IEEE Computational Intelligence Magazine - November 2019 - 72

other peers' controllers in order to effectively orchestrate and distribute intelligence. The Intelligence Controller must
also be able to run local control loops
and dynamically update the data flows
(inbound and outbound). The Intelligence Service manager executes operations such as FGIS composition, security
access enforcement and multi-purpose
policies negotiation. Additionally, it
accomplishes the crucial task of interfacing the intelligence and application layer
by setting up and handling this purpose's
specific functions.The FGIS may be hardcoded with a hardware implementation
or controlled by the Intelligence Controller. The FGIS composition is performed
by the Intelligence Service Manager's service composition functionality.

D. Binding Applications and
Intelligent Services

As explained before, the intelligence
layer delivers an intelligent service to
any application that has requested it. In
practical terms, the intelligent algorithm's output must comply with context and semantics that the application
can interpret and use in its own functions. This compliance can be ensured
by allowing the processed intelligence to
access to the application through a
microservice-based approach [20]. Specifically, the application layer gets access
to a dedicated API to connect to the
output delivered by the intelligence
layer. When defining such an API, additional aspects must be considered. For
example, many use cases include the

Intelligence Service Manager

Intelligence Layer
Intelligence Service
Manager
Intelligence Controller
Fine-Grained Intelligence
Service
Atomic Intelligence
Service 1
Atomic Intelligence
Service 2
Atomic Intelligence
Service 3

execution of several intelligent applications. Additionally, such use cases rely on
systems composed of several subsystems,
which in turn respond to specific functionality and requirements for appropriate processing, input, and actions. A
clear example of system complexity is
given by a modern car. In a car, there
are platforms dedicated for managing
engine and mechanical systems, while
others focus on the passenger's entertainment and comfort. Although such
subsystems belong to the same car's
integral system, each must interpret
very different input and output information and processing requirements.
Regardless of this specific example,
the way in which intelligence is handled
can differ from case to case. A single

It features several sub-components and provides a wide set of functionalities.
* It exposes intelligent services to the overlying application (that is, it abstracts the Application
Intelligent Service).
* It manages the service Lifecycle Management (e.g., by handling the intelligence over
the lifetime of a service).
* It executes Device Capability Discovery (e.g., gathers information from the intelligence
controller based on the available devices' capabilities).
* It performs the operation of Service Composition (e.g., composition of different Atomic
Intelligence Services). The service composition is also affected by the nature of the
(Abstracted) Application Intelligence Service and of the Device Capability Discovery.
Intelligence Controller
* It controls the FGIS based on parameters and limitations set by the Intelligence Service
Manager.
* It may expose hardware device capabilities to the Intelligence Service.

FIGURE 6 Intelligence Layer key components high-level description.

Fine-Grained Intelligent
Service
Intelligence Service
Manager

Fine-Grained Intelligent
Service

Intelligence Controller

Atomic Intelligent Service

Capable Device

Constrained Device

Cloud/Centralized Domain

Edge/Fog Domain

Device Domain

FIGURE 7 Example of Interactions between Intelligent Layers from several domains in a specific IoT deployment.

IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE | NOVEMBER 2019

IEEE Computational Intelligence Magazine - November 2019

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - November 2019