An Instantaneous 3D Ego-velocity Measurement Algorithm for Frequency Modulated Continuous Wave (FMCW) Doppler Radar Data
Carl C. Stahoviak · CU Scholar (University of Colorado Boulder) · 2019
This thesis describes an instantaneous 3D body-frame linear velocity measurement algorithm that operates solely on Doppler radar target data. A robust, model-free estimate of the ego-velocity vector of the sensor platform can be determined from a single radar scan, without requiring additional sensors or target data from previous scans. Additionally, this thesis lays the groundwork for the fusion of this velocity measurement into a visual-inertial-radar Simultaneous Localization and Mapping (SLAM) estimation scheme as a means to augment the traditional limitations of visual-inertial SLAM methods. Real world data sets captured onboard ground and aerial vehicle platforms are evaluated alongside a Vicon motion capture ground truth system. The ego-velocity measurement algorithm takes advantage of the radar's ability to measure both radial velocity, as well as returned signal intensity, and via the incorporation of a statistical re-sampling technique is shown to provide a robust measurement of the 3D body-frame linear velocity components of the sensor platform and their associated measurement uncertainties.