Moore Eigenstructure Assignment Technique

• Moore eigenstructure assignment is a parametric nullspace-based method that prescribes closed-loop eigenstructures for LTI systems.
• It systematically solves the Sylvester equation, enabling robust pole placement and optimization over feedback gain parameters.
• The technique supports modern applications such as multi-agent formation control and sparse, data-driven feedback design.

The Moore eigenstructure assignment technique is a foundational parametric approach for solving the pole placement and eigenstructure assignment problems in linear time-invariant (LTI) control systems. Its key contribution is a parametrization of all feedback gain matrices that yield a prescribed closed-loop eigenstructure, formulated via nullspace methods. Developed initially by B.C. Moore in the 1970s for systems of the form $\dot{x}(t) = A x(t) + B u(t)$, the technique plays a critical role in robust pole placement, multi-agent formation control, and sparse feedback design.

1. Mathematical Foundations and Problem Statement

Consider the LTI system $\dot{x} = A x + B u$, with $A \in \mathbb{R}^{n \times n}$ and $B \in \mathbb{R}^{n \times m}$ (full column rank). The objective is to design a static state-feedback law $u = F x$ so that the closed-loop matrix $A_{cl} = A + B F$ admits a prescribed spectrum $\Lambda = \{\lambda_1, ..., \lambda_v\}$ (with algebraic multiplicities summing to $n$) and a corresponding set of eigenvectors or generalized eigenvectors. The assignment requirement is succinctly captured by the Sylvester equation: $(A + B F) X = X \Lambda$, with $X$ nonsingular. The full eigenstructure assignment problem seeks feedback $\dot{x} = A x + B u$0 and eigenvector matrix $\dot{x} = A x + B u$1 such that this holds for the desired $\dot{x} = A x + B u$2 and $\dot{x} = A x + B u$3.

2. Classical Moore Eigenstructure Assignment Algorithm

Moore’s insight is that for each eigenpair $\dot{x} = A x + B u$4, enforcing $\dot{x} = A x + B u$5 leads to the constraint

$\dot{x} = A x + B u$6

Stacking $\dot{x} = A x + B u$7 as columns yields a block equation for all assigned eigenpairs. For each $\dot{x} = A x + B u$8, define the system matrix

$\dot{x} = A x + B u$9

and let $A \in \mathbb{R}^{n \times n}$0 be a basis for $A \in \mathbb{R}^{n \times n}$1. Introducing a block-diagonal parameter $A \in \mathbb{R}^{n \times n}$2 with $A \in \mathbb{R}^{n \times n}$3, the general nullspace solution is

$A \in \mathbb{R}^{n \times n}$4

where $A \in \mathbb{R}^{n \times n}$5 and $A \in \mathbb{R}^{n \times n}$6 are obtained by row partitioning. When $A \in \mathbb{R}^{n \times n}$7 is invertible,

$A \in \mathbb{R}^{n \times n}$8

parametrizes all feedbacks $A \in \mathbb{R}^{n \times n}$9 that assign the specified eigenstructure, encapsulating all degrees of freedom in the free parameter $B \in \mathbb{R}^{n \times m}$0 (Schmid et al., 2013).

For complex conjugate eigenpairs, real feedbacks are enforced by pairing and appropriate realification of the nullspace bases and parameters, ensuring $B \in \mathbb{R}^{n \times m}$1 is real (Schmid et al., 2013).

3. Connection to Nullspace and Sylvester Equation Methods

Moore’s construction elucidates the solvability conditions of the assignment problem: existence of a basis for the nullspace of $B \in \mathbb{R}^{n \times m}$2, and parameterization of all admissible eigenvectors through free columns of $B \in \mathbb{R}^{n \times m}$3 [(Schmid et al., 2013); (Motoyama et al., 2017)]. The Sylvester equation $B \in \mathbb{R}^{n \times m}$4 is equivalent to $B \in \mathbb{R}^{n \times m}$5 with $B \in \mathbb{R}^{n \times m}$6; solutions can be explicitly parameterized via nullspace methods. This approach both generalizes and structurally clarifies alternatives such as the Brockett–Davis–Marden and Klein–Moore algorithms.

