9.3 Sensitivity Analyses
9.3.3 ISM Latency
ISM latency is the period of time between two updates of one or more parameters contained in the ISM. The ISM latency which is denoted as ππππ‘ and the contents of the ISM are interdependent [Blanch et al 2013]. That means that, based on the assumed ππππ‘, the content of the ISM needs to be adapted accordingly. ISM latency has an impact on how the ARAIM system is designed with respect to the way the ISM is distributed to the user [WG-C ARAIM 2015]:
β’ The dissemination of the ISM at dispatch for example is only suitable if the validity period of the ISM con-tent is sufficiently long. A maritime user is very likely in transit for several days or even weeks.
β’ The dissemination via L-Band RNSS either from GNSS themselves or from SBAS is deemed a more complex solution. For such an approach the GNSS and the IRN need an interface for data exchange. The ISM con-tent and latency are constrained by the bandwidth of such an L-Band RNSS link. Furthermore, this would imply an adaption of the applicable signal in space standards for the RNSS respectively.
β’ The VHF Aeronautical Mobile Route Services (AMRS) allocation is considered to be out of scope for mari-time services as it is using instrastructure related to the aeronautics applications only [Eurocontrol 2010].
168 Advanced RAIM Related Considerations
Another aspect is the capabilitiy to advise the user of any potential threats (for example GNSS signals that must not be used) through the ISM. Having the requirement of a Time-to-Alert (TTA) of 10 seconds, a distinction is made between threats which require the TTA to be detected and those who do not. To be consistent with the nomenclature from [Blanch et al 2013] the threats will be classified as High Dynamics Threats (HDT) and Low Dynamics Threats (LDT) respectively. The LDT refers to anomalies with higher latency than the TTA while the HDT are assumed to occur within the TTA. Two monitoring approaches are identified:
β’ The task of the monitoring of LDT failures (long-term monitoring) is allocated to the ground segment. This monitoring covers the fault-free integrity risk [Martini et al 2013].
β’ The task of the monitoring of HDT failures is allocated to the user segment (short- term monitoring). This monitoring covers the faulty case integrity risk. In the frame of ARAIM, this class of failures has been al-located to RAIM FD/FDE techniques [Martini et al 2013].
The ISM latency is therefore a relevant parameter when it comes to notifying the user of failures in the system.
The shorter the ISM latency is defined, the higher the ability of the system reacting on failures and advising the user respectively. This again reduces the probabilities of failures that a user would experience. This could be done for example by integrating a flag that can be set to βuseβ or βdo not useβ per satellite into the ISM. However, this would again have an impact on the bandwidth of the ISM. The ISM latency is therefore a trade-off between several aspects to be considered.
As already stated above, the ISM parameters and latency are interdependent. A higher ππππ‘ yields more conserva-tive ISM parameters due to compensaconserva-tive reasons. This property holds basically for all ISM parameters: the error characterization parameters (URE, URA and bias) as well as the failure probabilities (ππ ππ‘, πππππ π‘). The latter are assumed to grow linearly with ππππ‘. Denoting the onset probability of a fault per time unit as ππππ ππ‘, then the probability of a satellite fault ππ ππ‘ should meet the following condition:
ππ ππ‘ β₯ ππππ‘β ππππ ππ‘
9.25 Figure 9-13 depicts the impact on ππ ππ‘ as function of the ISM latency ππππ‘. In this example, a ππππ ππ‘ probability of 8E-5 is assumed. This value corresponds to the satellite failure probability as derived in section 7.3. This graph shows the minimum required values for ππ ππ‘ as conservative values are allowed according to equation 9.25.
The validation of probabilities in the order of 1E-5 would certainly require several years of uncorrelated data.
