The field of Arti­fi­cial Intel­li­gence (AI) is devel­op­ing rap­idly, with new emer­ging applic­a­tions that are fil­ter­ing down into dif­fer­ent crit­ic­al sec­tors, such as the digit­al soci­et­ies, industry, trans­ports, health. To reduce elab­or­a­tion time, band­width con­sump­tion and to alle­vi­ate pri­vacy con­cerns, these applic­a­tions are being increas­ingly deployed at the edge of the infra­struc­ture, closer to IoT sensors and devices, where data is ori­gin­ated. While applic­a­tion developers are focused on the accur­acy of their mod­el, the com­ple­tion time or the user exper­i­ence, infra­struc­ture oper­at­ors are keen­er to ensure that the infra­struc­ture – which has lim­ited resources at the edge, with geo­graph­ic­ally-dis­trib­uted, some­times unre­li­able, het­ero­gen­eous devices – can accom­mod­ate as many applic­a­tions as pos­sible, to max­im­ise rev­en­ues and act­ive ser­vices or min­im­ise costs. These diver­ging per­spect­ives and goals are hinder­ing the evol­u­tion of AI mod­els into elast­ic ser­vices that can be deployed and auto­mated on a wide scale.

DECENTER aimed at filling this gap and delivered a mod­u­lar solu­tion to cre­ate AI-based cloud-nat­ive applic­a­tions and oper­ate them over inde­pend­ent or fed­er­ated, het­ero­gen­eous, cloud-to-edge infra­struc­tures.

In oth­er words, DECENTER has provided plat­forms, ser­vices and tools to trans­form AI mod­els into applic­a­tions and man­age them, includ­ing the data they elab­or­ate and gen­er­ate, with the same level of com­plex­ity of “tra­di­tion­al” cloud ser­vices, over an infra­struc­ture that spans from the data centre to resource-con­strained nodes loc­ated at the edge.

The fig­ure above shows the work­flow, in terms of func­tions, envi­sioned in DECENTER: the solu­tion provides ser­vices and tools (in red) to pre­pare AI for a cloud-to-edge deploy­ment, that is, it cre­ates a cloud-nat­ive applic­a­tion from optim­ised mod­els, which can also be exchanged and traded among parties. The cloud-nat­ive applic­a­tion is then orches­trated by means of plat­forms (in blue) into the cloud-to-edge con­tinuum. Data is brokered by means of IoT plat­forms, which are trans­formed into know­ledge bases with digit­al twin func­tion­al­it­ies shar­ing data and infer­ence res­ults across applic­a­tions. Deployed AI applic­a­tions are con­stantly mon­itored to check for poten­tial anom­alies and to ensure that Ser­vice Level Agree­ments (SLAs) are respec­ted, while optim­ising the util­isa­tion of the infra­struc­ture. The fig­ure also shows if the func­tion has been developed mainly in Europe or in Korea.

The tech­no­lo­gies pro­posed by DECENTER are pav­ing the way to the intro­duc­tion of nov­el ser­vices for industry, civil soci­ety and gen­er­al pub­lic. With its use cases, DECENTER demon­strated how AI can be deployed and auto­mated in crit­ic­al cloud-to-edge scen­ari­os, such as smart cit­ies, robot­ics logist­ics, con­struc­tion sites and ambi­ent intel­li­gence, in order to reduce time to react to events, increase pri­vacy and optim­ise costs. 

DECENTER Archi­tec­ture
Click on each com­pon­ent to learn more!

The fol­low­ing pic­ture show the inter­con­nec­tions among the com­pon­ents cre­ated dur­ing the DECENTER pro­ject, and the over­all matur­ity of the plat­form.

Legend:

Innov­a­tion in Fog Inter­net of Things Cloud Arti­fi­cial Intel­li­gence Block­chain Inter­net of Things
Fog Computing Platform to orchestrate Cloud-to-Edge resources

DECENTER will design and devel­op a robust and secure Fog Com­put­ing Plat­form, based on exist­ing Open Source frame­works that provides applic­a­tion-aware orches­tra­tion of resources and zero-touch pro­vi­sion­ing of Microservices for Arti­fi­cial Intel­li­gence applic­a­tions based on their spe­cif­ic require­ments and described by means of Qual­ity of Ser­vice mod­els.

Blockchain-based framework fostering cross-border dynamic binding of resources

DECENTER relies on a frame­work that cata­lyses offer and demand for Edge resources. The innov­at­ive solu­tion is based on Block­chain to draft and seal Smart Con­tracts, mon­it­or their ful­fill­ment and grant rewards.

