---
title: Output-Feedback Contraction in Control Theory
url: https://www.emergentmind.com/topics/output-feedback-contraction
type: topic
---

# Output-Feedback Contraction in Control Theory

Searching arXiv for recent and foundational papers on output-feedback contraction and related contraction-based feedback design.
Search query: "output-feedback contraction control contraction metrics observer contraction metric arXiv"
Output-feedback contraction denotes a family of control-theoretic constructions in which an output-feedback law is designed or analyzed so that the resulting closed-loop system is contracting, incrementally exponentially stable, or, in generalized settings, \(2\)-contracting. In finite-dimensional nonlinear systems, the term is most closely associated with control contraction metrics (CCMs), dual observer contraction metrics, and separation principles for dynamic output-feedback controllers [1406.1256]. In discrete time, the same theme appears in contraction-based LMIs for state-feedback and observer synthesis on learned Structured State-space Models (SSMs) with bi-Lipschitz input and output scaffoldings [2604.07069]. In infinite dimensions, contraction is formulated at the level of nonlinear contraction semigroups or contraction semigroups generated by closed-loop operators, and output-based integral action or static output feedback is used to obtain regulation and stabilization [2201.10146][2112.08105]. A broader usage also includes contraction certification for feedback interconnections, two-time-scale output-feedback loops, and learned-perception control architectures in which state estimates are produced from high-dimensional observations [2408.12790][2310.07966][2206.06553].

## 1. Differential and semigroup notions of contraction

In the continuous-time finite-dimensional setting, the basic plant is
\[
\dot x(t) = f(x(t),t) +B(t)u(t), \qquad y(t) = C(t)x(t),
\]
with differential dynamics
\[
\dot\delta_x(t) = A(x,t)\delta_x(t)+B(t)\delta_u(t), \qquad \delta_y(t) = C(t)\delta_x(t),
\]
where \(A(x,t)=\partial f(x,t)/\partial x\). A control contraction metric is a Riemannian metric of the form
\[
V(x,\delta_x,t)=\delta_x' M(x,t)\delta_x,\qquad \alpha_1I\le M(x,t)\le \alpha_2I,
\]
such that uncontrolled differential directions are strictly contracting [1406.1256]. In discrete time, contraction is expressed through
\[
(A_k + B_k K_k)^T M_{k+1} (A_k + B_k K_k) - (1-\rho) M_k \prec 0,
\]
for some \(0<\rho<1\), with \(M_k\succ 0\); this is the discrete-time analogue of standard contraction inequalities and implies that squared differential distances measured in metric \(M\) shrink by at least factor \(1-\rho\) per step [2604.07069].

The semigroup formulation is the infinite-dimensional analogue. For a nonlinear evolution
\[
\dot{w}+A(w)=0
\]
on a Hilbert space \(H\), strong monotonicity,
\[
(A(w_1)-A(w_2),\, w_1-w_2)_H \ge \alpha \|w_1-w_2\|_H^2,
\]
implies the exponential contraction estimate
\[
\|T_t w_1 - T_t w_2\|_H \le e^{-\alpha t}\,\|w_1-w_2\|_H,
\]
where \(\{T_t\}_{t\ge0}\) is the nonlinear strongly continuous contraction semigroup generated by \(-A\) [2201.10146]. In the linear operator-theoretic setting, a closed-loop operator generates a contraction semigroup precisely when it is maximal dissipative, by the Lumer–Phillips theorem; this is the fundamental mechanism behind several PDE feedback constructions [1403.3564].

These formulations share the same structural idea: contraction is not merely asymptotic stability of one trajectory, but a uniform decay of distances between trajectories in a chosen metric or norm. This suggests why output-feedback contraction is naturally tied to observer design, disturbance rejection, and trajectory tracking rather than only equilibrium stabilization.

## 2. Controller–observer synthesis in finite dimensions

The standard finite-dimensional realization is the CCM/observer-metric framework for nonlinear control-affine systems. A convex sufficient condition for universal exponential stabilizability by state feedback is the existence of \(W_c(x,t)\in S_+^n\), \(\rho(x,t)\ge 0\), and constants \(\alpha_2\ge\alpha_1>0\) such that
\[
\alpha_1I\le W_c(x,t)\le \alpha_2 I,
\]
and
\[
-\dot W_c + W_cA' + AW_c -\rho BB' \le -2\lambda W_c.
\]
Then the control contraction metric is \(M_c=W_c^{-1}\), and the differential gain is
\[
K(x,t) = -\frac{1}{2} \rho(x,t)\,W_c(x,t)^{-1}B',
\]
used together with a geodesic integration formula
\[
u(t) = u^\star(t)+\int_\gamma K(\gamma(s))\frac{\partial \gamma}{\partial s}ds
\]
to track a reference trajectory \((x^\star,u^\star)\) [1406.1256].

