Академический Документы
Профессиональный Документы
Культура Документы
Lecture – 30
Continuous-
Continuous-Discrete Kalman Filter
Xɺ (t ) = A (t ) X (t ) + B (t )U (t ) + G (t ) W (t )
Yk = C k X k + Vk
E W (t )W T (τ ) = Qk δ (t − τ )
0 k ≠ j
E Vk V jT = Rk δ kj δ kj =
1 k = j
Mechanism
X̂ 3−
Xˆ (t )
X̂ 1+ X̂ 2+
X̂ 1− X̂ 2− X̂ 3+
Xˆ 0− = Xˆ 0+
t0 t1 t2 t3
Time (t)
Expression for Pɺ
Xɺ = AX + BU + GW ɺ
Xɺɶ = Xɺ − Xˆ
ɺ
Xˆ = AXˆ + BU
⇒ = A Xɶ + GW
t
Xɶ (t ) = ϕ (t , t0 ) Xɶ 0 + ∫ ϕ (t , t0 ) G (τ )W (τ ) dτ
0
t
RWXɶ = E ∫ W (t ) W T (τ ) G (τ ) ϕ (t ,τ ) dτ
0
t
= ∫ Q δ (t − τ ) G T (τ ) ϕ (t ,τ ) dτ = 12 Q G T
0
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 6
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Expression for Pɺ
( )
T
ɺ ɺ ɺ ɺ
Pɺ = E Xɶ Xɶ T + Xɶ Xɶ T = E Xɶ Xɶ T + E Xɶ Xɶ T
E Xɺɶ Xɶ T = E ( A Xɶ + GW ) Xɶ T
= A E Xɶ Xɶ T + G E W Xɶ T
= AP + 12 GQG T
Pɺ = ( AP + 12 GQG T ) + ( AP + 12 GQG T )
T
Pɺ = AP + PAT + GQGT
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 7
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Summary
Model
Xɺ (t ) = A (t ) X ( t ) + B (t )U (t ) + G (t ) W (t )
Yk = C k X k + Vk
Xˆ ( t 0 ) = Xˆ 0
Initialization
P0− = E Xɶ ( t 0 ) Xɶ T ( t 0 )
Gain −1
Computation K e k = Pk− C k T C k Pk− C k T + R k
ɺ
Propagation Xˆ = AXˆ + BU
(using high
accuracy numerical
integration)
Pɺ (t ) = AP + PAT + GQG T
Note:
Facts to Remember
Nonlinear estimation problems are considerably more difficult than
the linear problem in general. (EKF is just an idea…not a cure for
everything !)
The EKF even though not truly `optimum’, has been successfully
applied in many nonlinear systems over the decades
The fundamental assumption in EKF design is that the
true state X (t ) is sufficiently close to the estimated state Xˆ (t )
at all time, and hence the error dynamics can be represented
fairly accurately by the linearized system about Xˆ (t )
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 12
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Continuous-
Continuous-Continuous EKF
Continuous-Continuous
Formulation
(Assumption: Measurement is continuously available)
Xɺ (t ) = f ( X , U , t ) + G (t ) W (t )
Y = h ( X , t ) + V (t )
∂h
h( X , t ) = h( Xˆ , t ) + ( X − Xˆ )
∂X Xˆ
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 15
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Formulation
∂f
E [ f ( X , U , t ) ] = E f ( Xˆ ,U , t ) + ( E [ X ] − Xˆ )
∂X Xˆ
= f ( Xˆ ,U , t ) + ( Xˆ − Xˆ )
∂f
∂X Xˆ
= f ( Xˆ ,U , t )
Similarly, E [ h( X , t )] = h( Xˆ , t )
ɺ
Observer dynamics: ( )
Xˆ (t ) = f Xˆ , U , t + K e ( t ) Y − h ( Xˆ , t )
Yˆ = h ( Xˆ , t )
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 16
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Error Dynamics
Xɶ (t ) ≜ X (t ) − Xˆ (t )
Xɺɶ (t ) ≜ Xɺ (t ) − Xɺˆ (t )
{
= { f ( X ,U , t ) + GW } − f ( Xˆ ,U , t ) + K e h( X , t ) + V − h( Xˆ , t ) }
= f ( X ,U , t ) − f ( Xˆ ,U , t ) − K e h( X , t ) − h( Xˆ , t )
+ GW − K eV
∂f ∂h
= Xɶ − K e Xɶ + GW − K eV
∂X Xˆ ∂X Xˆ
A( t ) C (t )
Error Dynamics
Xɺɶ (t ) = [ A(t ) − K e (t )C (t )] Xɶ + GW − K eV
A0 ( t )
Xɺ (t ) = f ( X , U , t ) + G (t ) W (t )
Model
Y = h ( X , t ) + V (t )
Initialization
Xˆ ( t 0 ) = Xˆ 0
P0 = E Xɶ ( t 0 ) Xɶ T ( t 0 )
Gain
Computation K e (t ) = P (t )C T (t ) R −1 (t )
Summary
ɺ
Xˆ (t ) = f ( Xˆ ,U , t ) + K e (t ) Y − h( Xˆ , t )
Continuous-Discrete EKF
Motivation: System dynamics is a continuous-time, whereas
measurements are available only at discrete interval of time.
