Modify

Opened 6 months ago

Closed 6 months ago

#182 closed task (fixed)

Kalman filter implementation clarification

Reported by: maximilian.gold@… Owned by: stuerze
Priority: normal Component: BNC
Version: Keywords: kalman filter
Cc:

Description (last modified by maximilian.gold@…)

What is the exact kalman filter implementation used in BNC? I was not able to find this exact square root filter implementation anywhere in the literature or understand it properly. The Cholesky decomposition used in combination with a QRZ Decomposition used on an Augmentation Matrix including weights, the square root of the filter covariances and a combination of the the square root of the filter covariances and Design Matrix seems unusual. A clarification would be appreciated a lot.

Attachments (2)

QRZ-SRF.PNG (122.5 KB ) - added by maximilian.gold@… 6 months ago.
Tiberius 1998 Recursive Data Processing for Kinematic GPS Surveying p.47-48
Mervart_GW_Koll_20140127-08-10.pdf (179.3 KB ) - added by stuerze 6 months ago.

Download all attachments as: .zip

Change History (9)

comment:1 by maximilian.gold@…, 6 months ago

Description: modified (diff)
Priority: minornormal

A source in form of a paper name or Link would be sufficient as well.

comment:2 by stoecker, 6 months ago

In 10501/ntrip:

(The changeset message doesn't reference this ticket)

in reply to:  2 comment:3 by anonymous, 6 months ago

Replying to stoecker:

In 10501/ntrip:

(The changeset message doesn't reference this ticket)

Hello stoecker, could you clearify what you mean? The .PPP or Log Files?
I am aware of the Filter states and values in there, but I would be interested in some implementation Details regarding the Kalman filter.

e.g. the usage of computed - observed (ll) for the updated States instead of the observations here:
xx += KT.t() * (ll - AA * xx);

the computation of the transposed Kalman Gain KT. (what is YY?) SHi is the inverted Decomposed Measurement sensitivity Matrix H? y = Hx, y: observations
Matrix KT = SHi * YY;

And Especially the usage of the QR Decomposition of an Augmented Matrix SRF for the measurement update. Here the closest I have gotten is in:
Tiberius 1998 Recursive Data Processing for Kinematic GPS Surveying p.47-48 (see attachment)
where it is used for the time update.

Any explanation on these topics would be appreciated.

Regards

Max

by maximilian.gold@…, 6 months ago

Attachment: QRZ-SRF.PNG added

Tiberius 1998 Recursive Data Processing for Kinematic GPS Surveying p.47-48

comment:4 by stoecker, 6 months ago

Sorry. Referenced wrong ticket.

comment:5 by stuerze, 6 months ago

Status: newassigned

The square root filter was implemented by Leos Mervart and is described e.g. on page 10 of the attachment.

Best regrads, Andrea

by stuerze, 6 months ago

comment:6 by anonymous, 6 months ago

Thank you very much for the clarification.
This helps a lot.
Regards, Max

comment:7 by stoecker, 6 months ago

Resolution: fixed
Status: assignedclosed

Modify Ticket

Change Properties
Action
as closed The owner will remain stuerze.
The resolution will be deleted. Next status will be 'reopened'.

Add Comment


E-mail address and name can be saved in the Preferences .
 
Note: See TracTickets for help on using tickets.