The dual observer problem replaces \(BB'\) by \(C'C\). An observer contraction metric \(W_o(x,t)\in S_+^n\) satisfies
\[
\dot W_o +A'W_o+W_oA-\rho C'C\le -2\lambda W_o,
\]
with the same type of metric bounds. In the constant-metric case this yields a Luenberger-type observer, while in the general case the observer uses a minimizing path from the estimate \(\hat x(t)\) to the output-consistent set
\[
\mathfrak X_y(t):=\{x \in \mathbb{R}^n : C(t)x=y(t)\}
\]
in the metric \(W_o\) [1406.1256]. The resulting theorem is a nonlinear separation principle: if a state-feedback CCM and a dual observer contraction metric both exist, then the dynamic output-feedback controller obtained by evaluating the CCM-based feedback at the state estimate is exponentially stable [1406.1256].

A discrete-time variant appears in SSM-based indirect data-driven control. The nonlinear SSM model
\[
x_{k+1} = A x_k + B \mathcal{S}_u(u_k),\qquad
y_k = \mathcal{S}_y\big(C x_k + D \mathcal{S}_u(u_k)\big)
\]
uses a linear recurrent unit together with bi-Lipschitz neural-network scaffoldings \(\mathcal{S}_u,\mathcal{S}_y\). Local controllability is inherited from controllability of \((A,B)\) and bi-Lipschitzness of \(\mathcal{S}_u\); local observability is inherited from observability of \((A,C)\) and bi-Lipschitzness of \(\mathcal{S}_y\) [2604.07069]. State-feedback synthesis is performed through the LMI
\[
\begin{bmatrix}
(1-\rho_c)P - \sigma B B^T & A Y + \alpha_u B X & 0\\
\star & Y^T+Y-P & \beta_u X^T\\
\star & \star & \sigma I
\end{bmatrix}\succ 0,
\]
with \(K=XY^{-1}\), while observer synthesis is performed through
\[
\begin{bmatrix}
(1-\rho_o) Q - \eta C^T C & (U A + \alpha_y V C)^T & 0\\
\star & U + U^T - Q & \beta_y V\\
\star & \star & \eta I
\end{bmatrix} \succ 0,
\]
with \(L=U^{-1}V\) [2604.07069]. The paper then proves a discrete-time separation principle for the full output-feedback loop. A common misconception is that contraction-based output feedback is restricted to continuous time; these SSM results show that the same architecture extends to learned discrete-time surrogates.

## 3. Interconnections, small-gain conditions, and generalized \(2\)-contraction

Output-feedback contraction is not restricted to \(1\)-contraction of a single plant–observer pair. For the feedback interconnection
\[
\dot x = f(x,z),\qquad \dot z = g(x,z),
\]
with Jacobian
\[
J(x,z)= \begin{bmatrix} \frac{\partial f}{\partial x} & \frac{\partial f}{\partial z}\\[0.3em]
\frac{\partial g}{\partial x} & \frac{\partial g}{\partial z} \end{bmatrix}
=\begin{bmatrix}A&B\\C&D\end{bmatrix},
\]
a sufficient condition for \(2\)-contraction is formulated in terms of a \(3\times 3\) Metzler matrix \(S(x,z)\) built from the \(2\)-contraction rates of the subsystems, the \(1\)-contraction of the subsystems, and the cross-gains \(\partial f/\partial z,\partial g/\partial x\) [2408.12790]. If there exists a matrix measure \(\mu_0\) induced by a monotonic norm \(|\cdot|_0\) such that
\[
\mu_0(S(x,z))\le -\eta<0\quad\forall(x,z),
\]
then the feedback system is \(2\)-contracting with rate \(\eta\) [2408.12790]. The associated small-gain reduction yields the scalar inequality
\[
\ell_z(f)\,\ell_x(g) < \frac{1}{2}\, \frac{(c_x(f)+c_z(g))\, c_x^{[2]}(f)\, c_z^{[2]}(g)} {c_x^{[2]}(f)+c_z^{[2]}(g)}.
\]

This generalized setting is significant because \(2\)-contraction does not imply a unique equilibrium. For time-invariant \(2\)-contracting systems with bounded trajectories, every bounded trajectory converges to the set of equilibria, yet multiple equilibria and multiple locally stable equilibria are allowed [2408.12790]. Accordingly, output-feedback contraction need not be synonymous with monostability. The paper’s FitzHugh–Nagumo network example makes this explicit: the network can be \(2\)-contractive but not \(1\)-contracting, and trajectories still converge to equilibria [2408.12790].

