Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 5(b) — for D = ∏ₛ D(s), L_D V = V has a unique bounded solution, the robust value over Markov policies with rules in D

Proved
RobustDP.Discounted.theorem5b_fixed_point

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

bellman-equationdynamic-programmingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1robust-optimization

Let (S,A,P,r,λ)(\mathcal S,\mathcal A,\mathcal P,r,\lambda)(S,A,P,r,λ) be a discounted ambiguous MDP, and for every state sss let D(s)⊆A(s)\mathcal D(s)\subseteq\mathcal A(s)D(s)⊆A(s) be nonempty; put D=∏s∈SD(s)\mathcal D=\prod_{s\in\mathcal S}\mathcal D(s)D=∏s∈S​D(s). Define

W(s)=sup⁡{π: dtπ∈D} inf⁡P∈TπEP[∑t=0∞λtr(st,dt(ht),st+1)],W(s)=\sup_{\{\pi:\,d^\pi_t\in\mathcal D\}}\ \inf_{\mathbf P\in\mathcal T^\pi}\mathbf E^{\mathbf P}\Big[\sum_{t=0}^\infty\lambda^t r(s_t,d_t(h_t),s_{t+1})\Big],W(s)={π:dtπ​∈D}sup​ P∈Tπinf​EP[t=0∑∞​λtr(st​,dt​(ht​),st+1​)],

the supremum over deterministic Markov policies π=(d0,d1,… )\pi=(d_0,d_1,\dots)π=(d0​,d1​,…) with every dt∈Dd_t\in\mathcal Ddt​∈D. Then WWW is bounded, LDW=W\mathcal L_{\mathcal D}W=WLD​W=W, and WWW is the only bounded solution of LDV=V\mathcal L_{\mathcal D}V=VLD​V=V.

With D=∏sA(s)\mathcal D=\prod_s\mathcal A(s)D=∏s​A(s) this identifies the solution of the robust optimality equation with the robust value over deterministic Markov policies.

Formalization Note The page states Theorem 5(b) for an arbitrary set D\mathcal DD of decision rules. Its proof of the upper bound (28) picks one d∈Dd\in\mathcal Dd∈D that is ϵ\epsilonϵ-greedy at every state simultaneously, which needs D\mathcal DD closed under state-wise pasting; for a non-product D\mathcal DD the statement fails (two states, D={d1≡a, d2≡b}\mathcal D=\{d_1\equiv a,\ d_2\equiv b\}D={d1​≡a, d2​≡b}). It is stated here for product sets D=∏sD(s)\mathcal D=\prod_s\mathcal D(s)D=∏s​D(s), which include both cases the paper uses (D={d}\mathcal D=\{d\}D={d} and D=∏sA(s)\mathcal D=\prod_s\mathcal A(s)D=∏s​A(s)). "Unique solution" is uniqueness among bounded functions.

Preamble
import Mathlib
import Definitions.Def_RobustDP_Discounted_Model
import Definitions.Def_RobustDP_Discounted_Value
import Definitions.Def_RobustDP_Discounted_Bellman
Formal statement
namespace RobustDP.Discounted

theorem theorem5b_fixed_point {S A : Type*} [Countable S] [Countable A] (M : Model S A)
    (Dset : S → Set A) (hDne : ∀ s, (Dset s).Nonempty) (hDsub : ∀ s, Dset s ⊆ M.Aset s) :
    let W : S → ℝ := fun s => ⨆ f : ℕ → productRules M Dset, V M (markovPolicy M (fun n => (f n).1)) s
    IsBounded W ∧ L M (productRules M Dset) W = W ∧
      ∀ U : S → ℝ, IsBounded U → L M (productRules M Dset) U = U → U = W := by sorry

end RobustDP.Discounted
Source
Iyengar, Robust dynamic programming, CORC Tech Report TR-2002-07 (rev. May 4, 2004), p. 9, Theorem 5(b); proof pp. 10–11, (24)–(28)
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me