Multi-State Measurement Processing with Factorized Stochastic Cloning
William N. Fife · AIAA SCITECH 2023 Forum · 2023
View Video Presentation: https://doi.org/10.2514/6.2023-2323.vid A stochastic cloning framework is presented for processing multi-state measurements within a factorized, consider/neglect extended Kalman filter. A multi-state measurement depends on two or more temporal instances of the same state, which breaks traditional assumptions of measurement models in Kalman filtering. Alterations to previous stochastic cloning developments for practical usage are discussed. An example of multi-state measurement dependence is presented from visual odometry. Performance of the stochastic cloning approach is illustrated in simulation for processing the visual odometry direction-of-motion measurement during a lunar descent-to-landing scenario. Monte Carlo analysis is used to demonstrate statistical consistency of the Kalman filter with the stochastic cloning augmentation.