We take up optimality results for robust Kalman filtering fromRuckdeschel[2001,2010] where robustness is understood in a distributionalsense, i.e.; we enlarge the distribution assumptions made in the ideal model bysuitable neighborhoods, allowing for outliers which in our context may besystem-endogenous/propagating or -exogenous/non-propagating, inducing thesomewhat conflicting goals of tracking and attenuation. Correspondingly, thecited references provide optimally-robust procedures to deal with each type ofoutliers separately, but in case of IO-robustness does not say much about theimplementation. We discuss this in more detail in this paper. Most importantly,we define a hybrid filter combining AO- and IO-optimal ones, which is able totreat both types of outliers simultaneously, albeit with a certain delay. Wecheck our filters at a reference state space model, and compare the resultswith those obtained by the ACM filter Martin and Masreliez[1977], Martin[1979]and non-parametric, repeated-median based filters Fried et al.[2006,2007].
展开▼
