unhide doc section in PDF
parent
08487f7a43
commit
ab88471c2a
|
@ -2291,15 +2291,11 @@ uncalibration
|
|||
used in the residual).
|
||||
\end_layout
|
||||
|
||||
\begin_layout Standard
|
||||
\begin_inset Note Note
|
||||
status collapsed
|
||||
|
||||
\begin_layout Section
|
||||
Noise models of prior factors
|
||||
\end_layout
|
||||
|
||||
\begin_layout Plain Layout
|
||||
\begin_layout Standard
|
||||
The simplest way to describe noise models is by an example.
|
||||
Let's take a prior factor on a 3D pose
|
||||
\begin_inset Formula $x\in\SE 3$
|
||||
|
@ -2353,7 +2349,7 @@ e\left(x\right)=\norm{h\left(x\right)}_{\Sigma}^{2}=h\left(x\right)^{\t}\Sigma^{
|
|||
useful answer out quickly ]
|
||||
\end_layout
|
||||
|
||||
\begin_layout Plain Layout
|
||||
\begin_layout Standard
|
||||
The density induced by a noise model on the prior factor is Gaussian in
|
||||
the tangent space about the linearization point.
|
||||
Suppose that the pose is linearized at
|
||||
|
@ -2431,7 +2427,7 @@ Here we see that the update
|
|||
.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Plain Layout
|
||||
\begin_layout Standard
|
||||
This means that to draw random pose samples, we actually draw random samples
|
||||
of
|
||||
\begin_inset Formula $\delta x$
|
||||
|
@ -2456,7 +2452,7 @@ This means that to draw random pose samples, we actually draw random samples
|
|||
Noise models of between factors
|
||||
\end_layout
|
||||
|
||||
\begin_layout Plain Layout
|
||||
\begin_layout Standard
|
||||
The noise model of a BetweenFactor is a bit more complicated.
|
||||
The unwhitened error is
|
||||
\begin_inset Formula
|
||||
|
@ -2516,11 +2512,6 @@ e\left(\delta x_{1}\right) & \approx\norm{\log\left(z^{-1}\left(x_{1}\exp\delta
|
|||
\end_inset
|
||||
|
||||
|
||||
\end_layout
|
||||
|
||||
\end_inset
|
||||
|
||||
|
||||
\end_layout
|
||||
|
||||
\end_body
|
||||
|
|
Binary file not shown.
Loading…
Reference in New Issue