SlideShare une entreprise Scribd logo
1  sur  22
Télécharger pour lire hors ligne
Course Calendar
Class DATE Contents
1 Sep. 26 Course information & Course overview
2 Oct. 4 Bayes Estimation
3 〃 11 Classical Bayes Estimation - Kalman Filter -
4 〃 18 Simulation-based Bayesian Methods
5 〃 25 Modern Bayesian Estimation Particle Filter
6 Nov. 1 HMM(Hidden Markov Model)
Nov. 8 No Class
7 〃 15 Supervised Learning
8 〃 29 Bayesian Decision
9 Dec. 6 PCA(Principal Component Analysis)
10 〃 13 ICA(Independent Component Analysis)
11 〃 20 Applications of PCA and ICA
12 〃 27 Clustering, k-means et al.
13 Jan. 17 Other Topics 1 Kernel machine.
14 〃 22(Tue) Other Topics 2
Lecture Plan
Classical Bayesian Estimation
Kalman Filter
1. Introduction
What is Kalman filter
Dynamical system representation
2. Discrete-time Gauss-Markov Model
Models, State and covariance transition
3. Bayesian Approach to Kalman Filter
State estimation problem, Deviation
1. Introduction
Kalman filter is an optimal recursive data processing algorithm
It processes available measurement data to estimate the current value
of the variables of interest, by using
1) dynamics of system and measurement models,
2) statistical knowledge of system and measurement noises,
3) available information about initial conditions of state variables
Recursive: Not require all previous data by keeping in storage,
repeatedly process every time a new measurement is acquired
Real-time processing
3
system
measurement
device
System error noise : w(t)
Measurements
y(t)
Optimal estimate
of state variables
Measurement
error noise : v(t)
unknown state
variables : x(t)
Deterministic
Input : u(t)
(Control)
Figure 1 Role of Kalman filter
Kalman
filter
4
 ˆ tx
5
         1 2 3 4
T
x t x t x t x t x t   
Linearized continuous-time state-space model for motion in the z-direction
     
dispacement angle
( )z t z t t t 
M : Mass of carriage,
m : mass of pendulum
l : length of pendulum
g : gravitational constant
 
 
     
1 1
23 43
232 2 2
3 3
2 4
434 4 4
1
2
3
4
0 1 0 0 0
,0 0 0
0 0 0 1 0 1 1
,
0 0 0
1 0 0 0 ( )
x x M m gmg
a aax x bd M Mlu t
x xdt
b b
ax x b M Ml
x
x
y t z t
x
x
       
         
       
      
        
      
 
 
  
 
 
 
2)Dynamical state-space representation
3)Measurement equation
1)State space variables (Vector)
Continuous-time state space model
Deterministic linear time-invariant state space model is given by
 
   
   
     
In time-varing systems, all elements of matrices , , are function
(State Dynamic Model)
(Measurement Model)
where : n-vector, : m-vector, : -vector,
A B C
x t
Ax t Bu t
dt
y t Cx t
x u y l  
 

of .t
6
Discrete-time state space model
     
   
1 1 (State Dynamic Model)
(Measurement Model)
where is integers (discrete-time index)
x t Ax t Bu t
y t Cx t
t
   

2.1 Random signals(noises) are applied to a discrete-time state-space
system with random initial conditions. The input u(t) is deterministic
and the system noise w(t) to be zero-mean, white*1), random Gaussian
First-order Markov Process: State at t depends only on the previous
state at t-1. (State-space representation)
       
    
             
         
           
:mean value of 0
1 1 1
where 1 0, 1 and
0 0 0 , 0 0
,
where . :
ww
T
xx
x x
x t Ax t Bu t Ww t
w t R t
x Pr x x P
x t Pr x t x t P t
P t R t Cov x t E x t x t
     
 


  
2. Discrete-time Gauss-Markov Models
*1 white signal :=signal value is not correlated in time, thus this implies signal has equal
power at all frequencies. 7
     
  
:
where, 0, vv
y t Cx t v t
v N R t
 
Measurement Model
The model is shown as the following block diagram.
 u t
 w t
 1x t   x t
 v t
 y tB 
A
C
W
1
z I
1
: delay operatorz
Figure 3 System diagram
8
9
   
 
   
 
 
     
     
1
/2 1/2
1
1 1 1
, exp
22
is D-dimensional random vector
:
: :
: Determinant of
Linear Transformation
, ,
T
xx xxD
xx
T
D
xx
xx xx
T
yy xx
Pr x x R x x R x x
R
x x x
x E x
R Cov x E x x x x
R R
y Ax
Pr y y R A x AR A

 
    
 

  



Appendix:
Multivariable Gaussian Density Distribution
Since the Gauss-Markov model is characterized by a Gaussian
distribution, it is completely specified statistically by its mean and
variance.
     
