ABSTRACT
This chapter explores the resilience of out-of-scope measurements (OOSM), on the one hand, and imputation, on the other, to the various forms of imperfections in sensor data in order to increase the awareness of the impact delayed measurements could have when building robotic prediction models. It focuses on the application of OOSM and multiple imputation to multi-target tracking prediction. One direct solution to the OOSM problem is simply to ignore and discard the OOSM in the tracking process more like the listwise deletion is a standard default approach for dealing with missing data in most statistical packages. Several applications require the maintenance of a high-fidelity estimate of the state of a dynamic system based on a sequence of noisy observations. Such applications demand the use of filtering mechanisms such as the Kalman filter to fit the observation sequence to a given model of the system dynamics.
