Pole Placement State Feedback Control

• Pole placement-based state feedback control is a design methodology that assigns closed-loop eigenvalues to achieve prescribed stability, transient, and robustness characteristics.
• It employs analytical tools such as Ackermann’s formula, Sylvester equations, and eigenstructure assignment to compute the static gain matrix for controllable LTI systems.
• Advanced applications include robust and data-driven control, sparse and minimum-gain feedback designs, and extensions to infinite-dimensional and periodic systems.

Pole placement-based state feedback control is a set of methodologies for synthesizing a static state feedback law that assigns the closed-loop eigenvalues (poles) of a given multi-input linear time-invariant (LTI) system to user-specified locations in the complex plane. Through careful construction of the gain matrix, this strategy ensures that the closed-loop system exhibits prescribed stability, transient, and robustness characteristics. Pole placement is fundamental to modern control theory and underlies advanced methods such as robust control, observer design, and feedback linearization.

1. Fundamental Problem Formulation

Given a controllable LTI system $\dot{x}(t) = A x(t) + B u(t)$ with $x \in \mathbb{R}^n$, $u \in \mathbb{R}^m$, the goal is to determine a feedback matrix $K \in \mathbb{R}^{m \times n}$ such that the closed-loop matrix $A_{\mathrm{cl}}=A-BK$ has an arbitrarily prescribed spectrum $$\sigma(A_{\mathrm{cl}})=\{\lambda_1,\,\dots,\lambda_n\}$$—modulo controllability limitations. This assignment can be extended to include desired geometric (invariant) structure for robustness, as well as pole multiplicity, under suitable controllability and structural conditions (Schmid et al., 2013, Schmid et al., 2014).

A crucial insight is that, for single-input systems, $K$ is unique (see companion-form and Ackermann’s formula), while for multi-input systems ($m>1$) the solution set is generally a proper affine or algebraic submanifold, parameterized by a number of degrees of freedom tied to the system’s controllability indices and the chosen closed-loop similarity class (Baragaña et al., 25 Nov 2025).

2. Classical and Modern Synthesis Approaches

2.1. Ackermann’s Formula and the Sylvester Equation

Ackermann’s formula gives a closed-form expression for $K$ in the single-input case under full controllability, based on the controllability matrix and the coefficients of the desired characteristic polynomial. For multi-input cases and more general spectral assignments (including repeated and defective eigenvalues), synthesis reduces to solving a parameterized family of Sylvester-type equations $AX - X\Lambda + BG = 0$, where $\Lambda$ is the Jordan form of the prescribed spectrum, $G$ is a block parameter matrix, and $K = GX^{-1}$ for an invertible $X$ (Schmid et al., 2014).

2.2. Moore Parametrization and Eigenstructure Assignment

The Moore eigenstructure assignment technique constructs all feedback matrices achieving both the desired eigenvalues and specified eigenvectors (or more generally, root-space structure). The core method proceeds by:

• For each $\lambda_i$, compute the system matrix $S(\lambda_i) = [A - \lambda_i I~|~B]$.
• Find bases $T_i$ of the null space for each $S(\lambda_i)$.
• Stack all $T_i$ to form $T$ and introduce a free parameter matrix $K$.
• Construct $M(K) = TK$ and partition $M(K)$ into $V(K)$ (upper $n$ rows) and $W(K)$ (lower $m$ rows).
• For almost every $K$, $V(K)$ is invertible, and every pole-placing gain is $F(K) = W(K)V(K)^{-1}$ (Schmid et al., 2013, Schmid et al., 2014).

This yields a complete parametric description of the solution set and efficiently spans all admissible feedback matrices.

2.3. Geometric and Manifold-Theoretic Characterizations

The set of all $K$ realizing a prescribed closed-loop similarity class forms a smooth manifold with an explicit local parametrization. A diffeomorphism is constructed between admissible feedbacks and orbits in the space of truncated observability matrices modulo the action of the closed-loop centralizer. The manifold’s dimension is $mn-N$, where $N$ is the dimension of the centralizer of a canonical representative in the closed-loop similarity class (Baragaña et al., 25 Nov 2025).