   
1 1x t Ax t Bu t
y t Cx t
   

Mean
     
     
(State Variance)
1 1
(Measurement)
T T
ww
T
yy vv
P t AP t A WR t W
R t CP t C R t
   
 
Variance
2.2 Mean and Variance Transitions
10
3. Bayesian Approach to Kalman filter
[Problem]
/State-space system model
/Gauss-Markov model of the state transition mechanism.
/Bayesian optimal estimation (MAP solution)
Extracting the unobserved or hidden dynamic state variables x(t)
from noisy measurement data {y(τ): τ=0,1, …, t}.
We see that the dynamic state variable x(t) at time t is obtained
through the transition probability based on the previous state
(Markovian property), and the knowledge of the underlying
conditional probability.
11
In a random process the next state depends only on the current state and not on the
sequence of events that preceded it. This specific kind of "memorylessness" is called the
Markov property.
A stochastic process has the Markov property if the conditional probability distribution
of future states of the process depends only upon the present state, not on the
sequence of events that preceded it. A process with this property is called a Markov
process. Quote from Wikipedia
Bayesian Approach – Prediction + Correction Scheme-
        
    
  
       
  
    
 
( )
1
1
: : 0 , 1 , , 1 ,
Bayesian Estimate:
ˆ arg
+ approach for computing
0 , 1 , , 1 ,
t
t
x t
t
t
t
Data Y y y y t y t
x t Max Pr x t Y
Pr x t Y
y y y t y t
Y
Pr y t x t
Pr x t Y
Pr y t


 


Prediction Correction
 
  
1
t
t
Pr x t Y
Y

12
Prediction
Phase
Correction
Phase
13
  
       
  
 
     
    
1
1
(1)
, ( ) ( 1), 1
? , ?
ˆ 1 ,
t
t
t
vv
ee
Pr y t x t Pr x t Y
Pr x t Y
Pr y t Y
N Cx t R t N x t t P t t
N y t t R t




 


           
       
       
1
1
From the Gauss-Markov model, we have
,
ˆ 1 , 1
ˆ 1 ,
vv
t
t ee
Pr y t x t C t x t R t
Pr x t Y x t t P t t
Pr y t Y y t t R t



  
 
     
     
   
1
ˆ ˆPrediction: 1 1 1
ˆPrediction error: 1 1
1 1 1
tx t t E x t Y Ax t t
x t t x t x t t
Ax t t Ww t

      
   
    
     
   
1 1 1
1 1 1
T
T
ww
P t t E x t t x t t
AP t t A R t
     
    
Using the process model, we have (ignore the input term Bu(t-1))
State prediction error covariance
14
              
           
         
1
1
1
Substituting these into Eq (1)gives
1
exp
2
1
ˆ ˆexp 1 1 1
2
1
ˆ ˆexp 1 1
2
T
vv
T
T
ee
p x t Y t y t Cx t R y t Cx t
x t x t t P t t x t x t t
y t y t t R y t y t t
 


 
      
 
        
 
      
     
     
     
   
Rewrite the terms above as follows:
:
ˆ1 1
ˆ: 1
ˆ 1
measurement noise v t y t Cx t
state estimation error x t t x t x t t
innovation e t y t y t t
y t Cx t t
 
   
  
  
15
Optimal Bayesian estimator is the processor that maximizes the
posterior above. Taking natural logarithms of both sides of Eq.(2),
           
           
   
1
1 1
1
ln Pr ( ) ln Pr ln Pr
1 1
1 1 1
2 2
1
R
2
t t t
T T
vv
T
ee
ln x t Y y t x t x t Y x t Y
LnK v t R t v t x t t p t t x t t
e t e t

 

  
     

The MAP estimate is obtained by differentiating eq.(3), setting it to
zero, and solving the equation. That is,
        
     
   
1
1
1
1
exp
2
1
exp 1 1 1
2
1
exp (2)
2
T
vv
T
T
ee
p x t Y t v t R v t
x t t P t t x t t
e t R e t
 


 
    
 
      
 
   
16
(3)
  
  
         
ˆ
1 1
ln Pr 0
ln Pr
1 1 0
MAP
x t
x X
x t
T
vv
x t Y
x t Y
C R t y t Cx t P t t x t t

 
 

       
     
       
       
       
1
1 1
1 1
1
1 1
1 1
1
ˆ1 1
then,
ˆˆ 1
ˆ1 1 (4)
T
vv
T
vv
T
MAP vv
T
vv
C R t C P t t x t
P t t x t t C R t y t
x t t X t C R t C P t t
P t t x t t C R t y t

 
 

 
 
   
   
     
     
This representation can be simplified by using a matrix inversion
lemma
   
1 11 1 1 1T T T
Matrix Inversion Lemma
A BD A A B I D A B D A
    
   
17
     
      
       
             
1
1 1
1 1
1 1
1 1
It is possible to show 1
thus, 1
1
Substituting above into eq. (4),
ˆˆ ˆ1 1
ˆ=
T
vv
T
vv
T
vv
T
map vv
P t t C R t C P t t
P t t P t t C R t C I
P t t P t t I P t t C R t C
x t t X t P t t P t t x t t C R t y t
x t

 
 
 
 
    
  
  
      
             
        
     
     
1 1
1
ˆ1 1
ˆ ˆ= 1 1
ˆ= 1
where
is called .
T T
vv vv
T
vv
t P t t C R t Cx t t P t t C R t y t
x t t K t y t Cx t t
x t t K t e t
K t P t t C R t
 

   
   
 

 
 
