More natural models

[Willem Van Onsem]

Experience with IDP shows that the generated model is not always the most intuitive one...

Example: the jailbreak assignment. When asked if all guards could escape, in the generated model the guards moved from one place to another before moving outside.

A reasonable query could be a query with approximate minimal entropy in the structure thus the functions/predicates are quite monotonic...

[Bart Bogaerts]

What do you mean with "monotonic"?

Should IDP be informed that there is a special type (Time), which are the fluents and which the actions and minimize number of actions/number of fluent-changes?

[Willem Van Onsem]

Wrong terminology, My bad...

A special parameter called time could be useful, but anyhow a function that changes quite frequently regarding its input is not desirable. Fluently is not necessary a good criterion (if the numbers represent for example rooms, fluent changes have no meaning)...

There are several definitions for such a metric... I will look for some good ones.

[Bart Bogaerts]

with fluent-changes I meant changes in time of that fluent... (when fixing all other input parameters)

I think that a functions that changes quite a lot is often exactly what you are looking for. (think of any problem with isomorphisms, think of graph coloring, ...)

Your requests seem to be very focussed on LTC theories, or more general, any kind of planning problems,..

[Willem Van Onsem]

Not necessary: I don't claim this should be the default behaviour of IDP, but it could be a type of inference that can be activated...