Viscous Burgers Equations with Neumann Feedback

• The paper introduces explicit nonlinear Neumann boundary feedback controllers that achieve exponential stabilization and robust output regulation for viscous Burgers equations.
• It leverages advanced tools such as Lyapunov functionals, energy identities, and variational methods to link continuous PDE dynamics with discrete numerical schemes.
• The approach effectively addresses tracking, disturbance rejection, and stabilization across various settings, including higher dimensions and memory effects, ensuring global existence and convergence.

The viscous Burgers equation with nonlinear Neumann boundary feedback control is a central paradigm for studying the boundary stabilization and robust output regulation of nonlinear parabolic PDEs. This framework enables exponential stabilization, robust tracking, and disturbance rejection—often through explicit boundary controller formulas—of both stable and unstable equations, including cases with reaction-type instabilities or external disturbances. The analytical tools developed for this class of systems exploit Lyapunov functionals, energy identities, and modern variational, finite element, and finite difference discretizations, establishing strong links between continuous and discrete stabilization properties.

1. Problem Setting and Model Structure

The canonical model is the one-dimensional viscous Burgers equation with control and disturbances on the spatial domain $x\in[0,1]$, $t\geq0$: $u_t = \nu u_{xx} - u u_x + a(x,t) u + u_d(x,t),$ with Neumann boundary conditions

$$\nu u_x(0,t)=f_0(t),\quad \nu u_x(1,t)=f_1(t),\quad u(x,0)=u_0(x),$$

where $\nu > 0$ is the viscosity, $a(x,t)\geq 0$ an instability (reaction-type) term, $u_d(x,t)$ a distributed disturbance, and $f_0,f_1$ are the boundary actuators to be designed. For stabilization about a steady state $w_d$, one sets $w=u-w_d$, yielding

$$w_t - \nu w_{xx} + w_d w_x + w w_x = \text{(forcing)}.$$

This structure generalizes to forced, multi-dimensional, and BBM–Burgers type PDEs, including memory effects and more general nonlinearity in the convective term. Nonlinear Neumann boundary conditions of cubic (or generally polynomial) nonlinear type are the key mechanism for output regulation and stabilization (Liu, 2021, Kundu et al., 2018, Kundu et al., 2019, Singh et al., 2023).

2. Nonlinear Neumann Boundary Feedback Control Design

Variable Decomposition and Problem Splitting

To address tracking and disturbance rejection for unstable systems, the solution is decomposed: $$u(x,t) = \hat{u}(x,t) + U(x,t), \quad f_i(t) = \hat{f}_i(t) + F_i(t)$$ where $(\hat{u},\hat{f}_i)$ solve a pure-diffusion stabilization problem and $(U,F_i)$ address a dynamical regulator PDE that absorbs nonlinearity, instability, and disturbance effects. This separation allows for a straightforward stabilization of the linear part and regulation of the output via explicit feedforward control at the boundary. Both cubic nonlinear feedback (e.g., $\hat{f}_0(t) = k[\hat{u}(0,t)+\hat{u}^3(0,t)]$) and linear feedback ($\hat{f}_0(t) = k\hat{u}(0,t)$) can be employed, with cubic feedback often providing sharper robustness (Liu, 2021).

General Nonlinear Feedback Laws

The standard feedback for stabilization about a constant or nonconstant steady state is of the form: $$w_x(0,t) = \frac{1}{\nu}\Bigl[(c_0+w_d)w(0,t) + \frac{2}{9c_0}w^3(0,t)\Bigr],\quad w_x(1,t) = -\frac{1}{\nu}\Bigl[(c_1+w_d)w(1,t) + \frac{2}{9c_1}w^3(1,t)\Bigr]$$ with constants $c_0,c_1>0$ chosen by design (Kundu et al., 2018, Singh et al., 29 Nov 2025, Singh et al., 1 Feb 2026). Variants exist for multi-dimensional domains,

$$\partial_n w = -\frac{1}{\nu}\left[(c_2+w_d)w + \frac{2}{9c_2}w^3\right] \quad \text{on }\partial\Omega$$

and for nonconstant or forced steady states, cubic terms remain key (Kundu et al., 2018, Kundu et al., 2019).

Feedforward terms enforce integral or output constraints in the presence of tracking or disturbance rejection objectives. For instance, the regulator feedforward at the boundary is selected to enforce control on the mean value or output functional of the system.

3. Lyapunov Analysis, Stability, and Convergence

Lyapunov Functionals and Exponential Stabilization

Lyapunov (energy-like) functionals tailored to the equation and boundary terms underlie all stabilization proofs. For the 1D system,

$E(t) = \frac{1}{2} \int_0^1 w(x,t)^2 dx,$

its dissipation under the nonlinear feedback law is given by a differential inequality: $$\dot{E}(t) \leq -\lambda E(t),\quad \lambda = \min\{\nu, k-1/6\}$$ for appropriately chosen feedback $k$ (Liu, 2021). For more general equations and higher dimensions, the Lyapunov functional is augmented with boundary energy and, if necessary, memory terms. The structure is preserved for BBM–Burgers equations, multidimensional domains, or equations with memory kernels, and analyzed using integrated energy identities with suitable trace inequalities (Kundu et al., 2018, Kundu et al., 2018, Kundu et al., 2019, Singh et al., 1 Feb 2026).