Strategy:
Gain
K e k ( t ) = Pk− C k− T C k− Pk− C k− T + R
Computation ∂h
w h e re, C k− =
∂ X Xˆ k−
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 23
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Summary:
Continuous-Discrete EKF
Xˆ = Xˆ + K Y − h ( Xˆ )
+ − −
k k
ek k k
Updation (
Pk+ = I − K ek C k )P (I − K
k
−
ek Ck )
T
+ K ek R k K eTk
( preferable )
(
= I − K ek C k )P k
−
( not preferable )
ɺ
Propagation
Xˆ (t ) = f ( Xˆ , U , t ); Xˆ k+ → Xˆ k−+1
Pɺ (t ) = AP + PAT + GQG T ; Pˆk+ → Pˆk−+1
where A(t ) =
∂f
∂X Xˆ ( t )
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 24
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Iterated EKF
f ( X , U , t ) = f ( X , U , t ) + A(t ) [ X − X ]
h( X , U , t ) = h( X , U , t ) + C (t ) [ X − X ]
∂f ∂h
where, A(t ) = ; C (t ) =
∂X X ∂X X
E [ f ( X , U , t ) ] = f ( X , U , t ) + A(t ) Xˆ − X
E [ h( X , U , t ) ] = h( X , U , t ) + C (t ) Xˆ − X
ɺ
Xˆ (t ) = f ( X , U , t ) + A ( t ) Xˆ − X +
K e (t ) Y − h ( X , U , t ) − C (t ) Xˆ − X
Yˆ = h ( X , U , t ) + C (t ) Xˆ − X
Recommendations/Issues in EKF
Design parameter selection:
o Fix R based on the sensor characteristics
o Select P0 to be “sufficiently high”
o Tune Q until obtaining satisfactory results
The filter should run sufficiently ahead of time prior to its usage, so
that the error stabilizes before its actual usage (else, initial error can
be very large and the associated control can destabilize the closed
loop system)
Keep the measurement equation linear wherever possible
Care should be taken to avoid numerical ill-conditioning. Methods
are available to address this issue (see Crassidis and Jenkins book).
Difficulties in Practice
Computer round-off errors
Unchecked error propagation
Asymmetry of covariance matrix: A
symptom of numerical degradation
Solution of Riccati equation
• Square-root filtering: Solution of Riccati
equation via Cholesky factorization
Large initial errors: Information filtering
OPTIMAL CONTROL, GUIDANCE AND ESTIMATION 32
Prof. Radhakant Padhi, AE Dept., IISc-Bangalore
Point to Remember:
Nonlinear estimation problems are considerably more difficult than
the linear problem in general (EKF is just and idea…not a cure for
everything!).
The EKF even though not truly `optimum’, has been successfully
applied in many nonlinear systems over the decades.
The fundamental assumption in EKF design is that the
true state X (t ) is sufficiently close to the estimated state Xˆ (t )
at all time, and hence the error dynamics can be represented
fairly accurately by the linearized system about Xˆ (t )
Limitations of EKF
Linearization can introduce significant error
Transformation of PDF:
A pictorial description
Reference: D. Simon,
Optimal State Estimation,
Wiley, 2006
System Dynamics
Selection of σ -points:
Note :
( A)T
A=A
UKF Implementation:
Measurement Update
Selection of σ -points:
Predicted measurements
References: Books
J. L. Crassidis and J. L. Junkins, Optimal Estimation of
Dynamic Systems, CRC Press, 2004. Second Edition:
2012.
D. Simon, Optimal State Estimation, Wiley, 2006.
P. Zarchan and H. Musoff, Fundamentals of Kalman
Filtering: A Practical Approach, AIAA, 2005.
F. L. Lewis, L. Xie and D. Popa, Optimal and Robust
Estimation, CRC Press, 2007.
A. Gelb, Applied Optimal Estimation, MIT Press, 1974.
M. S. Grewal and A. P. Andrews, Kalman Filtering:
Theory and Practice, Wiley, 2001.