On low-power analog implementation of particle filters for target tracking

We propose a low-power, analog and mixed-mode, implementation of particle filters. Low-power analog implementation of nonlinear functions such as exponential and arctangent functions is done using multiple-input translinear element (MITE) networks. These nonlinear functions are used to calculate the probability densities in the particle filter. A bearings-only tracking problem is simulated to present the proposed low-power implementation of the particle filter algorithm.

1. Introduction

Particle filters [1] are used in state estimation, where the underlying state-space models can be nonlinear and non-Gaussian. They use a large number of weighted samples, called particles, to represent probability distributions involved in the estimation. Because particle filters do not approximate the nonlinearities in the state-space systems, they are computationally complex. Hence, some particle filter applications require real-time hardware implementations that must be efficient in speed, accuracy, and power consumed.