Inside GNSS Media & Research

MAR-APR 2018

Issue link: https://insidegnss.epubxp.com/i/960969

Contents of this Issue

Navigation

Page 61 of 67

62 Inside GNSS M A R C H / A P R I L 2 0 1 8 www.insidegnss.com bis distance, the respective pseudorange is rejected. e integrity module calculating the projection levels uses the pseudoranges that passed the Mahalanobis distance check in the measurement rejection module as well as information from the track database. In consequence, the position solution provided by the integrity module is independent of the odometry, in the sense that the odome- ter readings do not enter in the position calculation. For projection level calcu- lation, a solution separation approach is used. In solution separation RAIM, the spread of subset position solutions with respect to the full set position solution is assessed. Subset position solutions are obtained by excluding subsets of satellites from the position calculation. us, the number of satel- lites that are excluded must reflect the number of simultaneous faults that need to be considered. e probability that a higher number of simultaneous faults occurs is very low, but might still be included in the integrity risk bud- geting. The major dif ference bet ween a c onvent ion a l s olut ion s e p a r at ion R AIM outlined above and the pro- posed integrity concept is that from the pseudoranges which have passed the measurement rejection module, a three dimensional position solution is not calculated, but rather a GNSS- based odometer distance results. Using this GNSS-based odometer distance, a three-dimensional position solution can then be obtained from the track database. Consequently, for the calcu- lation of protection levels, the spread of GNSS-based odometer distances is assessed, not the spread of position solu- tions. Obviously, calculating a GNSS- based odometer distance instead of a three dimensional position reduces the number of unknowns from four to two, which means that with two pseudor- anges only, a solution can be obtained. More details of this integrity algorithm are given by J. Wendel et alia (2016b). Filter Bank for GNSS and INS Integrity Within the INLU project, integrated navigation systems were also addressed, in which the soware GNSS receiver is combined with inertial sensors in loose, tight, and ultra-tight coupling architec- tures. Such architectures do not allow for the calculation of protection levels using ARAIM algorithms because these require that full set and subset position solutions are calculated using snapshot least squares. A variety of techniques can be found in literature which aim at the provision of integrity for integrated navigation systems. Batch processing approaches re-formulate the GNSS/INS data fusion as a least squares problem, which then allows us to apply RAIM or ARAIM techniques (M. Joerger and B. Pervan). Another option is to use a filter bank. Each elemental filter in the filter bank is robust with respect to a specific fault. In the simplest case, an elemental filter does not process the measure- ments of a specific satellite. In case the measurements of this satellite are faulty, this filter is not affected. is also avoids the need for a pseudorange fault model which is an advantage because such a model is in general rarely available. Examples of filter bank approaches can be found in (M. Brenner; J. Diesel and S. Luu), with the basic concept illustrated in Figure 4 with three filters only. Obvi- ously, a real GNSS/INS filter bank con- tains many more filters. For the single fault case, the number of elemental filters matches the number of satellites from which measurements are available, with possibly an additional elemental fil- ter assuming no faults are added. Each of the elemental filters propagates its state and covariance matrix forward in time using the measurements provided by an inertial measurement unit (IMU). When pseudorange and delta range measurements become available, they are processed by each elemental filter except for those the elemental filter assumes to be faulty. Hereby, the model probabilities are also updated. For each elemental filter, the model probabili- ties represent the likelihood that the assumptions of the filter are correct, i.e., that the satellites that the elemental WORKING PAPERS FIGURE 4 Simplified schematic of a GNSS/INS filter bank FIGURE 3 Block diagram of the INLU rail integrity concept

Articles in this issue

Links on this page

view archives of Inside GNSS Media & Research - MAR-APR 2018