Target tracking using interacting multiple models with particle filtering

Abstract

The problem of modeling accuracy in target tracking has been well studied in the past and is specially important when tracking maneuvering targets. One of the most simple and elegant ways of improving an algorithm in this sense is by using Interacting Multiple Model (IMM). IMM is a method that takes into account more than one model at the same time. This paper describes how it works and how it has been incorporated in tracking algorithms in the past, specially in the Extended Kalman Filter (EKF). We also introduce a novel way of using it with Particle Filters (PF). The original proposal found here is that we estimate the whole target state sampling particles from the Optimal Function.