Proposition 6.5(a) — pointwise optimal policies give an optimal stationary policy in
ProvedBertsekasShreve.Generalized.prop6_5a_pointwise_optimal_stationarycontractiondynamic-programminginfinite-horizonp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1
Consider the generalized abstract model of Section 6.1 under conditions A.1–A.5, Assumption and the exact selection assumption. If for each there is a policy in which is optimal at , i.e.
then there exists an optimal stationary policy , i.e. .
The restricted-class analogue of Proposition 4.3(b): pointwise optimality, possibly by different policies at different states, can be replaced by a single stationary policy from .
Preamble
import Mathlib import Definitions.Def_BertsekasShreve_Generalized_Model import Definitions.Def_BertsekasShreve_Generalized_Assumptions import Definitions.Def_BertsekasShreve_Generalized_Optimality
Formal statement
namespace BertsekasShreve.Generalized
open Filter Topology
/-- Proposition 6.5(a), p. 97. Under A.1–A.5, Assumption C̃ and the exact selection assumption,
if for each `x ∈ S` there is a policy in `Π̃` which is optimal at `x`, then there is an optimal
stationary policy in `Π̃`. -/
theorem prop6_5a_pointwise_optimal_stationary {S C : Type*} (P : Model S C)
(hA1 : P.A1) (hA2 : P.A2) (hA3 : P.A3) (hA4 : P.A4) (hA5 : P.A5)
(Bbar : Set (BertsekasShreve.Contraction.BFun S)) (m : ℕ) (ρ α : ℝ) (hC : P.AssumptionCtilde Bbar m ρ α)
(hES : P.ExactSelection)
(hopt : ∀ x, ∃ π : P.Policy, P.IsOptimalAt π x) :
∃ μ : P.Sel, P.IsOptimal (P.stationary μ) := by sorry
end BertsekasShreve.Generalized
Source
Bertsekas & Shreve, Stochastic Optimal Control: The Discrete-Time Case, Athena Scientific 1996, p. 97, Proposition 6.5(a)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.