Updated benchmarks
The updated benchmarks can be downloaded here. (The not up-to-date version is still available here)
Some guys have found a few bugs in problem generators (interestingly, only in optimal track..) and some issues in domain models. Below we list the changes.
- Thoughtful: "suit" in now only a predicate name; type name has been changed in "suittype".
- Barman in seq-opt: the generator provided 6 tasks that use non-declared objects in the goals. Such instances have been removed.
- Maintenance in seq-opt: the generator designed 15 unsolvable tasks. They have been removed.
- Tetris in seq-opt: 3 problems were unsolvable: they have been removed.
- action-costs requirements have been added to GED and Floortile.
- conditional-effects added to Cave and Maintenance. Notice that in requirements of these domains there is ":adl", which encompasses also conditional-effects ; including also conditional-effects is redundant, but seems to be helpful for some planners.
- RTAM domain model had one operator "move-in-city" (that is not needed for solving problems) which require a predicate that is not defined in benchmark instances. It has been commented away.
Generators of newly introduced domains can be found here .