Scribd
Upload
64 views3 pages

Outline: Linear Quadratic Regulator

This document provides an overview of optimal control and the linear quadratic regulator (LQR). It defines optimal control as selecting controls to minimize a performance measure while satisfying constraints, including the plant's state equation. The LQR specifically uses a quadratic performance measure involving a weighted sum of control effort and errors for a linear system. It describes how the LQR minimizes this performance measure and provides properties of the algebraic Riccati equation used to solve for the optimal control gains.

Uploaded by

Abdesselem Boulkroune
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
64 views3 pages

Outline: Linear Quadratic Regulator

This document provides an overview of optimal control and the linear quadratic regulator (LQR). It defines optimal control as selecting controls to minimize a performance measure while satisfying constraints, including the plant's state equation. The LQR specifically uses a quadratic performance measure involving a weighted sum of control effort and errors for a linear system. It describes how the LQR minimizes this performance measure and provides properties of the algebraic Riccati equation used to solve for the optimal control gains.

Uploaded by

Abdesselem Boulkroune
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Outline

• Optimal control.
Linear Quadratic Regulator • The linear quadratic regulator.
• The algebraic Riccati equation.
M. Sami Fadali • MATLAB commands.
Professor of Electrical Engineering
University of Nevada, Reno

1 2

Optimal Control Linear Quadratic Regulator

Optimal Control: select control that minimizes a


performance measure subject to constraints
including the state equation of the plant.
Choice of Performance Measure: Depends on the such that
application and design objectives. = state-weighting matrix
Ex. For the fastest possible response minimize time. = positive semidefinite matrix (eigenvalues  0)
Linear Quadratic Regulator: Quadratic
= input-weighting matrix
performance measure including a weighted sum of
the control effort and errors with a linear system. = positive definite matrix (eigenvalues > 0)

3 4
Simplification Interpretation of LQR
Minimize a weighted sum of the area under the
square error curve and the area under the
control effort curve
1

0.9

0.8

0.7

0.6

• Most common choice of weighting matrices.

x(t)
0.5

0.4

0.3

0.2

0.1

0
0 1 2 3 4 5 6
5 t 6

Solution Algebraic Riccati Equation


/
• If (A, B) is stabilizable and with ∗
/ /
is detectable, the unique
solution that stabilizes the closed-loop system is
∗
where • Nonlinear matrix equation.
∗ • Numerical solution.
• MATLAB
>> [P, ecl, K] = care(A, B, Q,R) % ecl=eig(A-B K)
• Algebraic Riccati Equation

7 8
Properties of LQR Solution
Properties of LQR Solution
• LQR minimization: analogous to minimizing a
/ quadratic function of and subject to constraints
• If (A, B) is stabilizable and is
but minimum is at functions and .
detectable, a unique solution that stabilizes the
closed-loop system exists. • Solution affected by the relative magnitudes of the
weights not their exact values.
/ /
• For , 120

/ Minimizing 100
f
and the pair is
with scalar: 80

observable by the rank test. at


60

f
40

• If (A, B) is controllable then it is stabilizable. at 20

0
-5 -4 -3 -2 -1 0 1 2 3 4 5
9 10
x*

MATLAB
Properties of LQR Solution (cont.) >> p=ss( [1,2;3,4],[0;1],[1,1],0)
>> Q = eye(2);R = 0.2;
• Weights reflect the relative importance of >>[K, P, e] = lqr(p,Q,R) % e = eig(A-B*K)
the and terms. If is large, say, then K=
solutions with larger are excluded since 9.5381 11.6911
they yield larger performance measures. P=
• LQR give a compromise between error and 2.8747 1.9076
1.9076 2.3382
control effort minimization. Minimizing
e=
error alone gives large (impulse) inputs.
-0.9357
-5.7554
11 12

You might also like