Knowledge Types of RALPH-MEA
Domain knowledge for an execution architecture can
be divided into six types, depending on the relationships encoded:
- Type A knowledge provides information about certain factors of
the current state given particular values for other current aspects. For
instance, knowing that the battery has failed can help us decide whether
the fuel gauge may also be low.
- Type B statements represent the effects of actions in terms of
the dependency of the future state on the current state and the chosen
action.
- Type C information provides utility functions on future states.
- Type D knowledge stores the best action choice given the
current state.
- Type E data specify the utility of a certain action if
performed in the current state.
- Type F statements represent goals by specifying a high utility
on a desired future state.
Knowledge of types A and B represent conditional probabilities, so as to
account for the uncertainty of a dynamic
and unpredictable world. Statements of
types C, D, E, and F represent utility functions, conditional on a future
state for C and F or on the current state and an action choice for D and E.
The conditionals do not consider the entire history of the agent, making
the assumption that only the current state is
relevant.
To return, press HOME.
To go to the next document, press NEXT.