Heuristics

Heuristic Strategy Families

Heuristics use the same IUlsSolver interface as exact methods, but return UlsSolveStatus.Feasible: they construct a valid plan without claiming an optimality proof.

Baselines and fixed-cycle rules

• LotForLotSolver
• PeriodicOrderQuantitySolver

Average-cost rules

• SilverMealSolver
• SegerstedtReformulatedSilverMealSolver
• LeastUnitCostSolver
• ChiuModifiedLeastUnitCostSolver
• HoChangSolisNetLeastPeriodCostSolver
• HoChangSolisImprovedNetLeastPeriodCostSolver

The two Ho-Chang-Solis methods replace Silver-Meal's calendar-period average by a net average over non-zero-demand periods. nLPC(i) adds the authors' improved tie-breaking stopping condition.

Part-period and EOQ-derived rules

• PartPeriodSimplifiedSolver
• PartPeriodBalancingSolver
• ChiuTingModifiedPartPeriodBalancingSolver
• PattersonLaForgeIncrementalPartPeriodSolver
• WemmerlovModifiedPartPeriodBalancingSolver
• McLarenOrderMomentSolver

McLaren's Order Moment combines an EOQ-derived time-between-orders estimate with a part-period target and a final marginal setup/holding test.

Look-Ahead / Look-Back

• WemmerlovPpbLookAheadLookBackSolver
• WemmerlovModifiedPpbLookAheadLookBackSolver

Marginal-cost rules

• GroffSolver
• FreelandColleySolver

Global merge rules

• KarniMaximumPartPeriodGainSolver

Karni's MPG is different from the forward heuristics above. It starts from a Lot-for-Lot plan and repeatedly removes the globally most attractive replenishment boundary according to the part-period gain criterion.

Common use

All public heuristics use the same call pattern:

IUlsSolver solver = new SilverMealSolver();
UlsSolveResult result = solver.Solve(problem);

For a heuristic, check result.Status == UlsSolveStatus.Feasible.

The stationary-cost heuristics reject incompatible time-varying cost structures instead of silently changing the published rule.