Dynamic Stabilisation of Boundary Control Systems

Technical Summary: Dynamic Stabilisation of Boundary Control Systems

Problem Statement
This paper addresses the dynamic stabilisation of abstract linear boundary control systems defined on Hilbert spaces. The primary focus is on systems governed by partial differential equations (PDEs) with boundary inputs, which are often modeled in the form:
$$\dot{x}(t) = Ax(t) + Biu(t), \quad Bx(t) = Qu(t), \quad y(t) = Cx(t)$$
where $x(t)$ is the state, $u(t)$ is the input, and $y(t)$ is the output. A significant challenge in this domain is that hyperbolic PDEs with boundary control and in-domain observation (or vice versa) frequently resist exponential stabilisation. While classical observer-based control relies on the separation principle, the authors note that in many such systems, one component of the closed-loop dynamics (typically the observer error dynamics) may fail to be exponentially stable, even if the state feedback component is. Consequently, the closed-loop system may only achieve non-uniform or polynomial stability, requiring a rigorous framework to establish well-posedness and decay rates in these weaker stability regimes.

Methodology
The authors employ the framework of boundary nodes to represent the system, which offers a direct connection to the underlying PDE models and simplifies the construction of boundary trace operators compared to well-posed linear system formulations. The core methodology involves designing an observer-based dynamic feedback controller of the form:
$$\dot{\hat{x}}(t) = A\hat{x}(t) + Biu(t) + Li(\hat{y}(t) - y(t))$$
$$B\hat{x}(t) = Qu(t) + L(\hat{y}(t) - y(t))$$
$$u(t) = K\hat{x}(t)$$
where $K$ is a state feedback gain, and $L, Li$ are output injection gains.

Key methodological features include:

1. Internal Loop Structure: The controller is treated as having an "internal loop" (in the sense of [44, 7]), allowing for well-defined dynamics even when disconnected from the plant. This is crucial for systems where the observer dynamics alone might not generate a strongly continuous semigroup.
2. Separation Principle Analysis: The stability of the closed-loop system is analyzed by decomposing it into the state feedback semigroup ($T_K$) and the observer error semigroup ($T_L$). The authors utilize results on cascaded systems to show that if one semigroup is exponentially stable and the other is polynomially (or non-uniformly) stable, the combined system inherits the slower decay rate.
3. Well-Posedness Verification: The paper establishes conditions under which the closed-loop system remains externally well-posed, ensuring that input-output maps are bounded and solutions exist in appropriate function spaces. This involves verifying that the transfer functions of the feedback and injection loops satisfy specific $H^\infty$ bounds.

Key Contributions

• Generalised Stability Conditions: The paper introduces conditions for exponential, strong, and polynomial stability for the closed-loop system. It specifically addresses the scenario where the separation principle yields a mix of exponential and non-uniform stability, proving that the closed-loop system achieves the slower decay rate (e.g., polynomial stability with a specific exponent).
• Unified Framework for Boundary Nodes: The authors extend and unify previous studies on observer-based stabilisation (previously limited to regular linear systems, well-posed systems, or specific PDEs) to the general class of abstract boundary control systems.
• Internal Loop Formulation: The work formally demonstrates that the proposed observer-based controller possesses an internal loop structure, guaranteeing well-defined dynamics without requiring the disconnected observer to be a well-posed system on its own.
• Application to Non-Exponential Cases: The methodology is applied to systems where exponential stabilisation is theoretically impossible due to geometric constraints or the nature of the control action, demonstrating that polynomial stabilisation is achievable.

Results
The paper presents three main theoretical results and three concrete applications:

1. Theoretical Results:

   • Theorem 3.1: Establishes that if the open-loop system is a well-posed boundary node and the feedback/injection gains satisfy specific $H^\infty$ conditions, the closed-loop system is a well-posed boundary node. It proves that the stability of the closed-loop semigroup is determined by the stability of the individual feedback ($T_K$) and observer ($T_L$) semigroups. Specifically, if one is exponentially stable and the other is polynomially stable (with exponent $\alpha$), the closed-loop system is polynomially stable with exponent $\alpha$.
   • Proposition 3.2: Provides sufficient conditions for polynomial stabilisation in systems with collocated inputs and outputs, showing that the closed-loop exponent is the minimum of the exponents of the feedback and observer components.
   • Corollary 3.4: Extends these results to cases where external well-posedness is not required, focusing solely on the generation of a stable semigroup for the closed-loop state.

2. Applications to PDEs:

   • Two-Dimensional Wave Equation: A wave equation on a square domain with in-domain control and observation on disjoint subdomains. Since the control region does not satisfy the Geometric Control Condition, exponential stabilisation is impossible. The authors design a controller achieving polynomial stability with exponent $\alpha = 1/2$.
   • One-Dimensional Wave Equation: An undamped wave equation with non-collocated boundary input and output. The system cannot be stabilised by static output feedback. The authors design an observer-based controller that achieves exponential stabilisation and proves the external well-posedness of the resulting closed-loop system.
   • Non-Uniform SCOLE Model: A model consisting of an undamped Euler–Bernoulli beam coupled with tip mass dynamics. Since the control acts only through the finite-dimensional part of the model, exponential stabilisation is not possible. The proposed controller achieves polynomial stabilisation with exponent $\alpha = 1/2$.

Significance
The paper claims significance in providing a robust, abstract framework for stabilising infinite-dimensional systems where traditional exponential stabilisation fails. By leveraging the boundary node formalism, the authors offer a method that is "relatively easy to construct" compared to regular or well-posed linear system settings and allows for immediate interpretation as a PDE system. The work is particularly notable for extending observer-based stabilisation to cases of polynomial and non-uniform stability, a regime that has been less explored in the context of general boundary control systems. The authors emphasize that their approach unifies earlier studies and successfully handles systems that are not well-posed in the classical sense, thereby broadening the scope of dynamic stabilisation for hyperbolic PDEs.