Joint Estimation of Target Range and Velocity for Radar System via Complex-Valued Kernel Quasi-Newton Method

Abstract

This article deals with the online joint estimation of the target range and velocity for the monostatic radar system with considering the nonlinear relationship between the target parameters and the observed echo. The nonlinear estimation problem of the two parameters is, therefore, formulated in the reproducing kernel Hilbert space (RKHS) with multiple inputs and multiple outputs. Herein, the complex-valued random Fourier feature (CRFF) method is applied to approximate the kernel method in a fixed-dimensional architecture to address the issue of dimensionality growth in complex-valued kernel learning. The <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">complex-valued random Fourier feature kernel quasi-Newton</i> (CRFFKQN) algorithm is proposed by solving a quadratic optimization problem in the random Fourier features space, which utilizes the information of all estimation errors up to the current iteration, thus resulting in the significant estimation performance improvement over the existing algorithms that only use the current information to estimate the target parameters. Numerical simulations are performed to illustrate the superiorities of the proposed algorithm against some counterparts in terms of estimation accuracy and efficiency.