Kalman gain
18
     
We can derive the following covariance-update relation.
1P t t I K t C P t t    
19
   
     
     
         
   
ˆ 0 0 , 0 0
ˆ ˆ1 1 1 1
1 1 1 1
ˆ ˆ1 1
1
ww
T
ee vv
x P
x t t Ax t t Bu t
P t t AP t t A R t
e t y t y t t y t Cx t t
R t CP t t C R
     
     
     
  
Initial Conditions :
Prediction :
Innovation:
Kalman Filter Algorithm
 
     
       
     
1
1
ˆ ˆ 1
1
T
ee
t
K t P t t C R t
x t t x t t K t e t
P t t I K t C P t t

 
  
     
Gain:
Update :
1t 
t y t
 1y t 
   ˆ 1 1 , 1 1x t t p t t   
 
 
ˆ 1 ,
1
x t t
p t t


   ˆ ,x t t p t t
Measurement
Measurement
Correction process
Correction process
Prediction by Model
Prediction by Model
Figure 4
Understanding Kalman filter operation and its Gain
( :Kalman gain)new old new
ˆ ˆX X KE K 
Prediction:
Use state space model
Correction:
Use measurement
     
 
 
 
 
is small
is large
small
Reliable model is small
large
large
Reliable measurement is large
small
T
vv
new old
new new
vv
vv
K t P t t C R t
ˆ ˆK X X
ˆK X KE
P t t
K
R t
P t t
K
R t




 
 
1
22
References:
[1] J. Candy, “Model-based Signal Processing”, John Wiley/IEEE Press, 2006
[2] J. Candy, “ Bayesian Signal Processing Classical, Modern, and Particle Filtering Methods”,
John Wiley/IEEE Press, 2009
[3] S. Maybeck, “Stochastic Models, Estimation, and Control”, Vol. 1, Academic Press, 1979,
1999
[4] R. Brown, “Introduction to random signal analysis and Kalman filtering”, John Weily, 1983

Contenu connexe

Tendances

Dynamical systems
Dynamical systemsDynamical systems
Dynamical systemsSpringer
 
The feedback-control-for-distributed-systems
The feedback-control-for-distributed-systemsThe feedback-control-for-distributed-systems
The feedback-control-for-distributed-systemsCemal Ardil
 
state space representation,State Space Model Controllability and Observabilit...
state space representation,State Space Model Controllability and Observabilit...state space representation,State Space Model Controllability and Observabilit...
state space representation,State Space Model Controllability and Observabilit...Waqas Afzal
 
The computational limit_to_quantum_determinism_and_the_black_hole_information...
The computational limit_to_quantum_determinism_and_the_black_hole_information...The computational limit_to_quantum_determinism_and_the_black_hole_information...
The computational limit_to_quantum_determinism_and_the_black_hole_information...Sérgio Sacani
 
Design and Analysis of a Control System Using Root Locus and Frequency Respon...
Design and Analysis of a Control System Using Root Locus and Frequency Respon...Design and Analysis of a Control System Using Root Locus and Frequency Respon...
Design and Analysis of a Control System Using Root Locus and Frequency Respon...Umair Shahzad
 
Controllability of Linear Dynamical System
Controllability of  Linear Dynamical SystemControllability of  Linear Dynamical System
Controllability of Linear Dynamical SystemPurnima Pandit
 
Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...
  Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...  Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...
Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...Belinda Marchand
 
Introduction to mathematical control theory - Dr. Purnima Pandit
Introduction to mathematical control theory - Dr. Purnima PanditIntroduction to mathematical control theory - Dr. Purnima Pandit
Introduction to mathematical control theory - Dr. Purnima PanditPurnima Pandit
 
Transfer fn mech. systm 2
Transfer fn mech. systm 2Transfer fn mech. systm 2
Transfer fn mech. systm 2Syed Saeed
 
Experimental realisation of Shor's quantum factoring algorithm using qubit r...
 Experimental realisation of Shor's quantum factoring algorithm using qubit r... Experimental realisation of Shor's quantum factoring algorithm using qubit r...
Experimental realisation of Shor's quantum factoring algorithm using qubit r...XequeMateShannon
 
Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...
Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...
Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...Amr E. Mohamed
 
Sistema de control discreto
Sistema de control discretoSistema de control discreto
Sistema de control discretoricardomadrid19
 
Modern Control - Lec07 - State Space Modeling of LTI Systems
Modern Control - Lec07 - State Space Modeling of LTI SystemsModern Control - Lec07 - State Space Modeling of LTI Systems
Modern Control - Lec07 - State Space Modeling of LTI SystemsAmr E. Mohamed
 
Introduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applicationsIntroduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applicationsSpringer
 
Introduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applicationsIntroduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applicationsSpringer
 
discrete time signals and systems
 discrete time signals and systems  discrete time signals and systems
discrete time signals and systems Zlatan Ahmadovic
 
Time domain definition 6
Time domain definition 6Time domain definition 6
Time domain definition 6Syed Saeed
 
Introduction to the Keldysh non-equlibrium Green's function technique
Introduction to the Keldysh non-equlibrium Green's function techniqueIntroduction to the Keldysh non-equlibrium Green's function technique
Introduction to the Keldysh non-equlibrium Green's function techniqueInon Sharony
 
Sliding Mode Observers
Sliding Mode ObserversSliding Mode Observers
Sliding Mode ObserversSolo Hermelin
 

Tendances (20)

Dynamical systems
Dynamical systemsDynamical systems
Dynamical systems
 
The feedback-control-for-distributed-systems
The feedback-control-for-distributed-systemsThe feedback-control-for-distributed-systems
The feedback-control-for-distributed-systems
 
state space representation,State Space Model Controllability and Observabilit...
state space representation,State Space Model Controllability and Observabilit...state space representation,State Space Model Controllability and Observabilit...
state space representation,State Space Model Controllability and Observabilit...
 
