Page 254 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

246 CHAPTER 12 Computational Intelligence in the Time

interacting with the physical world are providing the technological framework to
support such epochal shift in the humanemachine interaction.
A CPS is typically composed of a network of heterogeneous units strongly
interacting with the physical environment they are deployed in. In CPS individual
units interact with the physical world, creating the basis for “smart solutions,” which
revolutionize scenarios from health to industry, from workplace to home, from
transportation to entertainment, ultimately leading to an enhanced quality of life.
Designing such systems means actively participating in a new “digital revolution”
that enables augmented interaction with the real world.
CPSs are present at home and at work and provide the core technologies to
design smart homes, buildings, and cities; enable the Internet of Things (IoT);
support smart energy production; environmental protection; precise agriculture,
management, and metering; and facilitate smart transportation and healthcare,
just to provide a very concise list. The expected evolution of the field, as also
perceived by the industry [1,2], will focus on the integration of hardware and
software technologies to support application reconfiguration, enable autonomous
operations, make native the access to the Internet, and extend the usage and
operational models by introducing intelligent resources and application manage-
ment mechanisms.
We all agree that addressing fundamental architectural, technological, and
standard challenges, for example, the energy consumption of the transistor, the
communication level, and the IPV(6) protocol, will improve the unit’s efficiency
and networking ability [3]. However, fundamental advances in highly performing
hardware per se will not be enough to drastically change the way embedded
applications impact our lives. In fact, we should also design methodologies that,
by adaptively optimizing the use of existing embedded resources, provide the
application with intelligent functionalities granting adaptation abilities to environ-
mental changes, adaptive fault detection and mitigation facilities, and sophisticated
adaptive energy-aware modalities to prolong the system lifetime [4]. Intelligence is
also needed to prevent attacks from malicious software (malware) designed to steal
sensitive information from users, take control of systems, impair or even disable the
functionalities of the attacked devices, distort money from users, and so on.
Recently, malware Mirai was shown to be able to take control of IoT devices and
carry out a major cyberattack [5]; this represents a vulnerability issue that has to
be promptly addressed.
Current research addresses intelligent aspects mostly as independent research
lines either without any functionality harmonization effort or with little emphasis
on how to integrate the differentdchallengingdfacets of these fields within a solid
framework. Moreover, not rarely, strong assumptions are made to make derivations
amenable at the cost of a high loss in performance, efficiencydand sometimes
credibilitydwhenever real-world applications are taken into account. In particular,
we assume infinite energy availability, in the sense that energy and power consump-
tion is not an issue, stationarity/time invariance, implying that the process
generating sensor data (the physical world and the interaction with the sensor
transducer) is not evolving with time, correct data availability, claiming that