Multiple model spline probability hypothesis density filter

Abstract

The probability hypothesis density (PHD) filter is an efficient algorithm for multitarget tracking in the presence of nonlinearities and/or non-Gaussian noise. The sequential Monte Carlo (SMC) and Gaussian mixture (GM) techniques are commonly used to implement the PHD filter. Recently, a new implementation of the PHD filter using B-splines with the capability to model any arbitrary density functions using only a few knots was proposed. The spline PHD (SPHD) filter was found to be more robust than the SMC-PHD filter because it does not suffer from degeneracy, and it was better than the GM-PHD implementation in terms of estimation accuracy, albeit with a higher computational complexity. In this paper, we propose a multiple model extension to the SPHD filter to track multiple maneuvering targets. Simulation results are presented to demonstrate the effectiveness of the new filter.