Machine Learning for Emergent Middleware - Trustworthy Eternal Systems via Evolving, Software Data and Knowledge - page 20

environment by the NSs using legacy discovery protocols (e.g., SLP 4 )thereby
allowing the extraction of basic NS models based on the information exposed
by NSs, i.e., identification of the NS interface together with middleware used
for remote interactions. However, semantic knowledge about the NS must be
learned as it is not commonly exposed by NSs directly.
2. The Learning Enabler specifically enhances the model of discovered NSs with
the necessary functional and behavioural semantic knowledge. The Learning
Enabler uses advanced learning algorithms to dynamically infer the ontology-
based semantics of NSs' affordances and actions, as well as to determine the
interaction behaviour of an NS, given the interface description exposed by
the NS though some legacy discovery protocol. As detailed in subsequent
sections, the Learning Enabler implements both statistical and automata
learning to feed NS models with adequate semantic knowledge, i.e., func-
tional and behavioural semantics.

Partial NS Models

Discovery
Enabler

Learning
Enabler

NS 1 NS 2

NS Models

Monitoring
Enabler

NS 1 N S 2

δ

α β
ɣ

ρ

λ

ρ

ω

Emergent
Middleware

Networked System 2 Networked System 1

Synthesis
Enabler

Fig. 3. The Enablers supporting Emergent Middleware

3. The Synthesis Enabler dynamically generates the software (i.e., Emergent
Middleware) that mediates interactions between two legacy NS protocols to
allow them to interoperate. In more detail, once NS models are complete,
initial semantic matching of two affordances, that are respectively provided
and required by two given NSs, may be performed to determine whether the
two NSs are candidates to have an Emergent Middleware generated between
them. The semantic matching of affordances is based on the subsumption
relationship possibly holding between the concepts defining the functional
semantics of the compared affordances.
Given a functional semantic match of two affordances, the affordances' be-
haviour may be further analysed to ultimately generate a mediator in case
of behavioural mismatch. It is the role of the Synthesis Enabler to analyse
the behaviour of the two affordances and then synthesise—if applicable—the
mediator component that is employed by the Emergent Middleware to en-
able the NSs to coordinate properly to realise the given affordance. For this,

4 http://www.openslp.org/