Digital Twins

• virtualisation of real world entities (physical assets)
• faithful digital replicas
• deeply intertwined with their physical twin

Saracco, R.: Digital twins: Bridging physical space and cyberspace. Computer 52(12), 58–64 (2019)

Dual perspective

• DTs as virtual environment for MAS
• abstraction
• decoupling
• encapsulation of heterogeneity (data, behaviour, protocols, ...)
MAS as augmentation for DTs
• reasoning
• decision making
• autonomy

Separation of concerns

• DTs operate with full knowledge of the local context of their associated PA
• access to data
• available actions
• knows nothing about other PAs
Agents operate with partial knowledge of the global context of the application
• application goals
• available DTs
• knows nothing about how to operate on PA

This separation allows for assigning clear responsibilities during system design!

DTs for agents and MAS

• DTs are akin to artefacts of the A&A metamodel...
• used by agents
• structuring the environment
• mediating interaction
...but also fundamentally different
• bi-directional synchronisation with PAs
• faithful representation of the (physical) dynamics of the PA

Omicini, A., Ricci, A., Viroli, M.: Artifacts in the A&A meta-model for multi-agent systems. Autonomous Agents and Multi-Agent Systems 17(3), 432–456 (2008)

Individual perspective

Agents delegate to DTs tasks requiring knowledge of the local PA context

$$\rightarrow$$

System perspective

• The literature is abundant on what to expect from an individual DT
• Not much is said about networks of DTs
• can DTs be linked together to structure the virtual environment?
• should such structure be hierarchical?
• can it be changed dynamically?
• if so, by the application (agents) or by DTs themselves?

Web of Digital Twins: application-dependent knowledge graphs, dynamically established amongst DTs depending on CPS dynamics and application goals

Ricci, A., Croatti, A., Mariani, S., Montagna, S., Picone, M.: Web of digital twins.
ACM Trans. Internet Technol. (dec 2021). https://doi.org/10.1145/3507909

Agents and MAS for DTs

• Agents as enablers of intelligent behaviours
• domain-independent models available (e.g. BDI)
• encapsulation of (reusable) decision making
• "as-a-service" perspective orthogonal to applications
Agents as orchestrators and mediators
• handling DTs lifecyle
• monitoring and controlling execution
• enacting coordination mechanisms and protocols

Individual perspective

DTs delegate to agents "intelligent" tasks (reasoning, prediction, ...) that require
broader context w.r.t. the local one accessible to DTs

$$\leftarrow$$

System perspective

• Again, the literature is lacking about an open systems perspective:
• who is responsible for DTs lifecycle (e.g. binding to PAs, synching, etc.)?
• is there a "God-like" orchestrator?
• how can DTs interact in meaningful ways towards application goals?
• is DTs aggregation akin to service composition enough?

Agents as mediators (left): DTs express the intended interaction semantics (e.g. CFP), delegating to agents the figuring out of interaction sequences

Agents as orchestrators (right): DTs seamlessly execute while agents take care of
"DevOps" (e.g. creation, linking, migration, scaling, disposal, ...)

Research Directions 1/2

• Cross-fertilisation opportunities revolve around two core ideas:
• adopt DTs as the engineering abstraction to structure and encapsulate CPS resources and dynamics
• adopt agents as the engineering abstraction to encapsulate goal-oriented decision-making

As decision making depends on the CPS state (and viceversa), synergistic exploitation of agents and DTs is more of a requirement than a nice to have!

Research Directions 2/2

• Cognitive DTs: what if the DT can autonomously evolve through simulation and AI-enabled techniques?
• Anticipatory planning: how can DTs and agent cooperate to plan ahead through prediction and what-if analysis?
• Socio-technical systems: how about DTs virtualising people and organisation besides PAs?
• Mirror worlds: how about digital artefacts reified in our physical world (e.g. holograms)?
• Standardisation & interoperability: can DTs standardise agents means of interactions with environments?