Optimal State Regulator through the Matrix Riccati
Equation
Submitted by
Submitted to Priyanka(221515)
Dr. Shimi S.L. Ram Swaroop Bishnoi(221516)
Assistant Professor
Ravi Nath(221517)
M.E. in instrumentation and control
(Modular)
Department of Electrical Engineering NITTTR Chandigarh
OVERVIEW
• For an important class of control problems, which involves control of linear time-invariant
plants with the objective of minimizing the integral of a quadratic performance index, the
optimal control is possible by state-feedback control that results in a linear time-invariant
closed-loop system.
• Example – Error between the actual and desired state, other restrain such as Instantaneous
Energy Expenditure.
• The LQR algorithm minimizes the work-load of the control systems to optimize the controller.
• The system needs to be initiated withdraw parameters and test the efficiency of the control
system for the desired aim.
• The execution process is iterative and needs to find an optimal control through simulation and
then rearrange the parameters to achieve feasible output of the system.
• Many problems of industrial control belong to this class of problems—linear quadratic
optimal control problems.
To prove this important result, we consider the optimal control problem for a linear
multivariable completely controllable plant
where x is the nstate vector,
u is p 1 input vector;
A is n n real constant matrices,
B is n p real constant matrices,
and the null state x = 0 is the desired steady state.
The quadratic cost functional is defined
The objective is to find the optimal control law that minimizes the following
performance index, subject to the initial conditions
RU) (1)
The quadratic performance index
RU)
The quadratic and square function is employed for the performance index for three
reasons
(a)Squaring the error weights larger error more heavily than small error.
(b)Squared error represents system power consumption.
(c)Square error is much easier to handle analytically.
Norm
All states of system are equally important therefore we take only .
The energy expenditure performance index is RU.
is for ease of mathematical manipulations and also it does not affect the
minimization problem.
RU) (1)
where Q is n n positive definite, real, symmetric, constant matrix (It adds weightage to
states. In simplest form Q is a diagonal matrix )
R is p p positive definite, real, symmetric, constant matrix (R is a Restrain on the
input matrix U)
Since the (A, B) pair is completely controllable, there exists a state-feedback control law
where K is p n real, constant, unconstrained gain matrix, that results in an asymptotically
stable closed loop system
This implies that there is a Lyapunov function
for the closed-loop system ; that is, for some positive definite matrix P, the time derivative dV/dt
evaluated on the trajectories of the closed-loop system is negative definite.
The condition for to be optimal
If the state-feedback controller u = – Kx (t) is such that it minimizes the function
(2)
and the minimum value of f (u) = 0 for some then the controller is optimal P
Differentiating the above yields
+Q+Ru)
= (+ Q+Ru) , Now (=A
=Q+Ru))
==0=
Hence, a candidate for an optimal control law has the form
u=Px−Kx
K= P (3)
where K= P (3)
Note that
( )==R , a positive definite matrix
Thus, the second-order sufficiency condition, for minimize , is satisfied.
We now turn our attention to finding an appropriate P. The optimal closed-loop system
has the form
= (A – B P) ;
Our optimal controller satisfies the equation
u* =0
Where, controller u= u* is optimal
Q+ u* =0
We substitute u*= P given by into the above equation, and represent it as
The above equation should hold for any x. For this to be true, we have to have
(4)
The above equation is referred to as algebraic Riccati equation.
In conclusion, the synthesis of the optimal linear state-feedback controller, minimizing
the performance index
RU)
requires solving the matrix Riccati equation.
The Design steps are
(a) Solve the matrix Riccati Equation (4) for the positive definition matrix.
(b) Substitue the matrix P in equation (3) the resulting equation gives the optimal
control law.
COMMENTS
• The matrix R is assumed positive definite.
• The plant is completely controllable.
• The matrix Q is positive definte. The matrix Q can also be positive semi-
definite matrix.
• The system should be controllable and observable and asymptotically stable.
• The Recatti equations (4) is anon –linear algebraic equations .Since P is a
symmetrical matrix therefore we need to solve only equations.
The solution equation (4) is not unique. The desired solutions is obtained
when P be positive definite. The positive definite solution of equation (4)
is unique.
The gain elements of matrix K corresponding to non measurable state
variable may be constrained to zero value (otherwise a state observer
will be required).Also some gains are fixed by physical constraints of the
system.
APPLICATIONS
1 Optimal control of power systems and the Riccati
equation
The load-frequency control problem (LFCP), in modern control
theory, can be expressed as an optimal-regulator problem, the
solution of which is a feedback linear control. Moreover, the
gain matrix is a particular expression of the positive definite
solution of the algebraic Riccati equation associated to the
regulator problem. Therefore, if one wants a real-time
implementation, it is necessary to build up efficient algorithms.
This deals with the stability and efficiency of the Schur method
in the resolution of such equations; however, in the simulation
we have carried out (three-area model), we did not get
[Link] Fault-Tolerant Control of High-Speed Trains Based on the
Here a robust fault-tolerant control strategy based on the solution of Riccati equation are
developed for the velocity tracking control of the high-speed trains subjected to the actuator
faults. For each separate actuator with occurrence of faults, a robust fault-tolerant control is
performed by an appropriate solution of Riccati equation. According to the proposed robust
fault-tolerant control law, a new feedback controller which can stabilize the faults system is
designed. It is simulated and proved to be very useful and effective by the satisfactory numerical
simulation results of the robust fault-tolerant high-speed train control system.
3. Optimal control problems and Riccati Equations for systems with unbounded controls
and partially analytic generators-applications to boundary and point control problems
4. Optimal control problems and Riccati Equations for systems with unbounded controls
and partially analytic generators-applications to boundary and point control problems
References
2012 Third International Conference on Digital Manufacturing & Automation
THANK YOU
