Decision-Theoretic Control
in RALPH-MEA
Decision-Theoretic Control
in RALPH-MEA
The DT architecture uses standard decision-theoretic
reasoning to determine the ideal action,
using knowledge of types A, B,
and C, and the MEU principle.
Type A knowledge describes the current state, type B
knowledge describes the effects of proposed actions, and type
C knowledge evaluates the utility of new states.
The MEU principle calculates the best decision D
based on the utility of states that are the effect of D
and the probability of those states occurring if D is
executed in the current state.
Chaining type B knowledge together provides the ability to look ahead
several steps. Once the architecture generates an action, a
learning component can compile the knowledge for use by other
architectures.
Return to the top of this architecture.
Current Location: RALPH-MEA - Decision-Theoretic Control