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.
in practice, a user has to choose between two undesirable situations:<br />
• Integer <strong>validation</strong> is based on wrong assumptions, since the r<strong>and</strong>omness of the<br />
fixed <strong>ambiguities</strong> is incorrectly ignored;<br />
• No attempt is made to fix the <strong>ambiguities</strong> because of the low success rate, although<br />
there is still a probability that the <strong>ambiguities</strong> can be fixed correctly.<br />
Some <strong>integer</strong> <strong>validation</strong>s tests are defined such that the invalid assumption of a deterministic<br />
fixed ambiguity estimator is avoided. However, these tests lack a sound theoretical<br />
foundation. Moreover, often fixed critical values are used based on experience. But that<br />
implies that in many situations the tests will either be too conservative or too optimistic.<br />
1.2 Objectives <strong>and</strong> contribution of this work<br />
Obviously, <strong>validation</strong> of the <strong>integer</strong> ambiguity solution is still an open problem. In order<br />
to deal with the above two situations, two approaches are investigated:<br />
• A new class of ambiguity estimators is used, which results in estimators that in<br />
some sense are always superior to their float <strong>and</strong> fixed counterparts.<br />
This estimator is the Best Integer Equivariant (BIE) estimator. It is best in the<br />
sense that it minimizes the mean squared errors of the estimators. This can be<br />
considered a weaker performance criterion than that of the <strong>integer</strong> least-squares<br />
estimator, the maximization of the success rate. This might be an advantage<br />
if the success rate is not high, since then a user will generally not have enough<br />
confidence in the fixed solution, <strong>and</strong> stick to the float solution. It is investigated<br />
how BIE <strong>estimation</strong> can be implemented. <strong>The</strong> performance of this estimator is<br />
then compared with the fixed <strong>and</strong> float estimators.<br />
• A new ambiguity estimator is used by defining an a priori acceptance region, or<br />
aperture space.<br />
This new ambiguity resolution method is the so-called <strong>integer</strong> aperture <strong>estimation</strong>.<br />
<strong>The</strong> problem of ambiguity <strong>validation</strong> is incorporated in the <strong>estimation</strong> procedure<br />
with this approach. Instead of distinguishing only the probability of success <strong>and</strong><br />
of failure, also the probability of not fixing (undecided rate) is considered a priori.<br />
An <strong>integer</strong> acceptance region is then defined by putting constraints on the<br />
three probabilities; only if the float <strong>ambiguities</strong> fall in the acceptance region, the<br />
<strong>ambiguities</strong> will be fixed using one of the well-known <strong>integer</strong> estimators.<br />
<strong>The</strong> theoretical aspects of all approaches are considered, as well as implementation<br />
aspects. <strong>The</strong>reby it is avoided to go into too much detail with respect to proofs <strong>and</strong><br />
derivations; these are only included if they contribute to the underst<strong>and</strong>ing of the theory.<br />
<strong>The</strong> methods are implemented in Matlab R○.<br />
Furthermore, it is investigated how the various methods will perform. For that purpose,<br />
Monte-Carlo simulations are used, since then the ’true’ situation is known <strong>and</strong> it is<br />
possible to compare the performance of different estimators <strong>and</strong> validators.<br />
2 Introduction