Articles | Volume 14
Adv. Radio Sci., 14, 39–46, 2016
https://doi.org/10.5194/ars-14-39-2016
Adv. Radio Sci., 14, 39–46, 2016
https://doi.org/10.5194/ars-14-39-2016

  28 Sep 2016

28 Sep 2016

Extended Kalman Doppler tracking and model determination for multi-sensor short-range radar

Thomas J. Mittermaier1, Uwe Siart1, Thomas F. Eibert1, and Stefan Bonerz2 Thomas J. Mittermaier et al.
  • 1Chair of High-Frequency Engineering, Technical University of Munich, 80290 Munich, Germany
  • 2Ott-Jakob Spanntechnik GmbH, Industriestr. 3–7, 87663 Lengenwang, Germany

Abstract. A tracking solution for collision avoidance in industrial machine tools based on short-range millimeter-wave radar Doppler observations is presented. At the core of the tracking algorithm there is an Extended Kalman Filter (EKF) that provides dynamic estimation and localization in real-time. The underlying sensor platform consists of several homodyne continuous wave (CW) radar modules. Based on In-phase-Quadrature (IQ) processing and down-conversion, they provide only Doppler shift information about the observed target. Localization with Doppler shift estimates is a nonlinear problem that needs to be linearized before the linear KF can be applied. The accuracy of state estimation depends highly on the introduced linearization errors, the initialization and the models that represent the true physics as well as the stochastic properties.

The important issue of filter consistency is addressed and an initialization procedure based on data fitting and maximum likelihood estimation is suggested. Models for both, measurement and process noise are developed. Tracking results from typical three-dimensional courses of movement at short distances in front of a multi-sensor radar platform are presented.