Such numbers are always based on assumptions, e.g. decorrelation time or number of samples. The probability of failure for each satellite is a sum of all fractional probabilities of various pre-defined so-called Feared Events (see section 7.2). All these aspects would need to be characterized separately and summed up together to an overall ππ ππ‘ per satellite. This is deemed a constant effort and as GNSS matures, more consolidated probabilities can be derived for ππ ππ‘ as well as πππππ π‘.
Sensitivity Analyses 169
Figure 9-13: ππ ππ‘as function of ISM latency, based on ππππ ππ‘= 8β10β5
Table 9-5 summarizes the performance in terms of availability based on several scenarios with different assump-tions for each ππ ππ‘. For all the scenarios, two constellations (GPS+Gal) have been consistently assumed. The use of a single constellation does not allow the MHSS RAIM to meet the requirements (see section 8). Therefore, the dual constellation case has been defined as the worst case compared to the use of a third constellation. The aim of this analysis is to derive potential constraints on ππ ππ‘ for which the requirement is still met. Results are de-picted in Table 9-5 where the availability for integrity and continuity is shown over the various values for ππ ππ‘ and πππππ π‘.
Table 9-5: Availability versus ISM Parameters Const ISM Parameters Availability [%]
Psat Pconst Integrity Continuity
(15 minutes) Continuity (3 hours)
GPS+Gal 1E-3 1E-4 100 100 100
GPS+Gal 1E-2 *) 1E-4 100 100 100
GPS+Gal 1E-3 1E-3 100 100 100
GPS+Gal 1E-4 1E-3 100 100 100
*) 20Β°x20Β° grid sampling has been used due to high computational effort.
As can be seen from Table 9-5, even for a ππ ππ‘ up to 1E-2 the performance demands are still satisfied. On the other side, such a high value for ππ ππ‘ yields also disadvantages: as it is shown in Figure 9-14, the number of subsets to be considered is a function of the number of available satellites and ππ ππ‘. The increase of the values for ππ ππ‘ comes along with an increase of satellite subsets to be considered in the MHSS RAIM. This implies an in-crease of the computational effort as well.
170 Advanced RAIM Related Considerations
Figure 9-14: Number of subsets as function of Psat and number of satellites
A threshold of ππ ππ‘,π‘βπππ βπππ =οΏ½πΌπΌβπποΏ½ οΏ½2 οΏ½10 has been applied (see section 5.5). It can be seen that under extreme conditions (high ππ ππ‘ and high number of satellites) the number of subsets is in the order of 9E4. From the au-thorβs experience, it can be stated that the computational effort taking into account such high numbers of subsets increases massively and its feasibility in a maritime user receiver needs to be demonstrated. In order to keep the number of subsets in a reasonable order of magnitude, the value for ππ ππ‘ should not exceed 1E-3 (see red box in Figure 9-14). Within this range, the maximum number of subsets to be considered would be in an acceptable order of magnitude.
The original aim of this section is the derivation of a reasonable assumption for the ISM latency for maritime services. ππ ππ‘ is assumed to be the only parameter in the ISM being interdependent with its latency. Therefore, the performance level as function of different values for ππ ππ‘ has been assessed, motivated by identifying a rea-sonable maximum value for ππ ππ‘ still satisfying the demands for a maritime user. However, ππ ππ‘ turned out to be not a major driver for the achievable performance which has been investigated in more depth. This finding is valid for the horizontal position component but not necessarily for the vertical one. Nevertheless, another limita-tion has been identified that is the number of subsets to be considered in the MHSS RAIM algorithm. As shown above, if ππ ππ‘ is higher than 1E-2, the number of subsets increases to the level of several ten thousands for which the computational load is deemed critical for real-time operations. Of course, as the GNSS systems will mature with time, more adequate and reliable values for ππ ππ‘ (and πππππ π‘) will be derived. It is shown that performance demands are still satisfied under given assumptions and together with the limitation on the number of subsets to be considered, a reasonable maximum value for ππ ππ‘ of 1E-3 has been identified. This value would correspond to a maximum ISM latency in the order of 12 hours.
Conclusion 171