Pole placement

Pole placement (also called pole assignment or eigenvalue assignment) is a fundamental technique in control theory whereby a control engineer designs a state feedback controller to place the closed-loop eigenvalues of a system at desired locations in the complex plane. Given a linear time-invariant system $\dot{x} = Ax + Bu$, state feedback $u = -Kx$ yields the closed-loop dynamics $\dot{x} = (A - BK)x$. The goal of pole placement is to choose the gain matrix $K$ such that the eigenvalues of $(A - BK)$ — the closed-loop poles — achieve specified values, thereby enforcing desired transient response characteristics: settling time, overshoot, damping ratio, and natural frequency.

The theoretical foundation for pole placement was established independently by Wonham (1967) and others, who proved that arbitrary pole placement is possible if and only if the pair $(A, B)$ is controllable — meaning that the system's controllability matrix has full rank. This condition ensures that the input $u$ can influence every mode of the system, and thus that every eigenvalue can be shifted by feedback. When the system is not fully controllable, only the controllable eigenvalues can be placed; the uncontrollable eigenvalues remain fixed and must already be stable for the closed-loop system to be stabilizable.

In practice, pole placement is often computed using the Ackermann's formula for single-input systems, or via Bass-Gura and Kautsky-Nichols-Van Dooren algorithms for multi-input systems. The method remains a cornerstone of classical control design, though modern approaches increasingly favor optimal control (LQR) and robust control ($H_\infty$) methods that balance pole placement against other performance criteria.

Pole placement is the control engineer's fantasy of total authority: the belief that if you can measure the state and compute the gain, you can rewrite the system's destiny. This fantasy is mathematically correct for linear systems, but it is practically fragile. Pole placement assumes perfect state measurement, ignores actuator saturation, and treats model uncertainty as a nuisance rather than a fundamental constraint. The elegance of the pole placement theorem — that controllability implies arbitrary eigenvalue assignment — has seduced generations of control theorists into believing that instability is a choice rather than a property of the world.