Crate u_schedule 

A scheduling framework.

Provides domain models, constraints, validation, dispatching rules, and a greedy scheduler for scheduling problems. This crate defines the scheduling domain language — metaheuristic algorithms (GA, SA, CP) are provided by u-metaheur at a lower layer.

Modules

• models: Domain types — Task, Activity, Resource, Schedule, Assignment, Calendar, Constraint, TransitionMatrix
• validation: Input integrity checks (duplicate IDs, DAG cycles, resource refs)
• dispatching: Priority dispatching rules (SPT, EDD, ATC, etc.) and rule engine
• scheduler: Greedy scheduler and KPI evaluation
• ga: GA-based scheduling with OSV/MAV encoding
• cp: CP-based scheduling formulation

Architecture

This crate depends on u-metaheur and u-numflow, and contains only scheduling domain logic — no nesting, packing, or manufacturing concepts.

Solver enforcement matrix (0.4.0)

Model expressiveness and solver enforcement are distinct. What each execution path actually enforces:

Feature	SimpleScheduler	GA decode	CP builder
Fixed-assignment seeding (pin) 1	✅	❌	❌
Multi-requirement simultaneous hold	✅	❌ (single resource per activity)	❌ (first candidate)
Resource calendar	✅	❌	❌
Capacity > 1	✅	❌	❌
Setup/teardown duration components	✅	❌	❌
TransitionMatrix sequence-dependent setup	✅	✅	❌
Constraint::{TimeWindow, Synchronize}	validated only	validated only	❌ (skipped)
Constraint::TransitionCost	❌ (unsupported)	❌	❌

Every scheduler::SimpleScheduler result self-annotates via scheduler::check_schedule, so Schedule::is_valid() is meaningful on that path. Run the checker manually on GA/CP output to obtain honest violation reports.

References

• Pinedo (2016), “Scheduling: Theory, Algorithms, and Systems”
• Brucker (2007), “Scheduling Algorithms”
• Blazewicz et al. (2019), “Handbook on Scheduling”
• Haupt (1989), “A Survey of Priority Rule-Based Scheduling”
• Kolisch & Hartmann (1999), “Heuristic algorithms for the RCPSP” (serial SGS)

---

1. Pins are opaque reservations — they never update last_category, so the next TransitionMatrix-governed activity placed on the same resource computes its changeover against the category that was current before the pin, understating the true setup from the pin’s task. See scheduler::SimpleScheduler::with_fixed_assignments for the full writeup. Known limitation, not yet addressed: changeover-aware pin accounting — treat a pin as a last_category update so the following activity’s setup reflects a transition from the pinned task’s category. ↩

Modules

cp

CP-based scheduling formulation.

dispatching

Dispatching rules and rule engine for scheduling.

ga

GA-based scheduling optimization.

models

Scheduling domain models.

scheduler

Greedy schedulers and KPI evaluation.

validation

Input validation for scheduling problems.