Integration of interval analysis with constraints parsing (Difficulty: DIFFICULT)

[mike-gimelfarb]

Currently, we have interval analysis that infers bounds on fluents, i.e. state in [100, 500], reward in [-1.0, 5.0]...
We also have a constraints parser that parses inequality constraints on action/state fluents, i.e. action-fluent(?x) >= non-fluent(?x) + 100..
However, we do not currently have a way to integrate them. For example, if bounds on state fluents are not stated as constraints, but clipped in the cpfs? How can they be integrated into the constraints parser to get tighter gym bounds for RDDLEnv?