3. Robustness and Optimization Aspects

Pole placement is inherently non-unique in the multi-input case, enabling selection of a feedback matrix that optimizes additional criteria beyond simple spectral assignment:

• Robustness to perturbations: Typical measures include the condition number $\kappa_F(V) = \|V\|_F \|V^{-1}\|_F$ of the eigenvector (modal) matrix $V$, as closed-loop eigenvalue sensitivity is directly proportional to this (Bauer–Fike theorem).
• Feedback gain minimization: Minimize norm $\|F\|_F$ to reduce actuation effort.
• Departure from normality: The Schur-based robust pole placement minimizes $\|A+BF\|_F^2 - \sum |\lambda_j|^2$, seeking closed-loop matrices close to normal, which are less sensitive to perturbations (Zhen-chen et al., 2014, Schmid et al., 2014).
• Weighted objectives: Combined criteria of the form $$J(K) = \alpha \kappa_F(V(K)) + (1-\alpha)\|F(K)\|_F$$, with $0 \leq \alpha \leq 1$, enable explicit robustness–gain tradeoffs (Schmid et al., 2013, Schmid et al., 2014).

These optimization problems are generally smooth but non-convex and are typically attacked with gradient-based or quasi-Newton methods over the free parametrization variables (Schmid et al., 2013, Schmid et al., 2014).

4. Extensions and Variants

4.1. Data-Driven Pole Placement

Recently developed methodologies solve pole placement directly using finite trajectory data, for both continuous and discrete LTI systems:

• Data matrices are built from input–state samples under persistently exciting inputs.
• The Sylvester and Moore-type parameterizations are replicated using only data, enabling direct computation of $K$ through null-space constructions and data-driven Sylvester equations.
• Robustness is achieved by optimizing the condition number of the data-based modal matrix (Lopez et al., 2024, Bianchin, 2023).

4.2. Structured and Minimum-Gain Feedback

When additional structural constraints (such as sparsity) are imposed on $K$, the pole placement task becomes a constrained optimization problem (sparse MGEAP), which is addressed through projected gradient or alternating projection algorithms, using Sylvester-equation parametrizations to maintain spectral placement despite the structural restrictions (Katewa et al., 2018).

4.3. Infinite-Dimensional and Periodic Systems

Pole placement extends to infinite-dimensional or time-periodic systems by mapping the original system to an infinite-dimensional time-invariant harmonic (Toeplitz operator) model, then assigning the spectrum via an infinite Sylvester equation. Unique challenges include the loss of automatic invertibility and additional constraints on the feedthrough operator (Riedinger et al., 2022, Chen, 2024).

4.4. Applications in Advanced Control Architectures

• Nonlinear Feedback Linearization: Pole placement is directly employed after input–output linearization of nonlinear systems, often requiring special handling under state constraints, as constraints can destroy relative degree at boundaries. A switching feedback plus pole-placement solution is effective in this regime (Jin et al., 5 Sep 2025).
• Block Pole Placement in MIMO Systems: Block-solvent allocations, as for missile servomechanisms, generalize pole placement by assigning blocks of eigenvalues to achieve both cooperative decoupling and robust transient performance (Bekhiti et al., 2016).
• Hybrid and Large-Scale Systems: High-dimensional systems such as microgrids are synthesized with pole-placement-based feedback using large-scale versions of standard algorithms (such as Ackermann’s formula) (Tripathy et al., 1 Sep 2025).

5. Algorithms, Tools, and Comparative Performance

Several algorithmic approaches have been developed for efficient, robust pole placement:

• Span (Moore's parametric method): Exploits null-space parameterization and gradient-based robustification. Outperforms classical tools such as MATLAB’s place in terms of both eigenvalue sensitivity and gain, especially in systems with uncontrollable modes present in the desired spectrum (Schmid et al., 2013).
• Modified Schur Methods: Constructs a real-Schur form for the closed-loop matrix and explicitly minimizes departure from normality, leading to improved robustness for clustering and complex pole assignment (Zhen-chen et al., 2014).
• Sylvester-based and Klein–Moore Parametrizations: Form the analytical and computational core for robust and minimum-gain eigenvalue assignment in both dense and sparse feedback settings (Schmid et al., 2014, Katewa et al., 2018).