Smarter Internet of Things fabric

DECENTER enriches Inter­net of Things plat­forms with more intel­li­gent fea­tures able to cope with the data deluge com­ing from mul­tiple data sources includ­ing audio and video cap­able devices with built-in Arti­fi­cial Intel­li­gence algorithms for fea­ture extrac­tion, applic­a­tion-aware data fil­ter­ing redu­cing impact on the use of the net­work.

Hybrid decentralised Artificial Intelligence models exploiting DECENTER’s novel infrastructure

DECENTER evolves decent­ral­ised Arti­fi­cial Intel­li­gence mod­els that become pos­sible with a more react­ive and flex­ible resources infra­struc­ture, unlock­ing a huge poten­tial for innov­at­ive applic­a­tions. Among oth­ers, this innov­a­tion will address the needs for data loc­al­ity and time-crit­ic­al aspects of the Big Data prob­lem.

The table below links the innov­a­tions above to the pro­gress achieved in DECENTER real­ised through a spe­cif­ic archi­tec­tur­al com­pon­ent or research (with ref­er­ence deliv­er­able for fur­ther inform­a­tion), offer­ing obser­va­tions on prac­tic­al feas­ib­il­ity and multi-scope imple­ment­a­tion flex­ib­il­ity.

Pro­ject Research Imple­ment­a­tion Devel­op­ment Integ­ra­tion Val­id­a­tion Stand­ard­iz­a­tion Object­ives

Service Level Agreement

Define resource mod­els and asso­ci­ated Ser­vice Level Agree­ments for in-bor­der and cross-bor­der ser­vices.

Fog Computing platform

Devel­op a robust multi-tier Fog Com­put­ing plat­form to man­age Cloud-to-Edge resources.

Orchestration

Define and imple­ment resource orches­tra­tion strategies to sat­is­fy applic­a­tion require­ments.

Blockchain & Smart Contracts

Imple­ment Block­chain-based Smart Con­tracts and prob­ing mech­an­isms to val­id­ate their ful­fil­ment.

Privacy

Define pri­vacy-pre­serving mech­an­ism for rep­res­ent­ing users’ con­text based on Inter­net of Things data inter­pret­a­tion.

Internet of Things Management

Facil­it­ate inter­op­er­able Inter­net of Things device man­age­ment in the DECENTER infra­struc­ture.

Artificial Intelligence

Devel­op hier­arch­ic­al meth­ods to map Arti­fi­cial Intel­li­gence algorithms on to the Cloud-to-Things con­tinuum.

Data management

Devel­op applic­a­tion data man­age­ment on dis­trib­uted infra­struc­ture for cross-bor­der Arti­fi­cial Intel­li­gence.

Integration & Validation

Integ­rate and val­id­ate DECENTER innov­a­tions, and demon­strate them in real-world pilots.

Standardisation

Pro­mote the stand­ard­isa­tion and indus­tri­al applic­a­tion of the DECENTER res­ults.

DECENTER Leg­acy

As shown in the fig­ure above, the dif­fer­ent com­pon­ents developed dur­ing the DECENTER pro­ject have reached dif­fer­ent levels of matur­ity and integ­ra­tion.

The work of DECENTER will thus be con­tin­ued at dif­fer­ent level, from fur­ther research to pre-com­mer­cial exploit­a­tion, and through future inter­na­tion­al col­lab­or­a­tion.

To sus­tain the col­lab­or­a­tion bey­ond the pro­ject com­ple­tion, the con­sor­ti­um has selec­ted to organ­isa­tions in which the part­ners will con­tin­ue to con­trib­ute and col­lab­or­ate:

Hipeac is a net­work of spe­cial­ist from aca­demia, industry and policy makers. It has a strong expert­ise in advance com­puter archi­tec­ture and com­put­ing sys­tems research and devel­op­ment

After sev­er­al col­lab­or­a­tion, DECENTER has joined the organ­iz­a­tion as a pro­ject mem­ber. The part­ners of DECENTER will con­tin­ue to con­trib­ute to this ini­ti­at­ive.

The Urb­an Tech­no­logy Alli­ance is a non profit asso­ci­ation gath­er­ing cit­ies, aca­dem­ics and indus­tri­als around Smart City solu­tions.

The asso­ci­ation has a large inter­na­tion­al pres­ence in both Europe and Korea, but also Japan, US and Canada.