Output Tracking, Robust Regulation, and Disturbance Rejection

Feedforward/feedback boundary control, constructed from the reference signal $r(t)$, disturbance $u_d$, and the stabilization solution, achieves exponential tracking: $$|e(t)| = \left|\int_0^1 u(x,t)dx - r(t)\right| \leq \|u_0 - r(0)\| e^{-\lambda t/2}$$ for continuous disturbances and references, as rigorously proved in (Liu, 2021).

Achieved Regularity and Robustness

The methodology yields global existence and uniqueness in $H^2$ (and higher) Sobolev spaces for the closed-loop system, both continuously and under discretization. Exponential decay holds in all relevant norms ($L^2$, $H^1$, $H^2$), robust to model parameters, boundary discretization, and mesh refinement (Kundu et al., 2018, Singh et al., 2023, Kundu et al., 2018, Singh et al., 1 Feb 2026).

4. Numerical Methods and Discretization Effects

Finite Element and Finite Difference Schemes

Exponential stabilization and optimal convergence rates carry over to both finite element (FEM) and finite difference (FDM) discretizations:

• $C^0$-conforming FEM achieves $O(h^2)$ convergence in $L^2$, $O(h)$ in $H^1$, and for boundary control laws, often $O(h^2)$ or superconvergent $O(h^{3/2})$ rates, depending on the domain and regularity (Kundu et al., 2018, Kundu et al., 2018, Kundu et al., 2019, Singh et al., 1 Feb 2026).
• Fully discrete schemes using $\theta$-methods unify implicit and Crank–Nicolson time integration. For $\theta\geq1/2$, exponential stability is unconditional. Optimal error bounds $O(h^2 + k)$ in $L^2$ and $O(h + k)$ in $H^1$ and for control inputs are achieved (Singh et al., 29 Nov 2025, Singh et al., 1 Feb 2026).

Control Law Implementation in Discretization

Cubic boundary laws are incorporated via one-sided finite difference stencils or appropriate weak forms in FEM. Discrete energy estimates closely mimic the continuous proofs, and exponential decay for the fully discrete state and control is rigorously established (Singh et al., 29 Nov 2025, Singh et al., 1 Feb 2026).

Performance and Limitations

Numerical experiments consistently show rapid convergence of the mean (or output) to the reference, robust exponential decay of the state and control, and sharp correspondence of observed and predicted convergence rates. The overall convergence order is typically limited by the lower-order accuracy at the boundary where the control acts, despite higher-order accuracy in the interior (Kundu et al., 2018, Singh et al., 29 Nov 2025, Singh et al., 1 Feb 2026). High-order boundary discretizations consistent with cubic feedback remain an open engineering challenge.

5. Theoretical Extensions and Generalizations

Memory Terms and Adaptive/Unknown Viscosity

For Burgers equations with memory (Volterra type convolution terms), stabilization is achieved using feedbacks including auxiliary boundary terms dependent on the memory variable. Exponential decay in all relevant norms is proved for both known and adaptively estimated viscosity; parameter update laws ensure convergence even when $\nu$ is not known a priori (Singh et al., 1 Feb 2026).

Higher Dimensions and Nonconstant Steady States

In 2D (and 3D) settings, stabilization around constant and certain classes of nonconstant steady states is established by modifying the boundary feedback to include dependence on the local value (or trace) of the steady state. Under smallness conditions on the steady state, cubic feedback guarantees exponential decay in $H^2$ and $L^2$ norms (Kundu et al., 2018, Kundu et al., 2019). The method is robust to spatial discretization, with proven optimal rates for both state and boundary control.

Tracking Without Exosystem Assumptions

The methodology removes previous reliance on exosystem-generated references and disturbances with pure imaginary spectra: arbitrary continuous references and distributed disturbances are handled, provided they meet mild regularity conditions. This broadens the scope of robust output regulation for the unstable Burgers equation (Liu, 2021).

6. Key Theorems and Representative Results

The body of work establishes the following generic result (specialized from (Liu, 2021, Kundu et al., 2018, Singh et al., 1 Feb 2026)):

• For any $u_0 \in H^2$, continuous functions $a(x,t), u_d(x,t)$, and $r(t)\in C^1$, and for any feedback gain $k>1/6$ (cubic feedback), the closed-loop Burgers system under the described boundary controller possesses a unique global solution with exponentially decaying tracking error,

$$|e(t)| \leq \|u_0 - r(0)\| e^{-\lambda t / 2},\quad \lambda = \min\{\nu, k - 1/6\}.$$

For the discrete approximations, the scheme retains exponential decay and achieves first- or second-order accuracy consistent with discretization design.

7. Connections and Broader Implications

The nonlinear Neumann boundary feedback strategy for the viscous Burgers equation exemplifies a broader program in PDE control: stabilization and output regulation of nonlinear, possibly unstable, parabolic PDEs using explicit boundary feedback and feedforward laws. The techniques have been generalized to higher-order systems (BBM–Burgers), equations with memory, and to adaptive and robust scenarios where key system parameters are not exactly known. The strong agreement between theory and numerics underscores the utility and robustness of these methods for both infinite-dimensional control theory and computational PDE stabilization (Liu, 2021, Kundu et al., 2018, Kundu et al., 2018, Singh et al., 2023, Singh et al., 1 Feb 2026, Singh et al., 29 Nov 2025, Singh et al., 1 Feb 2026, Kundu et al., 2019).