Computing Library › Control Theory
Control Theory

Duality of Controllability and Observability

Controllability and observability are mirror images: each result about one becomes a result about the other by transposing the system.

A Structural Symmetry

Controllability and observability are not merely analogous, they are formally dual. The system (A, B, C) is observable if and only if the transposed system (A-transpose, C-transpose, B-transpose) is controllable. Every theorem about steering the state has a twin about reconstructing it.

The dual pairings

Kronos motion — control room

Why this is useful

Duality halves the theoretical work: prove a result once for controllability, transpose it, and the observability version follows for free. It also unifies software, since the same numerical routine that computes an optimal feedback gain can compute an optimal observer gain by feeding it the transposed matrices.

The regulator-estimator mirror

The clearest practical instance is the linear-quadratic-Gaussian design. The optimal control gain comes from a Riccati equation in (A, B); the optimal Kalman gain comes from a Riccati equation in (A-transpose, C-transpose). The two problems are mathematically identical up to transposition, which is why one Riccati solver serves both.

Conceptual payoff

Duality expresses a deep symmetry between influencing a system and observing it. Inputs push the state around; outputs read it out. The transpose operation swaps these roles cleanly, revealing that the geometry of what you can reach and the geometry of what you can see are two faces of the same structure.

Recognizing duality lets an engineer transfer intuition instantly: a hard-to-control mode and a hard-to-observe mode arise from the same kinds of structural degeneracy, and the fixes, adding an actuator or adding a sensor, are dual remedies.