Balanced realizations of regime-switching linear systems
read more
Citations
Matrix Differential Calculus with Applications in Statistics and Econometrics
John Barratt Moore 1941–2013
References
Principal component analysis in linear systems: Controllability, observability, and model reduction
Optimal Control: Linear Quadratic Methods
Deterministic and stochastic optimal control
All optimal Hankel-norm approximations of linear multivariable systems and their L, ∞ -error bounds†
Matrix Differential Calculus with Applications in Statistics and Econometrics
Related Papers (5)
Frequently Asked Questions (10)
Q2. What was the first study of balanced realizations?
In the 1980s, to bridge the gap between minimal realization theory and to treat the problem of finding lower order approximation, a “canonical” realization was first introduced in the seminal paper [13], where the problem was studied with principal component analysis of linear systems.
Q3. What is the meaning of the term balance?
The term “balance” was used since the realizations have certain symmetry between the input and the output maps characterized by the controllability and observability Grammians.
Q4. How can the authors use the -balancing to approximate (2.9)?
Using inherent fast and slow time scales, first let ε → 0, the authors obtain a limit ρ-balancing system, and then the authors can use the ρ-balancing to approximate (2.9) as in the last section.
Q5. Why are there infrequent transitions between ergodic classes?
due to the weak interactions resulted from the generator Q̂, there are still infrequent transitions from one ergodic class Mi to another ergodic class
Q6. What is the problem with -balancing systems?
Even if the authors construct ρ-balancing systems of equations, the large-scale nature of the systems can render the computation task infeasibledue to the large dimensionality.
Q7. Why is it not feasible for all t?
The reason is that the authors need to solve (2.9) for all t not just a fixed t or finitely many t. Directly solving this system is not feasible for all t. Besides, compared to the deterministic counter part, it is a coupled system of equations, which adds another fold of difficulty.
Q8. What is the simplest way to solve the Riccati equations?
(2.16)Although (2.9) provides an exact system of balancing equations, since it is time-varying, the solution must be obtained for all t.
Q9. What is the last term of (2.9)?
qm1In qm2In . . . qmmIn ∈ R(mn)×(mn),and(Q⊗ In)P (t)Îmn = ( m∑j=1q1jP (t, j), . . . , m∑j=1qmjP (t, j) )′def = Pmn(t) ∈ R(mn)×n.To obtain the last term in (2.9), note that(Q⊗ In)P (t)ÎmnI0 = ∑m j=1 q1jP (t, j) 0n · · · 0n ∑m j=1 q2jP (t, j) 0n · · · 0n· · · · · · · · · · · · ∑mj=1 qmjP (t, j) 0n · · · 0n ∈ R(mn)×(mn),where 0n is an n × n zero matrix, and I0 = (In, 0n, . . . , 0n) is an n × (mn) matrix.
Q10. What are the main applications of random jump linear systems?
Such random jump linear systems arise frequently in wireless communications, manufacturing systems, financial engineering, and other applications, where regime changes are utilized to formulate the random environment.