The computational limit_to_quantum_determinism_and_the_black_hole_information...
The computational limit_to_quantum_determinism_and_the_black_hole_information...The computational limit_to_quantum_determinism_and_the_black_hole_information...
The computational limit_to_quantum_determinism_and_the_black_hole_information...
 
Design and Analysis of a Control System Using Root Locus and Frequency Respon...
Design and Analysis of a Control System Using Root Locus and Frequency Respon...Design and Analysis of a Control System Using Root Locus and Frequency Respon...
Design and Analysis of a Control System Using Root Locus and Frequency Respon...
 
Controllability of Linear Dynamical System
Controllability of  Linear Dynamical SystemControllability of  Linear Dynamical System
Controllability of Linear Dynamical System
 
Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...
  Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...  Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...
Discrete Nonlinear Optimal Control of S/C Formations Near The L1 and L2 poi...
 
Introduction to mathematical control theory - Dr. Purnima Pandit
Introduction to mathematical control theory - Dr. Purnima PanditIntroduction to mathematical control theory - Dr. Purnima Pandit
Introduction to mathematical control theory - Dr. Purnima Pandit
 
Transfer fn mech. systm 2
Transfer fn mech. systm 2Transfer fn mech. systm 2
Transfer fn mech. systm 2
 
Experimental realisation of Shor's quantum factoring algorithm using qubit r...
 Experimental realisation of Shor's quantum factoring algorithm using qubit r... Experimental realisation of Shor's quantum factoring algorithm using qubit r...
Experimental realisation of Shor's quantum factoring algorithm using qubit r...
 
Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...
Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...
Modern Control - Lec 04 - Analysis and Design of Control Systems using Root L...
 
Sistema de control discreto
Sistema de control discretoSistema de control discreto
Sistema de control discreto
 
Modern Control - Lec07 - State Space Modeling of LTI Systems
Modern Control - Lec07 - State Space Modeling of LTI SystemsModern Control - Lec07 - State Space Modeling of LTI Systems
Modern Control - Lec07 - State Space Modeling of LTI Systems
 
Modern control 2
Modern control 2Modern control 2
Modern control 2
 
Introduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applicationsIntroduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applications
 
Introduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applicationsIntroduction to queueing systems with telecommunication applications
Introduction to queueing systems with telecommunication applications
 
discrete time signals and systems
 discrete time signals and systems  discrete time signals and systems
discrete time signals and systems
 
Time domain definition 6
Time domain definition 6Time domain definition 6
Time domain definition 6
 
Introduction to the Keldysh non-equlibrium Green's function technique
Introduction to the Keldysh non-equlibrium Green's function techniqueIntroduction to the Keldysh non-equlibrium Green's function technique
Introduction to the Keldysh non-equlibrium Green's function technique
 
Sliding Mode Observers
Sliding Mode ObserversSliding Mode Observers
Sliding Mode Observers
 

Similaire à 2012 mdsp pr03 kalman filter

2012 mdsp pr05 particle filter
2012 mdsp pr05 particle filter2012 mdsp pr05 particle filter
2012 mdsp pr05 particle filternozomuhamada
 
2012 mdsp pr06  hmm
2012 mdsp pr06  hmm2012 mdsp pr06  hmm
2012 mdsp pr06  hmmnozomuhamada
 
Numerical Methods
Numerical MethodsNumerical Methods
Numerical MethodsTeja Ande
 
An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...
An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...
An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...ijtsrd
 
A lambda calculus for density matrices with classical and probabilistic controls
A lambda calculus for density matrices with classical and probabilistic controlsA lambda calculus for density matrices with classical and probabilistic controls
A lambda calculus for density matrices with classical and probabilistic controlsAlejandro Díaz-Caro
 
Design of Linear Control Systems(0).ppt
Design of Linear Control Systems(0).pptDesign of Linear Control Systems(0).ppt
Design of Linear Control Systems(0).pptkhinmuyaraye
 
Random Matrix Theory and Machine Learning - Part 3
Random Matrix Theory and Machine Learning - Part 3Random Matrix Theory and Machine Learning - Part 3
Random Matrix Theory and Machine Learning - Part 3Fabian Pedregosa
 
On problem-of-parameters-identification-of-dynamic-object
On problem-of-parameters-identification-of-dynamic-objectOn problem-of-parameters-identification-of-dynamic-object
On problem-of-parameters-identification-of-dynamic-objectCemal Ardil
 
