LMIs in Control/pages/Entropy Bound for Affine Parametric Varying Systems

The System

${\begin{aligned}{\dot {x}}(t)&=Ax(t)+B_{w}w(t),\\z(t)&=C_{z}(\theta )x(t)+D_{zw}(\theta )w(t),\end{aligned}}$

where $C_{z}$ and $D_{zw}$ depend affinely on parameter $\theta \in \mathbb {R} ^{p}$.

The Data

The matrices $A,B_{w},C_{z}(.),D_{zw}(.)$.

The Optimization Problem:

Solve the following semi-definite program

${\begin{aligned}&\min _{\{P\succ 0,\gamma ^{2},\lambda ,\theta \}}\gamma ^{2}\\&\quad s.t.\quad D_{zw}(\theta )=0,{\begin{bmatrix}A^{\top }P+PA&PB_{w}&C_{z}(\theta )^{\top }\\B_{w}^{\top }P&-\gamma ^{2}I&0\\C_{z}(\theta )&0&-I\end{bmatrix}}\preceq 0,\quad {\rm {{Tr}(B_{w}^{\top }PB_{w})\leq \lambda .}}\end{aligned}}$

Implementation

https://github.com/mkhajenejad/Mohammad-Khajenejad/commit/02f31a2d7a22b2464dfe9212eb76409bda9439b1

Conclusion

The value function of the above semi-definite program returns a bound for $\gamma$-entropy of the system, which is defined as

${\begin{aligned}I_{\gamma }(H_{\theta })\triangleq {\begin{cases}{\frac {-\gamma ^{2}}{2\pi }}\int _{-\infty }^{\infty }\log \det(I-\gamma ^{2}H_{\theta }(i\omega )H_{\theta }(i\omega )^{*})d\omega ,\quad {\text{if}}\ \|H_{\theta }\|_{\infty }<\gamma \\\infty ,\quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad {\text{otherwise.}}\end{cases}}\end{aligned}}$

Remark

When it is finite, $I_{\gamma }(H_{\theta })$ is given by ${\rm {{Tr}(B_{w}^{\top }PB_{w})}}$ where $P$, is asymmetric matrix with the smallest possible maximum singular value among all solutions of a Riccati equation.

External Links

• LMI Methods in Optimal and Robust Control - A course on LMIs in Control by Matthew Peet.
• LMI Properties and Applications in Systems, Stability, and Control Theory - A List of LMIs by Ryan Caverly and James Forbes.
• LMIs in Systems and Control Theory - A downloadable book on LMIs by Stephen Boyd.