Repair-aware topology penalties for workforce scheduling in OR-Tools CP-SAT

[Soryie]

Hi everyone,

I’m building a workforce scheduling / rota optimisation system using OR-Tools CP-SAT and I’ve been experimenting with behavioural topology control inside the solver.

One problem I ran into was “irrecoverable tail fragments” during ownership allocation.

Example:

• venue open until 21:00
• solver assigns ownership ending at 18:30
• remaining uncovered block becomes 18:30–21:00
• but minimum shift length is 180 minutes
• therefore no legal repair candidate can ever exist

This created structurally impossible residual underfill even though total staffing availability was sufficient.

The approach I ended up testing was:

• penalising candidates that end within min_shift_minutes of close
• favouring ownership structures that naturally reach close or leave legally repairable tails

This significantly improved topology survivability and reduced brittle repair behaviour.

My question:
Has anyone here worked on similar “repair-aware topology” or “future-feasibility-aware” objective shaping in CP-SAT scheduling problems?

I’d be interested in hearing:

• modelling approaches
• objective strategies
• whether this is commonly solved via topology penalties
• or if there are cleaner structural formulations for this class of issue.