Quantitative performance comparisons on standard benchmarks show up to 60% improvement in modal conditioning and up to 30–50% reduction in gain over classical state-space pole placement methods. The most advanced algorithms are computationally efficient, handling large state dimensions (e.g., $n \sim 60$) without prohibitive cost (Schmid et al., 2013, Schmid et al., 2014, Zhen-chen et al., 2014).

6. Practical Implementation Considerations

When implementing pole placement for control synthesis, the following factors impact feasibility and performance:

• Controllability: Assignment is possible if and only if $(A,B)$ is controllable with respect to the prescribed pole set and multiplicities (Shaul et al., 2017).
• Numerical Conditioning: Assignment quality degrades if the eigenvector matrix is ill-conditioned; optimization over the free parameters mitigates this effect.
• Sensitivity and Robustness: Minimizing condition number or maximizing modal orthogonality of the closed-loop eigenmatrix is critical for robustness against unmodeled dynamics and parametric uncertainty (Schmid et al., 2013, Schmid et al., 2014).
• Structural Constraints: Imposing sparsity or block-diagonalization can be handled by projection and penalty methods in the optimization (Katewa et al., 2018).
• Data Requirements: In data-driven scenarios, input excitation must be persistently exciting of sufficiently high order to span the system’s controllable subspace (Lopez et al., 2024, Bianchin, 2023).

Failing to properly handle these aspects can lead to large control gains, poor robustness, or even infeasibility of the pole placement assignment.

7. Outlook and Advanced Topics

Areas of ongoing research and open questions include:

• Systematic characterization and algorithmic synthesis for high-dimensional, structured, and infinite-dimensional plant models.
• Efficient and numerically robust solutions under explicit feedthrough constraints and partial state measurements.
• Improved data-driven algorithms for closed-loop spectral assignment in the presence of noise and unmodeled dynamics, with provable trade-offs in robustness and performance.
• Geometric parametrization and Lie-theoretic classification of the entire feedback assignment manifold for use in controller tuning, observer design, and higher-order performance specification (Baragaña et al., 25 Nov 2025).

Pole placement-based state feedback remains a central, evolving methodology in control system design, underpinning both classical and modern techniques for robust, optimal, and data-driven control synthesis. Recent advances in parametric optimization, geometric analysis, and computational methods continue to extend its scope and practical impact across the spectrum of control applications.

---

References:

(Schmid et al., 2013) Robust pole placement with Moore's algorithm (Schmid et al., 2014) A unified method for optimal arbitrary pole placement (Zhen-chen et al., 2014) A Modified Schur Method for Robust Pole Assignment in State Feedback Control (Lopez et al., 2024) Data-Based Control of Continuous-Time Linear Systems with Performance Specifications (Bianchin, 2023) Data-Driven Exact Pole Placement for Linear Systems (Baragaña et al., 25 Nov 2025) A Local Parametrization of the State-Feedback Matrices in the Pole Assignment Problem (Katewa et al., 2018) Minimum-gain Pole Placement with Sparse Static Feedback (Riedinger et al., 2022) Harmonic Pole Placement (Bekhiti et al., 2016) 2-DOF block pole placement control application to: have-dash-II missile (Tripathy et al., 1 Sep 2025) A Mathematical Model of Hybrid Microgrid With Pole Placement Controller Using State Feedback For Stability Improvement (Shaul et al., 2017) Pole placement for overdetermined 2D systems (Jin et al., 5 Sep 2025) Feedback Linearisation with State Constraints (Chen, 2024) Pole Placement and Feedback Stabilization for Discrete Linear Ensemble Systems