Mixed H2/H∞ synthesis with pole placement constraints
[gopt,h2opt,K,R,S] = hinfmix(P,r,obj,region,dkbnd,tol)
h2hinfyn performs multi-objective
output-feedback synthesis. The control problem is sketched in this
denote the closed-loop transfer functions from w to z∞ and z2,
hinfmix computes a suboptimal solution
of the following synthesis problem:
Design an LTI controller K(s) that minimizes the mixed H2/H∞ criterion
∥T∞∥[[BULLET]] < γ0
∥T2∥2 < ν0
The closed-loop poles lie in some prescribed LMI region D.
Recall that ∥.∥∞ and ∥.∥2 denote the H∞ norm (RMS gain) and H2 norm of transfer functions.
P is any SS, TF, or ZPK LTI representation
of the plant P(s), and
a three-entry vector listing the lengths of z2, y,
and u. Note that z∞ and/or z2 can
be empty. The four-entry vector
obj = [γ0,
ν0, α, β] specifies the H2/H∞ constraints
and trade-off criterion, and the remaining input arguments are optional:
region specifies the LMI region
for pole placement (the default
region =  is
the open left-half plane). Use
lmireg to interactively
build the LMI region description
dkbnd is a user-specified bound
on the norm of the controller feedthrough matrix DK.
The default value is 100. To make the controller K(s)
strictly proper, set
dkbnd = 0.
tol is the required relative accuracy
on the optimal value of the trade-off criterion (the default is 10–2).
h2hinfsyn returns guaranteed H∞ and H2 performances
well as the
the LMI-optimal controller. You can also access the optimal values
of the LMI variables R, S via
the extra output arguments
A variety of mixed and unmixed problems can be solved with
In particular, you can use
hinfmix to perform pure
pole placement by setting
obj = [0
0 0 0]. Note that both z∞ and z2 can
be empty in such case.
Chilali, M., and P. Gahinet, "H∞ Design with Pole Placement Constraints: An LMI Approach," IEEE Trans. Aut. Contr., 41 (1995), pp. 358–367.
Scherer, C., "Mixed H2/H-infinity Control," Trends in Control: A European Perspective, Springer-Verlag (1995), pp.173–216.