I just downloaded the kFilter library (http://kalman.sourceforge.net/), and had some questions about its use that I couldn't find in the documentation. Has anyone used this library in the past?
My questions are basically these:
The Step function of the eKFilter receives two vectors (u, and v). What do these vectors represent? The only reference I could find was the comment saying
"// U u U-D Covariance Matrix (n, nn)"
I assume that one of these vectors is supposed to represent the new measurement (presumably v). Is the other supposed to represent the co-variance of the measurement? How are these values expected to be inserted?Normally, the Kalman Filter will not be expecting measurements on regular time-intervals. Rather, I would expect that a time would accompany every reading indicating the actual time that it occurs. In the examples given, a constant value (called Period) is used. Moreover, none of the virtual functions in the EKFilter class can recieve any inputs. How can one use t开发者_如何学运维ime as an input corresponding to a new measurement? Similarly, the example given has constant R and Q matrices. How can one use the co-variance as an input corresponding to a reading?
u
is the control input. It's usually something like linear and angular velocity.
v
is now called z
, it's the observation vector.
Data is usually interpolated so it's on a constant interval. Your measurement covariances, Q and R will be constant in the system as well.
精彩评论