4. Extensions, Special Cases, and Robustness Optimization

Because the Moore parametric form characterizes the full set of solutions, it enables optimization over secondary criteria such as robustness and gain magnitude. Robust feedback is typically associated with minimizing the sensitivity of the closed-loop eigenvalues, for instance, by minimizing the Frobenius condition number $B \in \mathbb{R}^{n \times m}$7. Optimization formulations include:

• $B \in \mathbb{R}^{n \times m}$8 for robustness,
• $B \in \mathbb{R}^{n \times m}$9 to trade off robustness and control effort (Schmid et al., 2013).

Gradient-based algorithms optimize $u = F x$0 via chain rule differentiation, exploiting the explicit parametric structure. Notably, these methods maintain full pole-placement accuracy and handle uncontrollable modes (by augmenting the nullspace dimension $u = F x$1).

5. Algorithmic Workflow and Computational Aspects

The Moore eigenstructure assignment process (for the generic case with simple eigenvalues) involves:

1. For each desired $u = F x$2, compute a basis $u = F x$3 for $u = F x$4.
2. Solve $u = F x$5 for $u = F x$6.
3. Set $u = F x$7 and set $u = F x$8.
4. Upon stacking, obtain $u = F x$9 (Motoyama et al., 2017, Celi et al., 2023).

For repeated or defective eigenvalues, the technique is extended to handle Jordan chains by solving $A_{cl} = A + B F$0 for lengths of the chain (Motoyama et al., 2017).

Implementation complexity is dominated by computation of basis matrices and $A_{cl} = A + B F$1 matrix inversions, with overall cost $A_{cl} = A + B F$2.

6. Applications: Robust Pole Placement, Multi-Agent Formation, and Sparse Feedback

Robust pole placement: The Moore technique has been adapted via parametric nullspace methods to robust pole placement, outperforming QR/Sylvester (MATLAB’s place), LMI-based (rfbt), and heuristic (roppole) methods in both matrix conditioning and control gain size. The “span” toolbox implementation achieved the best Frobenius condition and gain reduction across small benchmark and large random systems, particularly excelling in systems with uncontrollable modes where alternative methods fail (Schmid et al., 2013).

Multi-agent formation control: Moore’s method provides the structural toolset for assigning closed-loop eigenstructure in multi-agent systems, synthesizing sparse topologies (star, cycle, line) tailored via eigenvector selection. Hierarchical algorithms partition large networks, assigning group-level eigenstructures for highly efficient synthesis and scalable performance (e.g., $A_{cl} = A + B F$3 agents, achieved with $A_{cl} = A + B F$4s computation vs. $A_{cl} = A + B F$5s by centralized approaches) (Motoyama et al., 2017).

Sparse feedback and data-driven design: Recent advancements replace model-based assignment with data-driven characterization, constructing the allowable eigenvector subspaces directly from input–state trajectories, and providing closed-form feedback solutions, as well as optimization-based approaches for minimizing gain norm or enforcing sparsity. Classical Moore structure serves as the template for these generalizations (Celi et al., 2023).

7. Theoretical Completeness, Generalizations, and Comparison

Moore’s eigenstructure assignment is both necessary and sufficient for exact eigenstructure placement given controllability and compatible eigenvector choices. Extensions include:

• Surgical assignment: Assign selected eigenpairs (or Jordan chains) via “locking in” partial eigenstructure, then completing the spectrum without disturbing the assigned part, as detailed in the surgical assignment methodology (Maruf et al., 2020).
• Data-driven generalization: Admissible eigenvectors are characterized as elements of subspaces parameterizable from linear constraints on data gathered from open-loop experiments (Celi et al., 2023).
• Robust nullspace-based methods: These handle both controllable and uncontrollable modes comprehensively, offering theoretical guarantees not matched by heuristic or LMI approaches (Schmid et al., 2013).

A summary of comparative properties drawn from the cited literature is given below.

Method	Robustness (κ_X)	Handles Uncontrollable Modes	Sparse Assignment	Data-Driven Option
Moore (Nullspace)	High	Yes	Parametric	Extensions exist
QR/Sylvester (place)	Lower	No	No	No
LMI (rfbt)	Moderate	Limited	Yes (indirect)	No
Data-driven (Celi et al., 2023)	As designed	Yes (under controllability)	Direct	Yes

In all, the Moore eigenstructure assignment technique, through its parametric nullspace formulation, remains a theoretically complete and practically effective foundation for diverse pole placement, eigenstructure assignment, and robust/sparse feedback design problems [(Schmid et al., 2013); (Motoyama et al., 2017); (Celi et al., 2023); (Maruf et al., 2020)].