I have an overconstrained planning case.
My constraints setup uses a HardMediumSoftScore with several hard constraints, a medium constraint which penalizes the nullable variables to encourage the planner to assign them, and various soft constraints. This seems to work well in terms of finding a solution.
However, I would now like to obtain indictments from the planning score. All the interesting information is in the hard constraints, which prevented the nullable variables from being allocated i.e. a feasible solution with 1000 nullable values remaining ends up with a score like 0 hard/-1000 medium/x soft.
Since none of the hard constraints were violated in this feasible solution, the indictments just list the medium constraint on the nullable variable, which doesn't help me. What I really want to use the indictments for is to determine which hard constraints were most "important" in preventing the allocations to the nullable variable.
If the penalty on the nullable variable is a hard constraint itself (with a "harder" score than the other hard constraints), this "works" and I see the relevant indictments for the lesser hard scores, but now the planner is not able to create feasible solutions.
How can this be solved?