A related small-gain viewpoint appears in two-time-scale online feedback optimization. For the closed loop
\[
\begin{cases}
\dot u = -\nabla\phi(u) - G^\top \nabla\psi(z),\\
\epsilon \dot z = A z + B u + E w_z,
\end{cases}
\]
the plant is the fast contractive subsystem and the reduced gradient-flow controller is the slow contractive subsystem. The full shifted system is analyzed through a \(2\times 2\) Metzler gain matrix
\[
\Gamma_{\epsilon,\text{ofo}}=
\begin{bmatrix}
-\nu & \|G\|\ell_\psi\\
\|G\|\ell & \epsilon^{-1}\mu(A) + \|A^{-1}B G^\top\|\ell_\psi
\end{bmatrix},
\]
and if this matrix is Hurwitz the full output-feedback loop is strongly infinitesimally contractive in a composite norm [2310.07966]. This suggests that output-feedback contraction can be organized hierarchically: contraction margins of subsystems, interconnection gains, and time-scale separation jointly determine closed-loop incremental stability.

## 4. Infinite-dimensional formulations and contraction semigroups

In nonlinear infinite-dimensional systems, output-feedback contraction appears in a forwarding framework for robust output regulation. The plant is
\[
\dot{w} + A(w) = Bu(t),\qquad y=Cw,
\]
with \(-A\) generating a nonlinear strongly continuous contraction semigroup on \(H\). After introducing an integral state \(z\in Z\),
\[
\dot{w} + A(w) = Bu(t) + d,\qquad \dot{z} = Cw - y_{\mathrm{ref}},
\]
the design seeks a Lyapunov functional
\[
V(w,z) = \frac12 \|w\|_H^2 + \frac{\rho}{2}\,\|z - M(w)\|_Z^2
\]
based on a forwarding map \(M\in\mathcal C^1(H,Z)\) satisfying
\[
M(0)=0,\qquad dM(w)\,A(w) + Cw = 0
\]
[2201.10146]. The feedback law
\[
u(t) = B^* dM(w(t))^* [z(t) - M(w(t))]
\]
then yields a transformed closed loop in coordinates \(\eta=z-M(w)\) whose operator is locally strongly monotone, hence locally exponentially contractive in a weighted norm on \(H\times Z\) [2201.10146]. Under additional coercivity conditions, global asymptotic stability is obtained, and the equilibrium satisfies \(Cw^\star = y_{\mathrm{ref}}\) [2201.10146]. The paper explicitly notes that it does not construct an observer, so this is output-oriented integral action rather than pure output-only feedback.

A linear operator-theoretic counterpart is static output feedback for system nodes \(\Sigma\) with generating triple \((A,B,C)\) and transfer function \(G\). If there exists \(E\in{\mathcal L}(U)\) such that \(\Sigma_E\), obtained by replacing \(G(s)\) by \(G(s)+E\), is impedance passive, and if
\[
u(t) = -\kappa\,y(t) + v(t),
\]
with \(0<\kappa<\kappa_0\), then the closed-loop generator \(A^\kappa\) generates a contraction semigroup on \(X\) [2112.08105]. If, moreover, the plant is approximately observable or approximately controllable in infinite time, then the closed-loop semigroup is weakly stable; if \(\sigma(A)\cap i\mathbb R\) is countable, then the closed-loop semigroup and its dual are both strongly stable [2112.08105].

The feedback-theoretic semigroup construction makes the same point in a broader PDE context. A larger operator \(A_{\mathrm{ext}}\) is represented by a passive system node via the external Cayley system transform, and an internal maximal accretive operator \(S\) is mapped to a static output feedback operator
\[
K=(S-I)(S+I)^{-1}.
\]
If \(K\) is admissible, then the internal-loop operator \(A_S\) is the closed-loop main operator and generates a contraction semigroup on \(X_1\) [1403.3564]. This framework is used to show well-posedness of the heat equation, wave equations with viscous or structural damping, and degenerate parabolic equations [1403.3564]. In this literature, “output-feedback contraction” is therefore literally generation of a contraction semigroup by closing a passive system with a static output feedback law.

