Brockett openness laboratory

How much control can the openness profile support?

The open-loop field supplies a scale-dependent ball of velocities. A gain ceiling limits how far the feedback graph can reach into that field. Move the controls to compare the desired closed-loop ball with the open-loop capacity.

Graph-limited Brockett inequality

The feedback graph bounds the closed-loop profile.

Let Gu(x)=(x,u(x)) be the feedback graph and Fu=f ∘ Gu the closed-loop field. If ‖u(x)‖≤d(‖x‖), then

ΩFu(r) ≤ Ωf(√r2+d(r)2)

Exact radial example

Resize the image, target, and gain envelope.

Here the open-loop image is an ellipse whose largest centered ball is known exactly. A green result means this necessary test does not rule the target out; it is not a general sufficiency claim.

Parameters
Gain-envelope shape
At the selected radius

The open-loop image and its centered ball

f(BR) openness ball desired ball
Across nearby scales

Target openness versus available capacity

κ(r2+d(r)2) cr selected r
Control ceiling d(r) = 0.253
Joint radius reached R = 0.272
Desired ball cr = 0.080
Open-loop capacity κR2 = 0.059
Ruled out near the equilibrium This gain envelope is too small.

Why the geometry is exact

A simple system realizes the threshold.

Open-loop field fκ(x,u)=‖u‖Aκu

For Aκ=κ·diag(1.6,1), a joint ball of radius s maps to an ellipse with semiaxes 1.6κs2 and κs2. Therefore Ωf(s)=κs2 exactly.

Exact linear closed loop fκ(x,uc(x))=−cx

Taking y=−cAκ−1x and uc(x)=y/√‖y‖ produces the target field. Set uc(0)=0. Its magnitude is bounded by √c/κ‖x‖.

Asymptotic threshold K ≥ √(c/κ) = 1.00

This is a minimum required size for an upper gain envelope—not a maximum allowed gain. A linear envelope Kr is too small near zero for every finite K.

Outside this exact example, satisfying the displayed inequality only means that this Brockett obstruction is silent. Other obstructions may still prevent stabilization.