The syntax for problem definitions has been extended to allow for the specification of a probability distribution over initial states, and also to permit the association of a one-time reward with entering a goal state. It is otherwise identical to the syntax for PDDL2.1 problem definitions.
problem | ::= | ( define ( problem name ) | ||
( :domain name ) | ||||
[ require-def] | ||||
[ objects-def] | ||||
[ init] | ||||
goal ) | ||||
objects-def | ::= | ( :objects typed list (name) ) | ||
init | ::= | ( :init init-el* ) | ||
init-el | ::= | p-init-el | ||
| | ^{:probabilistic-effects} ( probabilistic prob-init-el* ) | |||
p-init-el | ::= | atomic formula (name) | ^{:fluents} ( = f-head (name) number ) | ||
prob-init-el | ::= | probability a-init-el | ||
a-init-el | ::= | p-init-el | ( and p-init-el* ) | ||
goal | ::= | goal-spec [ metric-spec] | metric-spec | ||
goal-spec | ::= | ( :goal GD ) [( :goal-reward ground-f-exp )] ^{:rewards} | ||
metric-spec | ::= | ( :metric optimization ground-f-exp ) | ||
optimization | ::= | minimize | maximize | ||
ground-f-exp | ::= | number | f-head (name) | ||
| | ( binary-op ground-f-exp ground-f-exp ) | |||
| | ( - ground-f-exp ) | |||
| | ( total-time ) | total-time | |||
| | ( goal-achieved ) | goal-achieved | |||
| | ^{:rewards} reward fluent |
Håkan L. S. Younes