

14

Key topics, state of the art, and current limitations (2).

Short-term barriers that need to be reduced, in particular for SMEs:

Complex and rapidly evolving tools and techniques to be mastered delays and costs in product/process design, deployment, test, and refinement

Deep learning require access to very large sources of curated datavery large sources of curated datavery large sources of curated datavery large sources of curated data, as well as significant significant significant significant computational resources for trainingcomputational resources for trainingcomputational resources for trainingcomputational resources for training

Need to be at the premises of the systems Need to be at the premises of the systems Need to be at the premises of the systems Need to be at the premises of the systems generating the big data, e.g., to locally monitor, control, and adapt the components of a manufacturing production line under tight latency and reliability requirements, while preserving an adequate degree of data privacy

Need of investments Need of investments Need of investments Need of investments in infrastructure at the server side (where relevant cloud/HPC resources are often needed for model learning and simulation), at the edge side (e.g., to extend manufacturing machinery and their gateways on the industry plant premises with edge computing functionality), at the communication infrastructure side (5G/6G, Time Sensitive Networking, …), and also in terms of integration efforts

15

More general and ambitious concrete actions.

What the EU, the BDVA, our scientific community can do to stimulate the emergence and stimulate the emergence and stimulate the emergence and stimulate the emergence and

consolidation of “small data” ecosystemsconsolidation of “small data” ecosystemsconsolidation of “small data” ecosystemsconsolidation of “small data” ecosystems?

Are we ready for an EUEUEUEU----centric cloud continuumcentric cloud continuumcentric cloud continuumcentric cloud continuum?

Which could be the role of national competence centers and

EU Digital Innovation Hubs in that?

…

More technically oriented:

Support research on federated learning and control applied to I4.0federated learning and control applied to I4.0federated learning and control applied to I4.0federated learning and control applied to I4.0

Support research on cloud continuum oriented to SLA on quality cloud continuum oriented to SLA on quality cloud continuum oriented to SLA on quality cloud continuum oriented to SLA on quality requirements, integration

with Time Sensitive NetworkingTime Sensitive NetworkingTime Sensitive NetworkingTime Sensitive Networking, virtualization/isolation with guaranteed execution propertiesguaranteed execution propertiesguaranteed execution propertiesguaranteed execution properties,

…

Definitely an open list for discussion…

www.iotwins.eu

@IoTwins_EU

info@iotwins.eu

Contacts.

Paolo Bellavista

Professor of Distributed and Mobile Systems

DISI – University of Bologna

BI-REX I4.0 Competence

paolo.bellavista@unibo.it

http://unibo.it/sitoweb/paolo.bellavista/en

Thank you.