bonetblai
commented
2 years ago You can put this PR on hold for the moment as there is something additional to do. The removal of irrelevant operators must be done in two steps. In the first step, all operators whose preconditions are false given the static atoms are removed. In the second step, all operators that are irrelevant given the goal are removed. This because when you want to preserve irrelevant operators, you still want to execute the first step and disable the second. I'll work on this later and then do the complete PR.
jendrikseipp
Here is the pull request for the :equality keyword. This was more difficult since Pyperplan doesn't support negative preconditions; i.e., differently from effects where each action has an add and del list, preconditions are simply a list of Predicates() so it wouldn't be that easy to accommodate "native" negative preconditions such as
(not (= ?x ?x))).Therefore, I did the following "hacky" trick. When :equality is specified, two new predicates are created: "=" and "not=". Then, each instance of
(not (= ?x ?x))in preconditions is translated into(not= ?x ?x), thus removing the negation. Checks that raise errors for the use of "=" and "not=" in effects, and the use of(not <predicate>)in non-equality preconditions are done. Constructs like(not (not (?x ?x)))and(not (not= ?x ?x))that are semantically right are refused, and constructs like(not= ?x ?x)are accepted.On the other hand, the initial situation of a problem is extended with denotations for "=" and "not=". That is, the initial situation is extended with atoms
(= o o)for each object o, and with atoms(not= o1 o2)for each pair of different objects o1 and o2.I have tried to adhere to the style.