Consider the optimal control problem of minimizing the functional
(1)
(2)
so that
We make the assumptions that the system (2) is time-invariant, has relative degree , is stabilizable, is detectable and that is nonsingular. (In lemma 3.1 below we show which systems can be represented in this form.) We want to study the limiting behavior of the solution to this problem, i.e. when the closed-loop system is stable, as tends to zero.
We begin with a heuristic analysis of the problem of finding a feedback control law that minimizes the functional (1) and stabilizes the system. Since is nonsingular, is completely controllable. Thus, we can use to control . If is bounded, the cost of control is small for small .
Suppose that is stable. Then we see that minimizes the functional. In this case we can find a bounded feedback control which makes the output zero. Hence the remaining dynamics
are just the zero dynamics of the original system (see section 2.1), which are stable by assumption. On the other hand, if is antistable, stabilizability of the system implies that the pair is controllable. Since is completely controllable, we can let be a control for and find an optimal by solving the reduced problem
(3)
(4)
The facts that is optimal and that is a feedback control of the variable suggest that the corresponding for the original problem (1)-(2) is optimal too. Unfortunately, since the system (4) is not detectable, an optimal solution to the reduced problem (3)-
(4) may not exist, see Wonham [2, p. 280]. However, it is easy to see that an optimal solution to the problem (3)-(4) does exist if we make the assumption that has no eigenvalues on the imaginary axis. Nevertheless, we will give a detailed description of the solution in the end of this section.
First, we motivate our analysis of this special case by stating the following lemma.
Lemma 3.1: Consider the -dimensional linear system
(5)
and suppose that the system has relative degree . (It is also natural to assume that the input matrix has
linearly independent columns.) Then the system (5) can always be transformed to the form (2)
Proof: We can always write
in the controllability canonical form (see e.g. Kwakernaak and Sivan [3, pp. 60-61]),
(6)
where , and is a non-singular matrix (since has full column rank). Furthermore, the pair is completely controllable. Partition as
The condition that the system has relative degree implies that and, by assumption, that the system has equal number of inputs and outputs (see section 2.1). Thus,
so that must be non-singular. Now we let
or
which, after substitution into (6), yields
(7)
Renaming the parameters in (7) shows the statement.
Solving the optimal control problem (1)-(2), we define the Hamiltonian function
where and satisfy the adjoint equations
(8)
Setting , we find that the optimal control law is
(9)
In order to study the behavior of the solution as goes to zero, we use singular perturbation methods (see section 2.2). We rescale the adjoint variable as , so that the adjoint equations (8) can be written
(10)
Multiplying the second equation in (2) by and using (9) and (10), we
obtain the singularly perturbated system
(11)
Since is nonsingular, we see that the system clearly satisfies the conditions of theorem 2.2 if we assume that the resulting Riccati equation has a unique positive semidefinite stabilizing solution. Therefore we set to obtain
(12)
(13)
Now there exists an invariant subspace such that
(14)
(15)
Substitution of (14) and (15) into (13) gives the reduced system
(16)
where P satisfies the Riccati equation
(17)
Referring to our heuristic solution, we show that the solution to the reduced problem (16)-(17) is exactly the solution to the problem
(18)
(19)
The Hamiltonian function for this problem is
where
The optimal control law is
Letting
and
we get the closed-loop system
where is the positive semidefinite solution (if it exists) to the Riccati equation
Thus, we have shown that the problem (1)-(2) reduces, as , to the reduced problem (18)-(19). Now we will turn to the question of existence and uniqueness of the solution to the reduced problem.
Proposition 3.1: Consider the Riccati equation (17). If the matrix does not have any eigenvalues
on the imaginary axis, there exists a unique, positive semidefinite solution to the Riccati equation which stabilizes the system (16).
Proof: Note that the detectability condition of theorem 2.5 is not satisfied for the problem (18)-(19). Therefore, we consider the three possible cases where is stable (all eigenvalues of lie in the open left-hand complex plane), antistable (all eigenvalues of lie in the open right-hand complex plane – we assumed that does not have any eigenvalue on the imaginary axis), and when has some eigenvalues in the open left-hand complex plane and some eigenvalues in the open right-hand complex plane.