## 5. Data-driven, learned, and perception-mediated realizations

A data-driven realization is given by unknown nonlinear systems of the form
\[
\dot x = A Z(x) + B u,
\]
where \(f(x)=AZ(x)\) is expressible through a known dictionary \(Z(x)=[x\;Q(x)]\). With data matrices \(U_0,X_1,Z_0\), the controller
\[
u = K Z(x)
\]
is synthesized by semidefinite programs whose feasibility implies a constant-metric contraction inequality for the closed-loop Jacobian [2401.07819]. In the noise-free case, the SDP uses the linear constraints
\[
Z_0 Y_1 = \begin{bmatrix} P_1 & 0\end{bmatrix},\qquad
Z_0 G_2 = \begin{bmatrix} 0 & I\end{bmatrix},
\]
together with an LMI involving \(X_1Y_1\), \(X_1G_2\), and a Jacobian bound on \(Q\) [2401.07819]. For disturbances that are linear combinations of sinusoids of known frequencies and constants, an added linear constraint
\[
M\begin{bmatrix} Y_1 & G_2 \end{bmatrix} = 0
\]
yields contraction conditions that do not depend on the magnitude of the disturbances [2401.07819]. The same machinery is extended to an integral controller
\[
\dot \xi = e,\qquad e=y-r,\qquad u=\mathcal K \mathcal Z(x,\xi),
\]
which makes the plant-plus-integrator extended state contracting and thereby achieves constant reference tracking and constant disturbance rejection [2401.07819]. Although this framework assumes the full state is measured, it is directly relevant to output-feedback contraction because the integral state is driven by output error and the extended closed loop is the object made contractive.

A complementary learned-perception architecture is developed for uncertain control-affine systems with high-dimensional outputs,
\[
\dot x(t) = f(x(t)) + Bu(t) + B_w(t)w_x(t),\qquad
y(t) = h(x(t), \theta) + B_yw_y(t).
\]
A learned inverse perception module \(\hat h(y,\theta)\) approximates a reduced observable state \(y_r=C_r x\), and contraction theory is used both to design a stabilizing state feedback controller and a convergent dynamic state observer [2206.06553]. The observer uses the learned perception system through
\[
\dot{\hat x} = f(\hat x) + B u(\hat x,x^\star,u^\star) + \frac{1}{2}\rho(\hat x) M_e(\hat x) C_r^\top \big( \hat h(h(x,\theta)+B_y w_y,\theta) - C_r \hat x \big),
\]
and the coupled tracking and estimation errors satisfy a Metzler differential inequality in the Riemannian distances \(d_c\) and \(d_e\) [2206.06553]. These bounds are incorporated into the CORRT planner so that the tracking tube \(\Omega_c(t)\) remains in the safe set and in the trusted domain where both the CCM/OCM conditions and the perception error bounds are valid [2206.06553]. This is a direct realization of output-feedback contraction from images: learned output inversion, contraction-based observer and controller, and set-based planning are combined in a single safety-critical loop.

## 6. Scope, limitations, and related directions

The literature uses the phrase “output-feedback contraction” in several nonequivalent senses. In the CCM tradition, it means a dynamic controller obtained by combining a state-feedback contraction design with a contracting observer, typically under a nonlinear separation principle [1406.1256][2604.07069]. In semigroup and PDE settings, it often means that static output feedback produces a maximal dissipative closed-loop generator and therefore a contraction semigroup [2112.08105][1403.3564]. In regulation problems with integral action, the feedback is output-oriented because the integrator is driven by \(y-y_{\mathrm{ref}}\), even when the plant state is still assumed available [2201.10146][2401.07819]. A common misconception is that the term always denotes pure static laws of the form \(u=k(y)\); the cited work shows that it may also denote observer-based dynamic feedback, integral action on an extended state, or static output feedback in operator-theoretic form.

A second misconception is that contraction always implies a unique equilibrium. Classical \(1\)-contraction does imply that all trajectories converge exponentially to one another and, in the time-invariant case with a forward invariant convex set, a unique globally exponentially stable equilibrium [2408.12790]. By contrast, \(2\)-contraction permits multistationarity: every bounded trajectory converges to the set of equilibria, yet multiple locally stable equilibria may coexist [2408.12790]. This suggests that generalized output-feedback contraction can be used to certify global asymptotic structure without excluding multistability.

The boundary of the term is illustrated by open-loop contraction design. For systems in normal form with scalar output \(y=x_1\), the existence of an output variable that ensures contraction of the inverse system facilitates the design of a contracting input perturbation [2209.04440]. The paper explicitly studies open-loop perturbations \(u^{**}(t)=u^*(t)+\Delta u(t)\) rather than feedback laws, but its structural message is closely related: a contracting inverse system makes output-based trajectory design tractable [2209.04440]. This suggests a plausible implication that output-feedback contraction and output-programming for contraction are adjacent rather than identical notions.

Across these strands, the recurring invariants are metric selection, incremental dissipation, small-gain compatibility of interconnections, and certification of the closed-loop geometry of trajectories. Output-feedback contraction is therefore best understood not as one theorem or one algorithm, but as a unifying viewpoint in which outputs, observers, integral states, or passive interconnections are organized so that the closed loop inherits contraction or a generalized contraction property from its constituent parts.

Source: https://www.emergentmind.com/topics/output-feedback-contraction