Fault Free Integrity of Mid-Level Voting for Triplex Differential GPS Solutions
Abstract
Landing systems for large unmanned air vehicles have
stringent integrity requirements as well as demanding
system continuity requirements that often lead to triplex
avionics architectures. Triplex avionics architectures are
designs that have triple redundancy for key functions.
Mid-level voting (MLV) algorithms that select the median
value from among the three solutions are commonly used
to select among available sensors and navigation
solutions. When each solution is computed using a single suite of avionics, such median values are robust to single
airborne sensor failures and provide improved unfaulted
accuracy as well. Robustness to single faults results
because a single faulted sensor will not impact the
solutions computed by the other two sets of avionics.
System accuracy is improved for zero mean error
solutions because the median value is more concentrated
about the truth than any of the single solutions. When
performing fault tree analysis for integrity risk in the
unfaulted case, it is common to treat sensors’ errors as
being mutually independent. In the case of multiple
carrier phase differential GPS (CDGPS) solutions, this
assumption is invalid due to common atmospheric errors
and common reference receiver errors. This paper aims to
quantify the unfaulted integrity risk from triplex
correlated CDGPS solutions for float, fixed, and almost
fixed baselines that use a MLV algorithm. The bound on
the integrity risk is compared with that of independent
solutions to show the impact of incorrectly assuming
independence of CDGPS solutions. Triplex performance
is compared to simplex to show improvement or
degradation in unfaulted availability of integrity.