Jump to the Comparative Capabilities page.
Jump to the Home page.
PLANNING
"Planning is the generation of an action sequence of action program for
an agent, such as a robot, that can change its environment. The
purpose of a lan is to achieve one or more explicit goals. The
essential inputs for planning are an initial world state, a repertoire
of actions for changing that world, and a set of goals. The form of
the plan is commonly just a linear sequence or acyclic directed graph
of actions, although the full range of programming control structures
are potentially relevant. For planning to be effective, the
environment in which the plan will be executed must be largely
predictable, but need not be completely deterministic. However,
planning will be ineffective for chaotic domains, and an agent can
only react to events."
From Encyclopedia of Artificial Intelligence
When we use the term planning we are refering to the fact of, whether
or not the architecture is capable of planning. Some systems have
actual planning subsystems, in which case it is fairly obvious that
the system can plan. Others are completely reactive and lack any
planning capabilities. Finally, there are that that fall in between,
we have used our discresion and the author's claims in determining
whether such systems are capable of planning.
Architectures that provide planning.