Efficient estimation of system states of a poorly modeled 2-D target tracking system using evolutionary strategy based particle filter algorithm

Abstract

High level of uncertainty present in the movement of an object in manoeuvering target tracking problem makes the system hard to model. Such a system is nonlinear as well due to the irregularity present in the availability of radar measurements. A conventional particle filter designed for this problem has the limitation of sample loss, which can be handled effectively by an evolutionary strategy based particle filter. Such filter can tackle complex nonlinear system such as the one just described, but its performance suffer for dealing with larger number of particles. Present work investigates one task scheduling scheme among processors, which helps in improving the estimation accuracy of one evolutionary particle filter by incorporating more measurements in its computation.