Theorem 5.1, pp. 18–19 — for Φ ∈ 𝚷_{n,T}, data are informative for quadratic stabilization iff (5.5) is feasible; under Φ₂₂ < 0 and full row rank, iff (5.6) is; K = LP⁻¹ stabilizes all of Σ
ProvedQuadMatIneq.Stabilization.theorem_5_1Let and consider data generated by with noise satisfying (2.3). Write
- (a) The data are informative for quadratic stabilization if and only if there exist , , and with
- (b) If in addition and , the data are informative for quadratic stabilization if and only if there exist and with
- Moreover, whenever satisfies (5.5), or satisfies (5.6), the gain stabilizes every : and is Schur stable.
This is the main data-driven control result of the paper: a necessary and sufficient LMI test for the existence of a single state-feedback controller, with a common quadratic Lyapunov function, that stabilizes every system consistent with noisy data, under any noise model in the class and with no rank or Slater condition in part (a).
Formalization Note "The data are generated by (2.1) with noise model (2.3)" is the hypothesis that some lies in ; it is essential (if , informativity holds vacuously). and are lmi55 and lmi56; informativity is InformativeQS (Definition 2.1); "stabilizing feedback gain for all " is StabilizesAll, stated with the of the feasible LMI: the Lyapunov inequality (2.4) together with . is Mathlib's matrix inverse, the true inverse since . The Moreover part is stated under the theorem's standing hypotheses only, without those of (b), as in the paper. No hypothesis is added: at both sides of (a) and (b) hold.
import Mathlib import Definitions.Def_QuadMatIneq_Stabilization_QMI import Definitions.Def_QuadMatIneq_Stabilization_Data open Matrix Filter Topology
namespace QuadMatIneq.Stabilization
/-- Theorem 5.1, pp. 18–19. `hgen` is "the data `(U₋, X)` are generated by (2.1) with noise
model (2.3)": the true system `(A_s, B_s)` lies in `Σ`. -/
theorem theorem_5_1 {n m T : ℕ}
(Φ : Matrix (Fin n ⊕ Fin T) (Fin n ⊕ Fin T) ℝ) (X : Matrix (Fin n) (Fin (T + 1)) ℝ)
(Um : Matrix (Fin m) (Fin T) ℝ) (hΦ : InPi Φ)
(hgen : ∃ As : Matrix (Fin n) (Fin n) ℝ, ∃ Bs : Matrix (Fin n) (Fin m) ℝ,
(As, Bs) ∈ Sigma Φ X Um) :
-- (a)
(InformativeQS Φ X Um ↔
∃ P : Matrix (Fin n) (Fin n) ℝ, ∃ L : Matrix (Fin m) (Fin n) ℝ, ∃ β : ℝ,
P.PosDef ∧ 0 < β ∧ (lmi55 Φ X Um P L β).PosSemidef) ∧
-- (b)
((-Φ.toBlocks₂₂).PosDef → (Matrix.fromRows (Xminus X) Um).rank = n + m →
(InformativeQS Φ X Um ↔
∃ P : Matrix (Fin n) (Fin n) ℝ, ∃ L : Matrix (Fin m) (Fin n) ℝ,
P.PosDef ∧ (lmi56 Φ X Um P L).PosDef)) ∧
-- Moreover, for (5.5)
(∀ (P : Matrix (Fin n) (Fin n) ℝ) (L : Matrix (Fin m) (Fin n) ℝ) (β : ℝ),
P.PosDef → 0 < β → (lmi55 Φ X Um P L β).PosSemidef → StabilizesAll Φ X Um P (L * P⁻¹)) ∧
-- Moreover, for (5.6)
(∀ (P : Matrix (Fin n) (Fin n) ℝ) (L : Matrix (Fin m) (Fin n) ℝ),
P.PosDef → (lmi56 Φ X Um P L).PosDef → StabilizesAll Φ X Um P (L * P⁻¹)) := by sorry
end QuadMatIneq.Stabilization
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.