Theorem 4.3, p. 17 — S*_t(x) ⊆ S*_t(x + 1), and S*_t(x) ⊆ S*_{t+1}(x) when V_t ⊆ V_{t+1}
ProvedRobustMNL.Dynamic.assortment_monotoneIn the robust capacity-allocation model over periods , with compact nonempty uncertainty sets , let denote an optimal assortment of smallest cardinality of the period- Bellman problem with capacity . Then:
- For every and ,
- For every and with ,
More remaining capacity leads to a larger optimal assortment; and, when uncertainty does not shrink over time, so does approaching the end of the horizon. This is the robust analogue of the nesting properties of single-leg revenue management.
Formalization Note Each part holds for every pair of smallest-cardinality maximizers (IsOptAssort). "For any " is read as ( is not defined by the Bellman equation); "for any " is , and in part 2 so that exists. The only hypothesis of part 2 beyond the standing assumptions is . is defined for every , so no capacity bound appears. Revenues are arbitrary reals (the paper's is dropped).
import Mathlib import Definitions.Def_RobustMNL_Dynamic_ValueFunction
namespace RobustMNL.Dynamic
theorem assortment_monotone {n : ℕ} (T : ℕ) (V : ℕ → Set (ℝ × (Fin n → ℝ)))
(r : Fin n → ℝ) (hV : IsUncertaintySeq T V) :
(∀ (t x : ℕ) (S S' : Finset (Fin n)), 1 ≤ t → t ≤ T → 1 ≤ x →
IsOptAssort T V r t x S → IsOptAssort T V r t (x + 1) S' → S ⊆ S') ∧
(∀ (t x : ℕ) (S S' : Finset (Fin n)), 1 ≤ t → t + 1 ≤ T → 1 ≤ x → V t ⊆ V (t + 1) →
IsOptAssort T V r t x S → IsOptAssort T V r (t + 1) x S' → S ⊆ S') := by sorry
end RobustMNL.Dynamic
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.