Hi all,

I found some typos and things that weren't clear to me in the tutorial. This thread is for corrections and suggestions regarding the "Getting started/Introduction to BFL" tutorial.

unclear:
2.2.1: Although the first example is fully linear, extendedkalmanfilter.h is included. This is somehow misleading. Why not just kalmanfilter.h?

consistency:
2.2.4: In the example files the prior density is named "prior_cont", but in the tutorial just "prior". The same applies to 2.3.4 where both names appear.

unclear:
2.2.7: When getting the posterior with PostGet() there is "my_filter" out of nowhere. Shouldn't it be just "filter"?

typo:
2.3.2: When creating the derivative there is a 'd' after the ';' - it reads "Matrix df(3,3);d"

unclear:
2.4.2: It says we need to implement only _one_ function called SampleFrom(). A few lines later there is talk about implementing "ExpectedValueGet()". I think the statement "one function" is wrong, as ExpectedValueGet() has also to be provided.

typo:
2.4.3: The constructor has two typos: NonLlinearMeasurementPdf (one 'l' too much) and ConditionlPdf (missing 'a').

unclear:
2.4.3: Same as in 2.4.2: It says we need to implement only _one_ function called ProbabilityGet(). A few lines later there is again talk about implementing "ExpectedValueGet()". Is this a copy/paste error?

consistency:
2.4.3: When returning the probability of the expected value there is an additional "-measurement" in the example files.
return _measNoise.ProbabilityGet(expected_measurement-measurement);

Additionally, before diving right into the examples a short outline of the procedure would have helped me. Like for example:

System model:

- create (gaussian) density to model system uncertainty
- create pdf class depending on your problem (not necessary in the linear case)
- implement necessary functions for intended filter
- instantiation of that pdf class
- creating system model from pdf

Measurement model:

- create (gaussian) density to model measurement uncertainty
- create pdf class depending on your measurment characteristics (not necessary in the linear case)
- implement necessary functions for intended filter
- instantiation of that pdf class
- create measurement model from pdf

Final steps:

- create prior density
- create filter using prior

I think BFL is a very fine piece of software and want to thank all the contributors!

Best regards,
weaker