Ego-motion estimation with a lowpower millimeterwave radar on a UAV
Pieter Meiresone, D. Van Hamme, Wilfried R. Philips, T. Verbelen · IET conference proceedings. · 2023
Radar sensors have been shown to be capable of performing simultaneous localization and mapping (SLAM) tasks. However, single-chip mmWave radar sensors have received little attention because of their limited resolution. In this paper, we present a novel approach to obtain a robust ego-motion estimation of a UAV using a low power single-chip millimeter wave (mmWave) FMCW radar sensor. By using a novel method to match local radar signal descriptors, we are able to achieve a robust trajectory estimation. We then propose to optimise the trajectory by extracting loop closures from low-dimensional latent space descriptors. We validate our solution in an industrial IoT lab with a drone, but it can be applied more broadly in power contrained platforms.