If I have m measurements to estimate an n dimensional state vector, and I am using Kalma

Blericker74 2022-07-05 Answered
If I have m measurements to estimate an n dimensional state vector, and I am using Kalman filter to do the filtering, then: Should I put all the m measurements together in the measurement matrix (measurement transformation matrix ) and perform the filtering or, should I filter each measurement sequentially? Please provide some supporting explanation for your choice.
For e.g: Let m = 2 and n = 3. The state vector is 3 dimensional and we need to use the two measurements to get the posterior estimate the state vector. Now I can use one of these two methods:

1. Use all these measurements together to form a gain matrix of size 3 × 2.
2. Use one measurement at one time and perform the filtering two times. The gain matrix will be 3 × 1 in this case.

Which of the two methods is a better choice?
You can still ask an expert for help

Expert Community at Your Service

  • Live experts 24/7
  • Questions are typically answered in as fast as 30 minutes
  • Personalized clear answers
Learn more

Solve your problem for the price of one coffee

  • Available 24/7
  • Math expert for every subject
  • Pay only if we can solve it
Ask Question

Answers (1)

Karla Hull
Answered 2022-07-06 Author has 20 answers
Short Answer:
This is an interesting question that I have wondered myself, but I have not worked out this problem. My intuition says that both approaches are equivalent as long as the measurements are uncorrelated. The reason is that if the measurements are correlated, then you must include the off diagonal terms of the covariance matrix. This wouldn't be possible if the updates are performed one after another.
However, we can know once and for all if we work out the equations, which shouldn't be too difficult. Here, we will consider scalar measurements for simplicity, but the concept applied for vector measurements as well.

Long Answer:
I suggest looking at the information form of the Kalman filter (KF) and extended Kalman filter (EKF) for more insight in this respect. The information matrix (i.e., inverse of covariance matrix) of the information form can be written in a single equation as follows where the time indices are omitted for clarity:
P 1 = ( F P 1 F T + Q ) 1 + H T R 1 H
where P is the covariance matrix, F and H are the Jacobians for the process and observation models, and Q and R are the covariance matrices for the process and observation noises.

Case 1 (Sequential Updates)
Let A = h A x i represent the Jacobian for the first measurement with measurement function h A , and let B = h B x i represent the Jacobian for the second measurement with measurement function h B . If we perform sequential updates, then we have the following:
P 1 = ( F P 1 F T + Q ) 1 + A T R A 1 A + B T R B 1 B
Now, assume the measurement is a scalar, then
A T R A 1 A = [ a 00 a 0 n a n 0 a n n ] , B T R B 1 B = [ b 00 b 0 n b n 0 b n n ]
where
a i j = R A 1 h A x i h A x j , b i j = R A 1 h B x i h B x j
Therefore,
A T R 1 A + B T R 1 B = [ a 00 + b 00 a 0 n + b 0 n a n 0 + b n 0 a n n + b n n ]

Case 2 (Simultaneous Updates) Let h=[hA,hB]T represent the measurement function, but in this case, the measurement function is a vector-valued function; therefore, the Jacobian and is given as follows:
H = [ h A x 0 h A x 1 h B x 0 h B x 1 ]
Now,
H T R 1 H = [ H 00 H 0 n H n 0 H n n ]
where
H i j = h A x j ( R A 1 h A x i + R B A 1 h B x i ) + h B x j ( R A B 1 h A x i + R B 1 h B x i )
where (with slight abuse of notation)
R 1 = [ R A 1 R A B 1 R B A 1 R B 1 ]
So, if the measurements are uncorrelated (i.e., R A B 1 = R B A 1 = 0), then H T R 1 H = A T R 1 A + B T R 1 B, which means that applying the updates sequentially or simultaneously givens the same results. However, in general, you cannot apply the updates sequentially for correlated measurements.

We have step-by-step solutions for your answer!

Expert Community at Your Service

  • Live experts 24/7
  • Questions are typically answered in as fast as 30 minutes
  • Personalized clear answers
Learn more

You might be interested in

asked 2022-06-18
Define the operator K f ( x ) = 0 1 k ( x , y ) f ( y ) d y . by
K f ( x ) = 0 1 k ( x , y ) f ( y ) d y .
where k ( x , y ) is a continuous complex valued function on the unit square. [ 0 , 1 ] × [ 0 , 1 ] Also assume that
s u p x [ 0 , 1 ] 0 1 | k ( x , y ) | d y < 1
How to show that given g ( C [ 0 , 1 ] , C ), the integral equation
f ( x ) = g ( x ) + 0 1 k ( x , y ) f ( y ) d y
has exactly one continuous solution f ( C [ 0 , 1 ] , C )?
Note: I only know introductory measure theory and functional analysis, so I am suppose to do this problem without heavy integral function theory.
One way to get started is to show that K is compact and its operator norm | | K | | < 1 because we are given that s u p x [ 0 , 1 ] 0 1 | k ( x , y ) | d y < 1 Then we know that ( I K ) is invertible and its inverse is given by the Neumann series. At this point, I'm not sure how to proceed.
asked 2022-06-30
I am a graduate student of Mathematics.In the book Measure,Integration and Real Analysis by Sheldon Axler there is a question that asks the reader to show that there exists no measure space ( X , S , μ ) such that { μ ( E ) : E S } = [ 0 , 1 ).I am not sure how to do it.I was thinking of taking a nested sequence { E n } of sets in S such that μ ( E n ) = 1 1 / n but I don't think that will work.Can someone give me any clue?
asked 2022-07-07
I need to find the volume of the intersection set A B whereby A = { ( x , y , z ) R | x 2 + y 2 + z 2 1 } and B = { ( x , y , z ) R | x 2 + y 2 1 / 2 } .
It is clear that A represents the unit ball centered at the origin and B represents the cylinder with radius 1 2 .. I should at some point use the Fubini theorem. I am puzzled by intersection set. Can somebody provide a solution proposal or a comment? Thanks.
asked 2022-04-15
How do you determine the level of measurement?
asked 2022-05-21
I need to include measurement uncertainties in this testing process. So I have a theoretical value A that is stated without uncertainty. My measured data gives me this rate value B with uncerainty ΔB. It used to be validated (B was stated without uncertainty) as if B < const*A then test passes.
I found this consistensy check that is used for comparing values with uncertainties (if ∣A - B∣ ΔA + ΔB∣ is true, then the compared values are consistent with each other within experimental uncertainty), so that would mean that in my situation, I just need to make sure that the difference ∣A-B∣ is smaller than ΔB. 1st question - if it doesn't pass this consistency test, are there two not comparable?
And second, what should I do next? How do I compare const*A and B ± ΔB? I guess the core of the answer will be in overlaping errors, but couldn't have found much on it.
asked 2022-03-16
Classify each of the following by type of variables and level of measurement.
a. Monthly payment: $1427
b. Annual family income: $86, 000
asked 2022-03-27
Describe how each of the following variables can be measured using the nominal, ordinal, interval, or ratio scale of measurement:
1. Socioeconomic status
2. Distance of household from the nearest health facility.

New questions