Cagatay Tanil, Amazon Prime Air

View Abstract Sign in for premium content


Timely detection and exclusion of slow faults in an integrated positioning sensor systems are crucial to deliver safe navigation. This paper describes a computationally cheap monitor that utilizes a sequence of a subset (partial) Kalman filter innovation vectors only associated with fault hypotheses. The paper also establishes an analytical recursive formulas to evaluate the partial innovations monitor. Both analytical and Monte Carlo results showed that the partial sequencing is superior to conventional methods using full innovation vector at each time epoch.