Download Full Issue in PDF - Academy Publisher
Download Full Issue in PDF - Academy Publisher
Download Full Issue in PDF - Academy Publisher
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
JOURNAL OF COMPUTERS, VOL. 8, NO. 6, JUNE 2013 1561<br />
axis of the rod angle to the center of rod mass, the angle<br />
of the rod from the vertical upward direction respectively.<br />
A. State Space Model of L<strong>in</strong>ear S<strong>in</strong>gle Inverted<br />
Pendulum<br />
The method of this work is based on l<strong>in</strong>ear constant<br />
state space model as:<br />
x<br />
= Ax + Bu<br />
. (1)<br />
y = Cx + Du<br />
The mathematical model of s<strong>in</strong>gle <strong>in</strong>verted pendulum can<br />
be obta<strong>in</strong>ed by mechanism analysis [1], shown <strong>in</strong> (2),<br />
where, u , is cart angular velocity. x andθ , are the same<br />
as shown <strong>in</strong> Fig 1.<br />
⎛x<br />
⎞ ⎛0 1 0 0⎞⎛x<br />
⎞ ⎛0⎞<br />
⎜<br />
x<br />
⎟ ⎜<br />
0 0 0 0<br />
⎟⎜ x<br />
⎟ ⎜<br />
1<br />
⎟<br />
⎜ ⎟<br />
<br />
= ⎜ ⎟⎜ ⎟+<br />
⎜ ⎟u<br />
⎜ θ ⎟ ⎜0 0 0 1⎟⎜θ<br />
⎟ ⎜0⎟<br />
⎜<br />
θ ⎟ ⎜ ⎟<br />
0 0 29.4 0 ⎜θ<br />
⎟ ⎜ ⎟<br />
⎝ ⎠<br />
<br />
⎝ ⎠<br />
⎝ ⎠ ⎝3⎠<br />
⎛x<br />
⎞<br />
⎜<br />
x 1 0 0 0 x<br />
⎟<br />
⎛ ⎞ ⎛ ⎞ ⎛0⎞<br />
y = ⎜ ⎟<br />
⎜ = + u<br />
θ<br />
⎟ ⎜<br />
0 0 1 0<br />
⎟⎜θ<br />
⎟ ⎜<br />
0<br />
⎟<br />
⎝ ⎠ ⎝ ⎠ ⎝ ⎠<br />
⎜ <br />
θ ⎟<br />
⎝ ⎠<br />
B. Controllability & Observability Analysis of S<strong>in</strong>gle<br />
Inverted Pendulum<br />
The controllability and observability of a system is<br />
prerequisite for analysis and controller design. Here,<br />
n−<br />
Uc = B AB A 1 B and<br />
Controllability matrix ( )<br />
n<br />
observability matrix Uo ( C CA CA −1<br />
)<br />
T<br />
(2)<br />
= were<br />
obta<strong>in</strong>ed and then rank criterion was employed to<br />
analysis its controllability and observability. It has been<br />
proved that the system as (2) has both controllability and<br />
observability.<br />
C. Stability Analysis of S<strong>in</strong>gle Inverted Pendulum<br />
The extended MVC 3 performance evolution and MPC<br />
tun<strong>in</strong>g system is based on the stabilization system. Hence,<br />
Stability analysis is necessary. The poles of the <strong>in</strong>verted<br />
pendulum as (2) are ( 5.4222 − 5.4222 0 0)<br />
, where<br />
positive real root appears. It shows that the system as (2)<br />
is <strong>in</strong>stable and this requires a stabilizer before design<strong>in</strong>g a<br />
MPC controller for the generalized controlled system [23].<br />
III. MPC PERFORMANCE EVOLUTION AND TUNING<br />
METHODE BASED ON EXTENDED MVC 3<br />
A. Introductions of LMI and Lemmas<br />
• About LMI:<br />
Assume a l<strong>in</strong>ear matrix <strong>in</strong>equality (LMI) can be stated<br />
as: F( x) = F0 + x1F1+⋅⋅⋅+ xmFm<br />
< 0 .Where, variable<br />
x constitutes a convex set, LMI can be solved us<strong>in</strong>g the<br />
method of convex optimization problem [24].<br />
1) Feasible solution of LMI:<br />
If there exists x makes, F( x ) < 0 , established, then the<br />
LMI is feasible [24].This can be expressed us<strong>in</strong>g the<br />
follow<strong>in</strong>g formulation:<br />
m<strong>in</strong>. t<br />
.<br />
s.. tF( x)<br />
< tI<br />
2) M<strong>in</strong>imization problem of LMI:<br />
The problem can be stated as a optimality problem that<br />
m<strong>in</strong>imize the largest eigenvalue, λ , of the matrix<br />
Gx ( ) under <strong>in</strong>equality constra<strong>in</strong>t H( x ) < 0 [24]:<br />
st ..<br />
Another expression is as:<br />
T<br />
m<strong>in</strong> λ<br />
G < λI<br />
.<br />
< 0<br />
( x)<br />
H ( x)<br />
T<br />
m<strong>in</strong> c x<br />
,<br />
st .. F < 0<br />
( x)<br />
where, c x is object function.<br />
• Lemmas:<br />
Consider a l<strong>in</strong>ear time-<strong>in</strong>variant state-space system:<br />
x( k + 1) = Ax( k) + Bu( k) + Ew( k)<br />
, (3)<br />
y( k) = Cx( k)<br />
where, x (k)<br />
, u ( k ) , y( k ) are state variable, manipulative<br />
variable and control output variable respectively,<br />
A , B , C and E are process model matrix, w ( k)<br />
denotes<br />
stationary, Gaussian noise with zero mean and covariance<br />
as ∑ w .<br />
State feedback controller can be expressed as:<br />
u( k) = Kx ( k)<br />
. (4)<br />
Then, the closed-loop system can be written as:<br />
x( k + 1) = ( A+ BK) x( k) + Ew ( k)<br />
. (5)<br />
Lemma1: LMI of MVC 3 [22]:<br />
For system (5), ∃ stabiliz<strong>in</strong>g K and ∑x ≥ 0 s.t.<br />
∑ , = 1... p<br />
yi i<br />
T<br />
T<br />
x w x<br />
T<br />
∑ y = ϕ<br />
i iC∑<br />
xC<br />
ϕi<br />
2<br />
∑ y < y , 1<br />
i i i = … p<br />
( A+ BK) ∑ ( A+ BK)<br />
+ E∑ E = ∑<br />
If and only if ∃ W, X > 0and Yi , = 1... ps.t.<br />
⎡<br />
T<br />
X − E∑<br />
E AX BW⎤<br />
w +<br />
⎢<br />
⎥><br />
0<br />
T<br />
⎢⎣( AX + BW ) X ⎥⎦<br />
⎡ Yi<br />
ϕiCX⎤ ⎢<br />
( ) T 0<br />
T<br />
⎥ ><br />
⎢⎣<br />
CX ϕi<br />
X ⎥⎦<br />
© 2013 ACADEMY PUBLISHER