Reliable Multi-Carrier Ambiguity Resolution in the Presence of Multipath

In this contribution it will be shown that a correct stochastic model will dramatically improve the degraded performance of carrier-phase ambiguity resolution in the presence of biases. The focus in this paper is on code biases caused by multipath, as it is known to be a dominant error source in high-accuracy GNSS applications. A bias in the observations will propagate in the parameter estimates, which may cause severe positioning errors. Therefore, it is essential to use the a-priori knowledge about errors in the model if possible. With the APosteriori Multipath Estimation (APME) method this is done by applying corrections to the observations prior to the parameter estimation. With the APME technique the multipath error affecting the code tracking is estimated, and this estimated error is subtracted from the code-phase measurements. APME also provides the standard deviation of the multipath estimates, so that the stochastic model can be adapted. In this contribution we will specifically investigate whether APME and the improved stochastic model may improve the reliability of ambiguity resolution, and/or may serve as an indication that reliable ambiguity resolution is not feasible. In conclusion, it will be shown that a good error mitigation method together with a correct stochastic model will significantly improve the reliability of ambiguity resolution, and if Ratio Test Integer Aperture estimation is applied APME automatically provides an indication when reliable ambiguity resolution is not possible. INTRODUCTION Carrier phase integer ambiguity resolution is a key aspect to (near) real-time high-accuracy and reliable GNSS positioning. The LAMBDA method is a very well-known mechanization of integer least-squares ambiguity estimation, which means that the highest probability of correct integer estimation is obtained. With LAMBDA the position solution is obtained in three steps. First, a standard least-squares adjustment is applied to obtain real-valued estimates for all unknown parameters. This solution is referred to as the float solution. In a second step, LAMBDA is applied to obtain the integer ambiguity estimates. Finally, the float position estimates are adjusted according to their correlation with the ambiguities.