porting to 32bit float

[leongzhang]

Hi, thanks for the work~
i see that the implementation here is with the double precision floats. I wonder if that was a choice based on accuracy or stability? have you tested/ported (or plan to test/port) to a 32bit float version?
I'm trying to identify a possible path to implement ukf on an small ARM M series microcontroller with 32 bit floating point support but not much for the double precision.
thanks!
gz

[ivo-georgiev]

Hi, you can try single precision behavior with simple change typedef on 39 line (double to become float) in System_Types.h.

[leongzhang]

Hi, you can try single precision behavior with simple change typedef on 39 line (double to become float) in System_Types.h.

Hi, i just downloaded it and tried on the matlab (2016a) that i have. it seems with single precision it is broken. some states become NaN.

(p.s. tried cfg1 and 2 both)

[ivo-georgiev]

Bad.... I have to check too.