Implementation of the Auxiliary Sampling Importance Resampling Particle Filter on Graphics Processing Unit

Abstract

In this paper, we investigate the performances of sampling importance resampling (SIR) particle filter and auxiliary sampling importance resampling (ASIR) particle filter. We mention the reasons behind ASIR particle filter which is the improved version of SIR particle filter. We use target tracking with radar application whose measurements are highly nonlinear in the experiments. At first, we implement the serial versions of these particles filters on a single processor. Although their qualities improve as the number of particle increases, their execution times deteriorate rapidly. Then we implement them on a graphics processing unit (GPU). GPU offers promising solutions for the particle filters as it has many cores in its architecture. We implement SIR and ASIR particle filters by using CUDA programming language and achieve important speed ups along with same qualities on GPU.