Issue1229

Title Isolate the Domain Transition Graph from the RHW Factory
Priority wish Status resolved
Superseder Nosy List clemens, davidspeck, gabi, jendrik, malte
Assigned To Keywords
Optional summary

Created on 2026-07-24.13:44:13 by davidspeck, last changed by clemens.

Messages
msg12141 (view) Author: clemens Date: 2026-07-31.15:20:38
Apologies for not keeping the tracker up to date. We've just merged pull reques thttps://github.com/aibasel/downward/pull/304 concerning this issue. What we ended up implementing is a new class within the landmarks/landmark_factory_rpg_sasp files which is concerned with building domain transition graphs for all variables.

Some things that we considered doing but left them for future work:
- consolidate different implementations of domain transition graphs which exists in various places, e.g., merge_and_shrink/fts_factory.cc, pruning/stubborn_sets_ec.cc and heuristics/domain_transition_graph.(h|cc))
- isolate reachability analysis within (domain transition) graphs, used here to identify more landmarks, but probably exists in slightly different forms in a few other places.
We didn't create new issues for these because nobody was eager enough to start working on these right now.

Thanks to everybody who contributed!
msg12114 (view) Author: davidspeck Date: 2026-07-24.13:44:13
Currently, the RHW factory has its own implementation of a Domain Transition Graph (DTG) which is used to computed atom landmarks based on other atom landmarks of the same variable. The code is rather verbose and we want to move it out of the factory to clean up. While it would be nice to use a generic DTG for multiple use cases (cg heuristic, cea heuristic, stubborn sets, landmarks, etc.), the current DTG implementation does specific things which makes it not usable at the moment.
History
Date User Action Args
2026-07-31 15:20:38clemenssetmessages: + msg12141
status: unread -> resolved
2026-07-24 13:44:13davidspeckcreate