How to prevent domain logic violations in Timefold?

[Koe-eigh]

Hi,

When using Timefold to optimize PlaningSolution, how can I make sure domain rules (like invariants) are not violated during planning?

Is it better to prevent invalid states by design (e.g. value range limits or setter checks), or should I rely only on hard constraints?

Thanks!

[greyhairredbear]

I'd say in general it helps a lot when it's impossible for a hard constraint violation to occur because of your design. Creating and performing moves that will result in solutions that violate hard constraints for certain (and having to calculate a new score after a move) wastes time and thus leads to worse solutions.

[kentbill]

Although avoiding solutions that inevitably violate hard constraints may reduce the size of the problem, does it limit the solver's behavioral freedom, making it hard to escape from local optimization?

See the following content in the user manual.

https://docs.timefold.ai/timefold-solver/latest/using-timefold-solver/modeling-planning-problems#valueRangeProviderOnPlanningEntity

There is a NOTE in this chapter:

By limiting the value range specifically of one planning entity, you are effectively creating a built-in hard constraint. This can have the benefit of severely lowering the number of possible solutions; however, it can also take away the freedom of the optimization algorithms to temporarily break that constraint to escape from a local optimum.

[triceo]

There are effectively two kinds of hard constraints - those you hard-code into your domain model, and those you enforce via the score. Deciding which is which IMO boils down to a judgment call.

Personally, I prefer hard-coding whatever I can, and later relaxing the model if I see the need. Modelling these problems requires experimentation, more often than not.