Identification of the Mathematical Models of Complex Relaxation Processes in ...
Identification of the Mathematical Models of Complex Relaxation Processes in ...Identification of the Mathematical Models of Complex Relaxation Processes in ...
Identification of the Mathematical Models of Complex Relaxation Processes in ...Vladimir Bakhrushin
 
Vu_HPSC2012_02.pptx
Vu_HPSC2012_02.pptxVu_HPSC2012_02.pptx
Vu_HPSC2012_02.pptxQucngV
 
BEC- 26 control systems_unit-II
BEC- 26 control systems_unit-IIBEC- 26 control systems_unit-II
BEC- 26 control systems_unit-IIShadab Siddiqui
 
Time series Modelling Basics
Time series Modelling BasicsTime series Modelling Basics
Time series Modelling BasicsAshutosh Kumar
 
Maneuvering target track prediction model
Maneuvering target track prediction modelManeuvering target track prediction model
Maneuvering target track prediction modelIJCI JOURNAL
 
Control digital: Tema 3. Análisis de sistemas discretos
Control digital: Tema 3. Análisis de sistemas discretos Control digital: Tema 3. Análisis de sistemas discretos
Control digital: Tema 3. Análisis de sistemas discretos SANTIAGO PABLO ALBERTO
 

Similaire à 2012 mdsp pr03 kalman filter (20)

2012 mdsp pr05 particle filter
2012 mdsp pr05 particle filter2012 mdsp pr05 particle filter
2012 mdsp pr05 particle filter
 
2012 mdsp pr06  hmm
2012 mdsp pr06  hmm2012 mdsp pr06  hmm
2012 mdsp pr06  hmm
 
Numerical Methods
Numerical MethodsNumerical Methods
Numerical Methods
 
An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...
An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...
An Exponential Observer Design for a Class of Chaotic Systems with Exponentia...
 
A lambda calculus for density matrices with classical and probabilistic controls
A lambda calculus for density matrices with classical and probabilistic controlsA lambda calculus for density matrices with classical and probabilistic controls
A lambda calculus for density matrices with classical and probabilistic controls
 
Design of Linear Control Systems(0).ppt
Design of Linear Control Systems(0).pptDesign of Linear Control Systems(0).ppt
Design of Linear Control Systems(0).ppt
 
Random Matrix Theory and Machine Learning - Part 3
Random Matrix Theory and Machine Learning - Part 3Random Matrix Theory and Machine Learning - Part 3
Random Matrix Theory and Machine Learning - Part 3
 
On problem-of-parameters-identification-of-dynamic-object
On problem-of-parameters-identification-of-dynamic-objectOn problem-of-parameters-identification-of-dynamic-object
On problem-of-parameters-identification-of-dynamic-object
 
Identification of the Mathematical Models of Complex Relaxation Processes in ...
Identification of the Mathematical Models of Complex Relaxation Processes in ...Identification of the Mathematical Models of Complex Relaxation Processes in ...
Identification of the Mathematical Models of Complex Relaxation Processes in ...
 
PhDretreat2011
PhDretreat2011PhDretreat2011
PhDretreat2011
 
Vu_HPSC2012_02.pptx
Vu_HPSC2012_02.pptxVu_HPSC2012_02.pptx
Vu_HPSC2012_02.pptx
 
BEC- 26 control systems_unit-II
BEC- 26 control systems_unit-IIBEC- 26 control systems_unit-II
BEC- 26 control systems_unit-II
 
14599404.ppt
14599404.ppt14599404.ppt
14599404.ppt
 
Time series Modelling Basics
Time series Modelling BasicsTime series Modelling Basics
Time series Modelling Basics
 
ACS 22LIE12 lab Manul.docx
ACS 22LIE12 lab Manul.docxACS 22LIE12 lab Manul.docx
ACS 22LIE12 lab Manul.docx
 
legendre.pptx
legendre.pptxlegendre.pptx
legendre.pptx
 
QMC Program: Trends and Advances in Monte Carlo Sampling Algorithms Workshop,...
QMC Program: Trends and Advances in Monte Carlo Sampling Algorithms Workshop,...QMC Program: Trends and Advances in Monte Carlo Sampling Algorithms Workshop,...
QMC Program: Trends and Advances in Monte Carlo Sampling Algorithms Workshop,...
 
pRO
pROpRO
pRO
 
Maneuvering target track prediction model
Maneuvering target track prediction modelManeuvering target track prediction model
Maneuvering target track prediction model
 
Control digital: Tema 3. Análisis de sistemas discretos
Control digital: Tema 3. Análisis de sistemas discretos Control digital: Tema 3. Análisis de sistemas discretos
Control digital: Tema 3. Análisis de sistemas discretos
 

Plus de nozomuhamada

2012 mdsp pr13 support vector machine
2012 mdsp pr13 support vector machine2012 mdsp pr13 support vector machine
2012 mdsp pr13 support vector machinenozomuhamada
 
2012 mdsp pr12 k means mixture of gaussian
2012 mdsp pr12 k means mixture of gaussian2012 mdsp pr12 k means mixture of gaussian
2012 mdsp pr12 k means mixture of gaussiannozomuhamada
 
