Environmental Engineering Reference
In-Depth Information
system self-management, requiring automation of responsibility including
some decision making for the successful operation of the system). Thus, au-
tonomicity will often be in addition to self-governance to meet the system's
own required goals. It may be argued at the systems level that the success
of autonomy requires successful autonomicity. Ultimately, ensuring success in
terms of the tasks requires that the system be reliable [
158
].
For instance, the goals of a system may be to find a particular phe-
nomenon using its onboard science instrument. The system may have auton-
omy (the self-governance/self-direction) to decide between certain parameters.
The goals to ensure the system is fault tolerant and continues to operate under
fault conditions, for instance, would not fall directly under this specific dele-
gated task of the system (its autonomy), yet ultimately the task may fail if the
system cannot cope with uncertain dynamic changes in the environment. From
this perspective, the autonomic and self-management initiatives may be con-
sidered specialized forms of autonomy, that is, the autonomy (self-governance,
self-direction) is specifically to manage the system (to heal, protect, configure,
optimize, and so on).
nomic, autonomous essentially means “self-governing” (or “self-regulating,”
“self-directed”) as defined [
109
,
162
]. “Autonomic” is derived from the noun
“autonomy,” and one definition of autonomous is autonomic, yet the main dif-
ference in terms of the dictionary definitions would relate to speed; autonomic
being classed as “involuntarily,” “reflex,” and “spontaneous.”
“Autonomic” became mainstream within computing in 2001 when IBM
launched their perspective on the state of information technology [
63
]. Auto-
nomic computing has four key self-managing properties [
69
]:
•
Self-configuring
•
Self-healing
•
Self-optimizing
•
Self-protecting
With these four properties are four enabling properties:
Self-aware
:
of internal capabilities and state of the managed component
Self-situated
:
environment and context awareness
Self-monitor and self-adjust
:
through sensors, effectors, and control loops
In the few years since autonomic computing became mainstream, the “self-x”
list has grown as research expands, bringing about the general term “selfware”
or “self-
,” yet the four initial self-managing properties along with the four
enabling properties cover the general goal of self management [
156
].
The tiers for Intelligent Machine Design [
101
,
136
,
139
,
140
] consist of a
top level (reflection), a middle level (routine), and a bottom level (reaction).
Reaction is the lowest level where no learning occurs, but is the immediate re-
sponse to state information coming from sensory systems. Routine is the mid-
dle level, where largely routine evaluation and planning behaviors take place.
∗
