An Analysis of ARAIM Performance Sensitivity to the Ground System Architecture Definition

Receiver Autonomous Integrity Monitoring (RAIM) has been certified to provide lateral guidance in flight operations ranging from En-route to Non-Precision Approach (NPA) [1]. Recent developments in the RAIM algorithm science, namely Advanced RAIM (ARAIM), have suggested a future role in vertically guided operations down to LPV with a decision height of 200ft [2, 3]. However, to meet the more stringent requirements of such operations, in particular the necessary integrity, a supporting ground network is envisioned whose role will be to verify the standards provided by the GNSS service providers. The exact allocation of responsibility and risk between this ground segment, the accompanying airborne segment and the existing space segment remains to be fully defined [3, 4]. The ground monitoring may ultimately perform real-time signal monitoring akin to the SBAS concept or none at all as is the case in traditional RAIM. More likely, an offline or near real-time bounding methodology will be employed. This paper analyses in further detail how this bounding methodology may be applied. What is certain is that the airborne component will require inputs of the probabilities of failure for both individual satellites known as narrow faults, and constellations known as wide faults. In addition, inputs are needed for the bounds on maximum biases and the nominal and worst case error variances applied at the airborne receiver [5]. How these inputs will be provided to the airborne receiver algorithms will directly depend on the bounding methodology outlined above, whether fixed in the standards in the case of offline or no monitoring; or conversely, provided by a real time communications channel via a satellite or ground link. This paper addresses the question of how ARAIM performance will vary as a function of the ground network characteristics. This analysis determines how the ground network architecture - which depends on the risk allocated to the ground segment - impacts the probability of failure assumed by the airborne component. The paper quantifies the probability of satellite failure for two potential architectures [3]: firstly, based on offline monitoring employed purely in the long term to validate that the space segment assumptions are not unsound; secondly, a bounding methodology which shares the allocation of integrity risk between the ground and airborne segments and with a medium latency message update rate. In the offline monitoring case, the real-time monitoring responsibility is fully allocated to the airborne component, the total navigation system integrity risk being shared with the space segment. As such, the real probability of failure at a particular epoch is a function of the failure onset probability and temporal correlation of the nominal errors. The paper presents a derivation of the onset probability by considering the relationship of the true probability of failure onset, the empirically estimated probability of failure onset (as done historically for GPS RAIM) and the proposed ARAIM probability of failure onset [6]. In the case of near real-time monitoring, the intention of the ARAIM ground infrastructure is not to replicate the relatively dense monitoring of the current single frequency SBAS. A number of reference network baselines have been defined which vary in their density and coverage, be it regional or global [4]. This paper considers that each satellite has single reference station coverage with various elevation masks, such a network falls under the sparse global setting. Assuming this network enables an equally simple but informative ground monitoring algorithm to be chosen. Moreover, its monitoring accuracy may be determined analytically which enables the failure detection power to be computed. Naturally, many variations in the design of ground monitoring algorithms are possible and the approach taken is not intended to be optimal in any sense. Rather, it is chosen for its simplicity but also reflects the real observability of ranging errors whilst remaining conservative in the capabilities of the ground segment. The monitoring accuracy obtained as an output is a function of the residual error sources at the reference station, which in turn depend upon the minimum satellite elevation (or assumed station mask). Using the single station model, we derive the monitoring power as a function of the coverage (i.e. minimum satellite elevation of each satellite) under this simple ground segment model. In addition, simulations of the ARAIM airborne component have been performed as part of a CNES ARAIM project. These simulations have been used to generate statistics on the Minimum Hazardous Bias (biases in the multiple failure case), or equivalently, the magnitude of failure which has the potential to cause an integrity failure for the user [7, 8]. Using this input in combination with the monitoring accuracy from the ground segment, the risk of undetected failure at the time of ISM is computed. This is achieved for two cases of ISM dissemination: the first approach includes the monitoring accuracy metric described above in addition to satellite integrity flags; and the second approach is limited to the flag. In the latter limited data bandwidth scenario, the user computes a worst case monitoring accuracy based on the assumptions outlined in the paper. The final result of this analysis is to derive the threat of a single failure which occurred prior to the last ISM message arriving at the user. The risk of failure onset after the ISM message as a function of the latency is then added to give the total probability of single failure used as input to the airborne component. This system analysis cumulates in a presentation of the feasibilities and performances of the ARAIM system including the sensitivities in terms of the onset probability assumptions, the ground network dimensioning and the ISM message content and latency.