2012 mdsp pr11 ica part 2 face recognition
2012 mdsp pr11 ica part 2 face recognition2012 mdsp pr11 ica part 2 face recognition
2012 mdsp pr11 ica part 2 face recognitionnozomuhamada
 
2012 mdsp pr10 ica
2012 mdsp pr10 ica2012 mdsp pr10 ica
2012 mdsp pr10 icanozomuhamada
 
2012 mdsp pr09 pca lda
2012 mdsp pr09 pca lda2012 mdsp pr09 pca lda
2012 mdsp pr09 pca ldanozomuhamada
 
2012 mdsp pr08 nonparametric approach
2012 mdsp pr08 nonparametric approach2012 mdsp pr08 nonparametric approach
2012 mdsp pr08 nonparametric approachnozomuhamada
 
2012 mdsp pr07 bayes decision
2012 mdsp pr07 bayes decision2012 mdsp pr07 bayes decision
2012 mdsp pr07 bayes decisionnozomuhamada
 
2012 mdsp pr04 monte carlo
2012 mdsp pr04 monte carlo2012 mdsp pr04 monte carlo
2012 mdsp pr04 monte carlonozomuhamada
 
2012 mdsp pr02 1004
2012 mdsp pr02 10042012 mdsp pr02 1004
2012 mdsp pr02 1004nozomuhamada
 
2012 mdsp pr01 introduction 0921
2012 mdsp pr01 introduction 09212012 mdsp pr01 introduction 0921
2012 mdsp pr01 introduction 0921nozomuhamada
 
招待講演(鶴岡)
招待講演(鶴岡)招待講演(鶴岡)
招待講演(鶴岡)nozomuhamada
 

Plus de nozomuhamada (13)

2012 mdsp pr13 support vector machine
2012 mdsp pr13 support vector machine2012 mdsp pr13 support vector machine
2012 mdsp pr13 support vector machine
 
2012 mdsp pr12 k means mixture of gaussian
2012 mdsp pr12 k means mixture of gaussian2012 mdsp pr12 k means mixture of gaussian
2012 mdsp pr12 k means mixture of gaussian
 
2012 mdsp pr11 ica part 2 face recognition
2012 mdsp pr11 ica part 2 face recognition2012 mdsp pr11 ica part 2 face recognition
2012 mdsp pr11 ica part 2 face recognition
 
2012 mdsp pr10 ica
2012 mdsp pr10 ica2012 mdsp pr10 ica
2012 mdsp pr10 ica
 
2012 mdsp pr09 pca lda
2012 mdsp pr09 pca lda2012 mdsp pr09 pca lda
2012 mdsp pr09 pca lda
 
2012 mdsp pr08 nonparametric approach
2012 mdsp pr08 nonparametric approach2012 mdsp pr08 nonparametric approach
2012 mdsp pr08 nonparametric approach
 
2012 mdsp pr07 bayes decision
2012 mdsp pr07 bayes decision2012 mdsp pr07 bayes decision
2012 mdsp pr07 bayes decision
 
2012 mdsp pr04 monte carlo
2012 mdsp pr04 monte carlo2012 mdsp pr04 monte carlo
2012 mdsp pr04 monte carlo
 
2012 mdsp pr02 1004
2012 mdsp pr02 10042012 mdsp pr02 1004
2012 mdsp pr02 1004
 
2012 mdsp pr01 introduction 0921
2012 mdsp pr01 introduction 09212012 mdsp pr01 introduction 0921
2012 mdsp pr01 introduction 0921
 
Ieice中国地区
Ieice中国地区Ieice中国地区
Ieice中国地区
 
招待講演(鶴岡)
招待講演(鶴岡)招待講演(鶴岡)
招待講演(鶴岡)
 
最終講義
最終講義最終講義
最終講義
 

Dernier

Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...Neo4j
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘RTylerCroy
 
GenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdfGenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdflior mazor
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessPixlogix Infotech
 
Developing An App To Navigate The Roads of Brazil
Developing An App To Navigate The Roads of BrazilDeveloping An App To Navigate The Roads of Brazil
Developing An App To Navigate The Roads of BrazilV3cube
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024The Digital Insurer
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CVKhem
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsMaria Levchenko
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...Martijn de Jong
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsJoaquim Jorge
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024Rafal Los
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024The Digital Insurer
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerThousandEyes
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfsudhanshuwaghmare1
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdfhans926745
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?Antenna Manufacturer Coco
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processorsdebabhi2
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAndrey Devyatkin
 

Dernier (20)

Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
GenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdfGenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdf
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your Business
 
Developing An App To Navigate The Roads of Brazil
Developing An App To Navigate The Roads of BrazilDeveloping An App To Navigate The Roads of Brazil
Developing An App To Navigate The Roads of Brazil
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CV
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdf
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of Terraform
 

