[Boutilier,Reiter,Soutchanski,Thrun]:
``Decision-Theoretic, High-level Robot Programming in
the Situation Calculus'', AAAI-2000;
(but in that paper, it was not allowed to mention
sense actions explicitly in a Golog program.)
Golog programs provide natural constraints to make the search for an optimal policy computationally feasible.
DTGolog = Decision Theoretic Golog
The semantics of a DTGolog program is defined by the predicate
- prog is a Golog program,
- s is a starting situation,
- 
is the optimal conditional policy,
- u is the expected utility of that policy,
- prob is the probability that pol will execute successfully,
- h is a prespecified horizon (a finite number of steps).
An off-line intepreter implementing this
definition will be called with
prog, 
, h.
The
arguments , u, prob will be instantiated by
the interpreter.
Once the off-line DTGolog interpreter computed an optimal policy, the policy can be given to the robotics software to control a real mobile robot.