Growth rate conditions for uniform asymptotic stability of cascaded time-varying systems
Authors: Panteley, Loría · Year: 2001 · Venue: Automatica 37(3):453–460 (Brief Paper)
Raw: md
Summary
This paper gives sufficient conditions for uniform global asymptotic stability (UGAS) of a cascaded nonlinear time-varying (non-autonomous) system, where a UGAS “driving” subsystem Σ₂ perturbs a UGAS “driven” subsystem Σ₁ through an interconnection term g(t,x)x₂. The key idea is to classify cascades by the relative growth rate of the drift f₁(t,x₁) versus the interconnection g(t,x) as ‖x₁‖→∞, yielding three complementary theorems. The central technical payoff is that these growth-rate conditions are typically easier to check than constructing a strict (negative-definite-derivative) Lyapunov function for the full cascade, and they extend the autonomous folklore result “GAS + GAS + bounded solutions ⇒ GAS” to the non-autonomous setting.
Key Claims
- Lemma 2 (reduction principle): If Σ₂ (eq. 2) and the unperturbed driven system ẋ₁=f₁(t,x₁) (eq. 4) are each UGAS, and the solutions of the cascade (1)–(2) are globally uniformly bounded, then the cascade (1)–(2) is UGAS. This reduces every theorem to a boundedness argument.
- Theorem 3 (Case 1, f₁ dominates g): Under A1, A2 and the little-o condition as ‖x₁‖→∞ (A3), the cascade is UGAS; moreover if W(x₁) ≥ α₃(‖x₁‖) for some α₃∈𝒦, then Σ₁ is ISS w.r.t. input x₂.
- Theorem 4 (Case 2, f₁ majorizes g): Under A1, A2, the integral growth-rate condition A4 (eqs. 22–23), and A5 (‖[L_gV]‖ ≤ λW(x₁) for ‖x₁‖≥η on each ball ‖x₂‖<r), the cascade is UGAS — even when Σ₁ is not ISS w.r.t. x₂. A4’s divergent integral ∫ds/α₆(s)=∞ prevents finite-escape time.
- Theorem 5 (Case 3, g grows faster than f₁): Under A1, A2, A4 and the integrability condition A6, ∫_{t₀}^∞ ‖x₂(t)‖ dt ≤ φ(‖x₂(t₀)‖) with φ∈𝒦, the cascade is UGAS. Here mere decay of x₂ is insufficient; integrability of the input trajectory is what buys boundedness.
- Example 4 is a counterexample showing that in general GAS + GAS + forward-completeness does not imply GAS: ẋ₁=−sat(x₁)+x₁x₂, ẋ₂=−x₂³ is forward complete with A1,A2,A4 holding, yet trajectories can grow unbounded because x₂(t)=(2t+1/x₂₀²)^{−1/2} is not integrable (fails A6).
- These conditions are sufficient but not necessary (stated explicitly in the Conclusions).
Method
Model: the time-varying cascade
with x₁∈ℝⁿ, x₂∈ℝᵐ, x = col[x₁,x₂], all functions continuous and locally Lipschitz in x uniformly in t, f₁ continuously differentiable, and a nondecreasing G with ‖g(t,x)‖ ≤ G(‖x‖).
Standing assumptions: A1 — system (4) ẋ₁=f₁(t,x₁) is UGAS, admitting a 𝒞¹ Lyapunov function V(t,x₁) with α₁(‖x₁‖) ≤ V ≤ α₂(‖x₁‖), V̇₍₄₎ ≤ −W(x₁) (only semi-definite W is required), and ‖∂V/∂x₁‖ ≤ α₄(‖x₁‖). A2 — Σ₂ is UGAS, hence ‖x₂(t)‖ ≤ β(‖x₂₀‖, t−t₀) for some β∈𝒦ℒ.
The taxonomy uses two comparison notions: little-o (Def. 1: φ=o(ϱ) iff ‖φ‖ ≤ λ(‖x‖)‖ϱ‖ with λ→0) and majorization (Def. 2: limsup ‖φ‖/‖ϱ‖ < ∞). The three theorems then partition cascades by whether the Lie-derivative interconnection term L_gV is little-o of, comparable to, or dominant over the decay rate W. Proofs run through the Lyapunov derivative along (1),
then establish forward completeness via A4 (no finite escape) and global uniform boundedness, after which Lemma 2 delivers UGAS.
Regime note (free-flying vs free-floating): this is a pure nonlinear-systems / Lyapunov-theory paper — there is no spacecraft, manipulator, or base-actuation model, so the free-flying vs free-floating distinction does not arise here. Its relevance to a free-flying space manipulator is as a tool: the motivating application class cited is trajectory-tracking control of robot manipulators and Euler–Lagrange systems, whose tracking-error dynamics naturally take a non-autonomous cascade form.
Relevance to thesis
For a free-flying space manipulator, the closed-loop tracking-error dynamics under a nominal guidance/control law are typically non-autonomous (the reference trajectory enters explicitly through time) and frequently decompose into a cascade — e.g. an attitude/base-pose error loop driving a manipulator-configuration error loop, or an estimator error driving a controller. This paper supplies exactly the right machinery to certify UGAS of such a tracking design without having to build a single strict Lyapunov function for the coupled system: prove each loop UGAS separately, then verify a growth-rate condition (Theorem 3, 4, or 5) on the interconnection. The ISS corollary of Theorem 3 and the integrability condition A6 of Theorem 5 are directly usable when later adding bounded disturbances or vanishing perturbations in the risk-aware layer.
Connections
Topics: cascaded_systems · lyapunov_stability · trajectory_tracking · input_to_state_stability
Key Equations / Quotes
Cascade and interconnection bound (eqs. 1–3):
Reduction lemma:
“If systems (2) and (4) are UGAS and the solutions of (1) and (2) are globally uniformly bounded then (1) and (2) is UGAS.” (Lemma 2, p. 455)
Case-1 little-o growth condition (A3, eq. 21):
Case-2 non-escape integral condition (A4, eqs. 22–23):
Case-3 integrability of the driving signal (A6, eq. 25):
On the motivation:
“The advantage of this type of analysis is that it is often easier to verify such conditions than to find a Lyapunov function with a negative-definite derivative.” (Abstract, p. 453)
Open Questions
- The authors state the three growth-rate classes are sufficient but not necessary for UGAS of the cascade; finding necessary conditions is flagged as “an important and challenging subject of future research” (Conclusions, p. 459).
- Conditions are given for the factored interconnection g(t,x)x₂; Corollary 1 extends to a non-factored g(t,x) under additional structure (eq. 32), but a fully general characterization is not provided.
- How tight is the integrability requirement A6 relative to merely-decaying (non-integrable) inputs in practical tracking designs? Example 4 shows decay alone fails, but the gap between A6 and necessity is not quantified.