2012 mdsp pr03 kalman filter

  • 1. Course Calendar Class DATE Contents 1 Sep. 26 Course information & Course overview 2 Oct. 4 Bayes Estimation 3 〃 11 Classical Bayes Estimation - Kalman Filter - 4 〃 18 Simulation-based Bayesian Methods 5 〃 25 Modern Bayesian Estimation Particle Filter 6 Nov. 1 HMM(Hidden Markov Model) Nov. 8 No Class 7 〃 15 Supervised Learning 8 〃 29 Bayesian Decision 9 Dec. 6 PCA(Principal Component Analysis) 10 〃 13 ICA(Independent Component Analysis) 11 〃 20 Applications of PCA and ICA 12 〃 27 Clustering, k-means et al. 13 Jan. 17 Other Topics 1 Kernel machine. 14 〃 22(Tue) Other Topics 2
  • 2. Lecture Plan Classical Bayesian Estimation Kalman Filter 1. Introduction What is Kalman filter Dynamical system representation 2. Discrete-time Gauss-Markov Model Models, State and covariance transition 3. Bayesian Approach to Kalman Filter State estimation problem, Deviation
  • 3. 1. Introduction Kalman filter is an optimal recursive data processing algorithm It processes available measurement data to estimate the current value of the variables of interest, by using 1) dynamics of system and measurement models, 2) statistical knowledge of system and measurement noises, 3) available information about initial conditions of state variables Recursive: Not require all previous data by keeping in storage, repeatedly process every time a new measurement is acquired Real-time processing 3
  • 4. system measurement device System error noise : w(t) Measurements y(t) Optimal estimate of state variables Measurement error noise : v(t) unknown state variables : x(t) Deterministic Input : u(t) (Control) Figure 1 Role of Kalman filter Kalman filter 4  ˆ tx
  • 5. 5          1 2 3 4 T x t x t x t x t x t    Linearized continuous-time state-space model for motion in the z-direction       dispacement angle ( )z t z t t t  M : Mass of carriage, m : mass of pendulum l : length of pendulum g : gravitational constant           1 1 23 43 232 2 2 3 3 2 4 434 4 4 1 2 3 4 0 1 0 0 0 ,0 0 0 0 0 0 1 0 1 1 , 0 0 0 1 0 0 0 ( ) x x M m gmg a aax x bd M Mlu t x xdt b b ax x b M Ml x x y t z t x x                                                               2)Dynamical state-space representation 3)Measurement equation 1)State space variables (Vector)
  • 6. Continuous-time state space model Deterministic linear time-invariant state space model is given by                 In time-varing systems, all elements of matrices , , are function (State Dynamic Model) (Measurement Model) where : n-vector, : m-vector, : -vector, A B C x t Ax t Bu t dt y t Cx t x u y l      of .t 6 Discrete-time state space model           1 1 (State Dynamic Model) (Measurement Model) where is integers (discrete-time index) x t Ax t Bu t y t Cx t t     
  • 7. 2.1 Random signals(noises) are applied to a discrete-time state-space system with random initial conditions. The input u(t) is deterministic and the system noise w(t) to be zero-mean, white*1), random Gaussian First-order Markov Process: State at t depends only on the previous state at t-1. (State-space representation)                                                  :mean value of 0 1 1 1 where 1 0, 1 and 0 0 0 , 0 0 , where . : ww T xx x x x t Ax t Bu t Ww t w t R t x Pr x x P x t Pr x t x t P t P t R t Cov x t E x t x t              2. Discrete-time Gauss-Markov Models *1 white signal :=signal value is not correlated in time, thus this implies signal has equal power at all frequencies. 7
  • 8.          : where, 0, vv y t Cx t v t v N R t   Measurement Model The model is shown as the following block diagram.  u t  w t  1x t   x t  v t  y tB  A C W 1 z I 1 : delay operatorz Figure 3 System diagram 8
  • 9. 9                           1 /2 1/2 1 1 1 1 , exp 22 is D-dimensional random vector : : : : Determinant of Linear Transformation , , T xx xxD xx T D xx xx xx T yy xx Pr x x R x x R x x R x x x x E x R Cov x E x x x x R R y Ax Pr y y R A x AR A                  Appendix: Multivariable Gaussian Density Distribution
  • 10. Since the Gauss-Markov model is characterized by a Gaussian distribution, it is completely specified statistically by its mean and variance.           1 1x t Ax t Bu t y t Cx t      Mean             (State Variance) 1 1 (Measurement) T T ww T yy vv P t AP t A WR t W R t CP t C R t       Variance 2.2 Mean and Variance Transitions 10
  • 11. 3. Bayesian Approach to Kalman filter [Problem] /State-space system model /Gauss-Markov model of the state transition mechanism. /Bayesian optimal estimation (MAP solution) Extracting the unobserved or hidden dynamic state variables x(t) from noisy measurement data {y(τ): τ=0,1, …, t}. We see that the dynamic state variable x(t) at time t is obtained through the transition probability based on the previous state (Markovian property), and the knowledge of the underlying conditional probability. 11 In a random process the next state depends only on the current state and not on the sequence of events that preceded it. This specific kind of "memorylessness" is called the Markov property. A stochastic process has the Markov property if the conditional probability distribution of future states of the process depends only upon the present state, not on the sequence of events that preceded it. A process with this property is called a Markov process. Quote from Wikipedia
