Environmental Engineering Reference
In-Depth Information
5
Core Technologies for Developing Autonomous
and Autonomic Systems
This chapter examines the core artificial intelligence technologies that will
make autonomous, autonomic spacecraft missions possible. Figure
5.1
is a
pictorial overview of the technologies that will be discussed. The
plan
tech-
nologies will be discussed first followed by the
act
and
perceive
technologies,
and finally technologies appropriate for testing.
It is dicult to make definitive statements on the functionality, strengths,
and weaknesses of software systems in general, since designers have tremen-
dous latitude on what they do. This chapter explains and discusses the at-
tributes seen in a majority of the systems described. It should be understood
that exceptions may exist.
5.1 Plan Technologies
The planning portion of the autonomy cycle is responsible for examining the
environment and choosing appropriate actions in light of the goals and mission
of the system. Sometimes this choice requires interactions with other systems.
Planners are a central technology in all computerized planning, and many
techniques have been developed to support planning, such as formal collab-
oration languages, evidential reasoning, and learning techniques. The rest of
this section will discuss these techniques and planning in general.
5.1.1 Planner Overview
A defining characteristic of an autonomous system is the ability to indepen-
dently select appropriate actions to achieve desired objectives. Planner sys-
tems are the software component commonly used to achieve this capability.
Work on software planners goes back to 1959 and, over the intervening years,
many types of planners have been developed.
Figure
5.2
shows a high level view of a planner and its context in a system
architecture. All planners begin with a set of initial mission objectives that are
