Scilab 5.4.1
- Scilab help
- CACSD (Computer Aided Control Systems Design)
- Formal representations and conversions
- Plot and display
- abinv
- arhnk
- arl2
- arma
- arma2p
- arma2ss
- armac
- armax
- armax1
- arsimul
- augment
- balreal
- bilin
- bstap
- cainv
- calfrq
- canon
- ccontrg
- cls2dls
- colinout
- colregul
- cont_mat
- contr
- contrss
- copfac
- csim
- ctr_gram
- damp
- dcf
- ddp
- dhinf
- dhnorm
- dscr
- dsimul
- dt_ility
- dtsi
- equil
- equil1
- feedback
- findABCD
- findAC
- findBD
- findBDK
- findR
- findx0BD
- flts
- fourplan
- freq
- freson
- fspec
- fspecg
- fstabst
- g_margin
- gamitg
- gcare
- gfare
- gfrancis
- gtild
- h2norm
- h_cl
- h_inf
- h_inf_st
- h_norm
- hankelsv
- hinf
- imrep2ss
- inistate
- invsyslin
- kpure
- krac2
- lcf
- leqr
- lft
- lin
- linf
- linfn
- linmeq
- lqe
- lqg
- lqg2stan
- lqg_ltr
- lqr
- ltitr
- macglov
- minreal
- minss
- mucomp
- narsimul
- nehari
- noisegen
- nyquistfrequencybounds
- obs_gram
- obscont
- observer
- obsv_mat
- obsvss
- p_margin
- parrot
- pfss
- phasemag
- plzr
- pol2des
- ppol
- prbs_a
- projsl
- reglin
- repfreq
- ric_desc
- ricc
- riccati
- routh_t
- rowinout
- rowregul
- rtitr
- sensi
- sident
- sorder
- specfact
- ssprint
- st_ility
- stabil
- sysfact
- syslin
- syssize
- time_id
- trzeros
- ui_observer
- unobs
- zeropen
Please note that the recommended version of Scilab is 2025.0.0. This page might be outdated.
See the recommended documentation of this function
lqe
linear quadratic estimator (Kalman Filter)
Calling Sequence
[K,X]=lqe(P21)
Arguments
- P21
syslin
list- K, X
real matrices
Description
lqe
returns the Kalman gain for the filtering
problem in continuous or discrete time.
P21
is a syslin
list representing the system P21=[A,B1,C2,D21] P21=syslin('c',A,B1,C2,D21) or P21=syslin('d',A,B1,C2,D21)
The input to P21
is a white noise with variance:
[B1 ] [Q S] BigV=[ ] [ B1' D21'] = [ ] [D21] [S' R]
X
is the solution of the stabilizing Riccati
equation and A+K*C2
is stable.
In continuous time:
K=-(X*C2'+S)*inv(R)
In discrete time:
X=A*X*A'-(A*X*C2'+B1*D21')*pinv(C2*X*C2'+D21*D21')*(C2*X*A'+D21*B1')+B1*B1'
K=-(A*X*C2'+B1*D21')*pinv(C2*X*C2'+D21*D21')
xhat(t+1)= E(x(t+1)| y(0),...,y(t))
(one-step predicted x
)
satisfies the recursion:
xhat(t+1)=(A+K*C2)*xhat(t) - K*y(t).
Examples
//Assume the equations //. //x = Ax + Ge //y = Cx + v //with //E ee' = Q_e, Evv' = R, Eev' = N // //This is equivalent to //. //x = Ax + B1 w //y = C2x + D21 w //with E { [Ge ] [Ge v]' } = E { [B1w ] [B1w D21w]' } = bigR = // [ v ] [D21w] // //[B1*B1' B1*D21'; // D21*B1' D21*D21'] //= //[G*Q_e*G' G*N; // N*G' R] //To find (B1,D21) given (G,Q_e,R,N) form bigR =[G*Q_e*G' G*N;N'*G' R]. //Then [W,Wt]=fullrf(bigR); B1=W(1:size(G,1),:); //D21=W(($+1-size(C2,1)):$,:) // //P21=syslin('c',A,B1,C2,D21); //[K,X]=lqe(P21); //Example: nx=5;ne=2;ny=3; A=-diag(1:nx);G=ones(nx,ne); C=ones(ny,nx); Q_e(ne,ne)=1; R=diag(1:ny); N=zeros(ne,ny); bigR =[G*Q_e*G' G*N;N'*G' R]; [W,Wt]=fullrf(bigR);B1=W(1:size(G,1),:); D21=W(($+1-size(C,1)):$,:); C2=C; P21=syslin('c',A,B1,C2,D21); [K,X]=lqe(P21); //Riccati check: S=G*N;Q=B1*B1'; (A-S*inv(R)*C2)*X+X*(A-S*inv(R)*C2)'-X*C2'*inv(R)*C2*X+Q-S*inv(R)*S' //Stability check: spec(A+K*C)
Report an issue | ||
<< linmeq | CACSD (Computer Aided Control Systems Design) | lqg >> |