  • 12. Bayesian Approach – Prediction + Correction Scheme-                                    ( ) 1 1 : : 0 , 1 , , 1 , Bayesian Estimate: ˆ arg + approach for computing 0 , 1 , , 1 , t t x t t t t Data Y y y y t y t x t Max Pr x t Y Pr x t Y y y y t y t Y Pr y t x t Pr x t Y Pr y t       Prediction Correction      1 t t Pr x t Y Y  12 Prediction Phase Correction Phase
  • 13. 13                            1 1 (1) , ( ) ( 1), 1 ? , ? ˆ 1 , t t t vv ee Pr y t x t Pr x t Y Pr x t Y Pr y t Y N Cx t R t N x t t P t t N y t t R t                                     1 1 From the Gauss-Markov model, we have , ˆ 1 , 1 ˆ 1 , vv t t ee Pr y t x t C t x t R t Pr x t Y x t t P t t Pr y t Y y t t R t        
  • 14.                 1 ˆ ˆPrediction: 1 1 1 ˆPrediction error: 1 1 1 1 1 tx t t E x t Y Ax t t x t t x t x t t Ax t t Ww t                            1 1 1 1 1 1 T T ww P t t E x t t x t t AP t t A R t            Using the process model, we have (ignore the input term Bu(t-1)) State prediction error covariance 14
  • 15.                                      1 1 1 Substituting these into Eq (1)gives 1 exp 2 1 ˆ ˆexp 1 1 1 2 1 ˆ ˆexp 1 1 2 T vv T T ee p x t Y t y t Cx t R y t Cx t x t x t t P t t x t x t t y t y t t R y t y t t                                                        Rewrite the terms above as follows: : ˆ1 1 ˆ: 1 ˆ 1 measurement noise v t y t Cx t state estimation error x t t x t x t t innovation e t y t y t t y t Cx t t             15
  • 16. Optimal Bayesian estimator is the processor that maximizes the posterior above. Taking natural logarithms of both sides of Eq.(2),                             1 1 1 1 ln Pr ( ) ln Pr ln Pr 1 1 1 1 1 2 2 1 R 2 t t t T T vv T ee ln x t Y y t x t x t Y x t Y LnK v t R t v t x t t p t t x t t e t e t               The MAP estimate is obtained by differentiating eq.(3), setting it to zero, and solving the equation. That is,                    1 1 1 1 exp 2 1 exp 1 1 1 2 1 exp (2) 2 T vv T T ee p x t Y t v t R v t x t t P t t x t t e t R e t                           16 (3)
  • 17.                 ˆ 1 1 ln Pr 0 ln Pr 1 1 0 MAP x t x X x t T vv x t Y x t Y C R t y t Cx t P t t x t t                                             1 1 1 1 1 1 1 1 1 1 1 ˆ1 1 then, ˆˆ 1 ˆ1 1 (4) T vv T vv T MAP vv T vv C R t C P t t x t P t t x t t C R t y t x t t X t C R t C P t t P t t x t t C R t y t                               This representation can be simplified by using a matrix inversion lemma     1 11 1 1 1T T T Matrix Inversion Lemma A BD A A B I D A B D A          17
  • 18.                                    1 1 1 1 1 1 1 1 1 It is possible to show 1 thus, 1 1 Substituting above into eq. (4), ˆˆ ˆ1 1 ˆ= T vv T vv T vv T map vv P t t C R t C P t t P t t P t t C R t C I P t t P t t I P t t C R t C x t t X t P t t P t t x t t C R t y t x t                                                               1 1 1 ˆ1 1 ˆ ˆ= 1 1 ˆ= 1 where is called . T T vv vv T vv t P t t C R t Cx t t P t t C R t y t x t t K t y t Cx t t x t t K t e t K t P t t C R t                   Kalman gain 18       We can derive the following covariance-update relation. 1P t t I K t C P t t    
  • 19. 19                               ˆ 0 0 , 0 0 ˆ ˆ1 1 1 1 1 1 1 1 ˆ ˆ1 1 1 ww T ee vv x P x t t Ax t t Bu t P t t AP t t A R t e t y t y t t y t Cx t t R t CP t t C R                      Initial Conditions : Prediction : Innovation: Kalman Filter Algorithm                       1 1 ˆ ˆ 1 1 T ee t K t P t t C R t x t t x t t K t e t P t t I K t C P t t             Gain: Update :
  • 20. 1t  t y t  1y t     ˆ 1 1 , 1 1x t t p t t        ˆ 1 , 1 x t t p t t      ˆ ,x t t p t t Measurement Measurement Correction process Correction process Prediction by Model Prediction by Model Figure 4
  • 21. Understanding Kalman filter operation and its Gain ( :Kalman gain)new old new ˆ ˆX X KE K  Prediction: Use state space model Correction: Use measurement               is small is large small Reliable model is small large large Reliable measurement is large small T vv new old new new vv vv K t P t t C R t ˆ ˆK X X ˆK X KE P t t K R t P t t K R t         1
  • 22. 22 References: [1] J. Candy, “Model-based Signal Processing”, John Wiley/IEEE Press, 2006 [2] J. Candy, “ Bayesian Signal Processing Classical, Modern, and Particle Filtering Methods”, John Wiley/IEEE Press, 2009 [3] S. Maybeck, “Stochastic Models, Estimation, and Control”, Vol. 1, Academic Press, 1979, 1999 [4] R. Brown, “Introduction to random signal analysis and Kalman filtering”, John Weily, 1983