Case 1:
Suppose that . Then the Riccati equation (17) has the unique positive semidefinite solution and (16) becomes
Remark:
We see that this is just the zero dynamics of the system. Using the method outlined in section 2.1 on the system (2)
(20)
we set
, so that too. Then
Since is nonsingular we choose
Changing coordinates according to
(21)
we obtain
(22)
Since , the zero dynamics of the system are described by
Case 2:
Suppose that . Then must be controllable due to the stabilizability condition. Consider the Lyapunov equation
(23)
By integrating the expression
using the Lyapunov equation (23), the stability of and the fact that , we see that
is a solution. This is just the controllability Grammian of . (See e.g. Wonham [2,~pp.~277-278].) Furthermore, if is controllable, we have that
is positive definite for every (see Wonham [2, p. 38]).
Thus is positive definite and therefore exists. If we let and multiply the Lyapunov equation by from the right and from the left, we see that this is just the reduced Riccati equation (17). Since is nonsingular, we have from (17)
so that (16) becomes
Since the transpose of a matrix has the same eigenvalues as the original matrix and the similarity transformation does not change the eigenvalues, the reduced system has the same eigenvalues as , but reflected in the imaginary axis.
Writing the equation as
and referring to (13), we see that this is just the adjoint equation of .
Case 3:
Finally, suppose that is unstable with some stable and some unstable eigenvalues. Define the modal subspaces
(see Francis [4, pp. 85-86]) of as
(24)
where is the characteristic polynomial of . The factor has all its zeros in Re and has all its zeros in Re . (There are no zeros on the imaginary axis.) It can be shown that is spanned by the generalized real eigenvectors of corresponding to eigenvalues in
Re and similarly for . These two modal subspaces are complementary, i.e. they are independent and their sum is all of . Thus
Now, because is supposed not to have any eigenvalues on the imaginary axis, there exists a similarity transformation such that has the form
where is stable and is antistable. Partitioning and accordingly, the algebraic Riccati equation (17) can then be written
(25)
If we can find a positive semidefinite matrix which stabilizes the system (16), i.e. makes stable, and satisfies the Riccati equation (25), must be unique. Suppose that has the form
We see that we have the conditions
(26)
From case 1 and 2, we see that the Riccati equation (25) and the stability conditions (i) and (ii) are satisfied if we choose and as the positive semidefinite solution to the equation
(27)
Such a solution exists, since the pair is stabilizable by assumption. Thus, we have the resulting system
(28)
It is easily seen that the closed-loop system is stable since its eigenvalues are exactly those of the stable matrix and those of the antistable matrix with negative sign.
To interpret this result, first we observe that due to stabilizability, the unstable subspace must lie in the controllable subspace, , but we can not conclude that the intersection of the stable subspace and the
controllable subspace is empty.
Consider the decomposition of case 3 into the stable and unstable modal subspaces. Suppose that , although this in general is not true. Then we simply have and we can put the system (19) in its controllability canonical form, i.e.
(29)
where the zeros in the matrices follow from and the -invariance of . Therefore we have
(30)
which corresponds to in case 3. This would give the reduced system
(31)
Furthermore, we observe that , where is the controllability Grammian. Thus, we can interpret the above results as follows:
In the subspace , the optimal control gives the zero dynamics restricted to the stable subspace, in we have the adjoint equation subject to a similarity transformation by the inverse of the controllability Grammian of , i.e. of restricted to . Finally, the subspace is naturally also affected by the minimum control which, however, does not change the stability properties of the matrix .
To conclude, we observe that the conditions for -bounding are satisfied. Considering the decomposition and lemma 2.2 of section 2.3 we see that
Furthermore, from the comment following theorem 2.4 and the fact that is square and has full rank, the condition that has no zeros on the imaginary axis is exactly the condition that does not have any eigenvalues on the imaginary axis, since the eigenvalues of are the transmission zeros of the system.