Open DillonZChen opened 6 months ago
The state space generator will soon be deprecated. We are currently working on a complete rework of Mimir which will also allow more control about state space expansion. I can recommend using pymimir from its main branch for state space exploration. Note that it supports a smaller PDDL fragment. The future version in the dynamic branch will support PDDL 3.1. with the exception of object fluents.
For learning policies, it would be nice if the state space feature also extracted the optimal actions for each state in a problem on top of the optimal cost to go. Although it is possible to extract optimal actions from the optimal cost values, this takes a bit of code to do.