The GNSS integer ambiguities: estimation and validation
The GNSS integer ambiguities: estimation and validation
The GNSS integer ambiguities: estimation and validation
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>The</strong>n Ua = {â ∈ R n |<br />
p<br />
∩<br />
i=1 |vi| ≤ 1<br />
2 }. <strong>The</strong>refore, the probability P (â ∈ Ua) equals<br />
the probability that component-wise rounding of the vector v produces the zero vector.<br />
This means that P (â ∈ Ua) is bounded from above by the probability that conditional<br />
rounding, cf.(Teunissen 1998d), produces the zero vector, i.e.:<br />
Ps,LS ≤ P (â ∈ Ua) ≤<br />
p<br />
<br />
<br />
1<br />
2Φ( ) − 1<br />
2σvi|I i=1<br />
(3.41)<br />
with σv i|I the conditional st<strong>and</strong>ard deviation of vi. <strong>The</strong> conditional st<strong>and</strong>ard deviations<br />
are equal to the diagonal entries of the matrix D from the LDL T -decomposition of the<br />
vc-matrix of v. <strong>The</strong> elements of this vc-matrix are given as:<br />
σvivj = cT i Q−1<br />
â cj<br />
ci 2 Qâ cj 2 Qâ<br />
In order to avoid the conditional st<strong>and</strong>ard deviations becoming zero, the vc-matrix of<br />
v must be of full rank, <strong>and</strong> thus the vectors ci, i = 1, . . . , p ≤ n need to be linearly<br />
independent.<br />
<strong>The</strong> procedure for the computation of this upper bound is as follows. LAMBDA is<br />
used to find the q >> n closest <strong>integer</strong>s ci ∈ Z n \ {0} for â = 0. <strong>The</strong>se q <strong>integer</strong><br />
vectors are ordered by increasing distance to the zero vector, measured in the metric<br />
Qâ. Start with C = c1, so that rank(C) = 1. <strong>The</strong>n find the first c<strong>and</strong>idate cj for which<br />
rank(c1 cj) = 2. Continue with C = (c1 cj) <strong>and</strong> find the next c<strong>and</strong>idate that results in<br />
an increase in rank. Continue this process until rank(C) = n.<br />
In Kondo (2003) the correlation between the wi is not taken into account, which means<br />
that instead of the conditional variances, simply the variances of the vi are used. <strong>The</strong>n<br />
the following is obtained:<br />
p<br />
i=1<br />
<br />
2Φ( 1<br />
2σvi<br />
with the Ps,i equal to<br />
Ps,i =<br />
2<br />
√<br />
2πσvi<br />
<br />
) − 1 =<br />
1<br />
2<br />
0<br />
p<br />
i=1<br />
<br />
exp − 1<br />
2<br />
Ps,i<br />
x 2<br />
σ 2 vi<br />
It is known, cf.(Teunissen 1998d), that<br />
p<br />
i=1<br />
<br />
2Φ( 1<br />
2σvi<br />
<br />
) − 1 ≤ P (â ∈ Ua)<br />
<br />
dx (3.42)<br />
This means that it is only guaranteed that Kondo’s approximation of the success rate<br />
is a lower bound if P (â ∈ Ua) is equal to the success rate. This will be the case if p is<br />
Quality of the <strong>integer</strong> ambiguity solution 41