Scribd
Upload
35 views

Algorithm (Kalman Filter/Extended-Kalman Filter) : Why Need of Kalman?

The document outlines the steps of the Kalman filter algorithm for nonlinear systems. It discusses: 1. The need for Kalman filters to handle nonlinear system behavior by linearizing nonlinear functions. 2. The 6 main steps of the Kalman filter algorithm: predict state and error covariance, predict output, calculate Kalman gain, update state estimate using measurements, and update error covariance. 3. Key concepts and equations used in the Kalman filter, including predicted and estimated values, prediction errors, covariance, and deriving the equations for each of the 6 steps.

Uploaded by

Ehtisham Khan Niazi
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
35 views

Algorithm (Kalman Filter/Extended-Kalman Filter) : Why Need of Kalman?

The document outlines the steps of the Kalman filter algorithm for nonlinear systems. It discusses: 1. The need for Kalman filters to handle nonlinear system behavior by linearizing nonlinear functions. 2. The 6 main steps of the Kalman filter algorithm: predict state and error covariance, predict output, calculate Kalman gain, update state estimate using measurements, and update error covariance. 3. Key concepts and equations used in the Kalman filter, including predicted and estimated values, prediction errors, covariance, and deriving the equations for each of the 6 steps.

Uploaded by

Ehtisham Khan Niazi
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

ALGORITHM (KALMAN

FILTER\EXTENDED-KALMAN FILTER)
WHY NEED OF KALMAN?
As our cell model behavior is non-linear, kalman filters helps us recursion of non-linearity in linearity.

STEPS INVOLED:-
1. Predict the state which is to be measured.
2. Determine the error covariance of predicated state
3. Predict system output
4. Estimate the gain matrix LK
5. State estimate measurements update
6. Error covariance of update measurements
BACKGROUND KNOWLEDGE FOR EMPLEMENTATION OF KALMAN
FILTER

Superscript “ - ” indicates a predicted quantity based only on past measurements.


Superscript “ + ” indicates an estimated quantity based on both past and present measurements

Symbol “ ^ ”(hat) indicates a predicted or estimated quantity: 𝑥ℎ𝑎𝑡 + or 𝑥ℎ𝑎𝑡 −

Symbol “ ~ ”(tilde) indicates an error: the difference between a true(x) and predicted or estimated
quantity:

xtilda= x – xhat

+
Estimated state (CALCULATING 𝒙𝒉𝒂𝒕

For this mean squared error method which say that

‘’finding an value of xhat that minimize the expected square length of an error vector conditioned on
previous system measurements’’.
Result from above derivation

PREDICTION ERROR

It is defined as true value minus the predicated value result is predicated error

INTITUION (we consider the estimated value based on present measurements to be true value

𝑥𝑘 = 𝑥ℎ𝑎𝑡𝑘+ )

Also

RE-ARRANGING THE EQUATION FOR Xhat+

FINDING AN EXPRESSION FOR E[ xtilda(-) | yk ] Pathway to finding


kalman gain matrix (Lk)

Expected value for two Gaussian vectors ( x,y) is given as


Applying this formula to find

Putting all these pieces together results in general updated state estimate equation

UPDATED STATE ESTIMATE COVARIANCE

Now as we have knowledge of state estimated or predicated quantity as well as its covariance we can
proceed to implementing kalman filter.

KALMAN FILTER ASSUMES THAT SYSTEM BEING MODELED CAN BE REPRESENTED IN STATE SPACE FORM

State space is given as

Input is u_(k) , output is y_(k), process noise w_(k) , sensor noise is u_(k)

DERIVING PREDITCITED STATE step(1)


This is step (1) of kalman filter , as derived earlier xhat_k(-) is equal to
Putting of state space value for function for x_(k)

DERIVING STEP (2) ERROR COVARIANCE OF PREDICATED STATE

As discussed earlier that prediction error is equal to

Therefore covariance of predicated error is

DERIVING 3rd STEP (PREDICT SYSTEM OUTPUT)

h(k)=y(k)

The general steps are not implementable as computer programs, since they involve statistical operations
(expected values) . However, the specialized Kalman-filter steps we have developed so far are
completely valid program operations.
DERIVING STEP 4 ( FINDING KALMAN GAIN MATRIX)
As matrix was

Now we will finding

COVARIENCE FOR Ytilda_(k)

ALSO

Now combining

and in kalman gain matrix gives us

DERIVING STEP 5(STATE ESTIMATE UPON L_(K))


DERIVING STEP 6(ERROR COVARIENCE OF UPDATE STATE ESTIMATE)

SUMMARY OF 6 STEPS

You might also like