Autonomous UAV Navigation using Millimeter Wave Radar

Abstract

We present the design, implementation and evaluation of MilliNavigator, an autonomous navigation system for drones capable of mapping, path-planning, self-localizing, and navigating in indoor environments by leveraging strategically-placed millimeter wave anchors. Autonomous drones are an increasingly relevant tool for completing and automating hard-to-reach tasks. State of the art navigation systems rely primarily on cameras and GPS for environmental perception and self-localization. These solutions can impose restrictions on existing systems, which limit their navigable environment to well-lit, outdoors, and unobstructed paths. This thesis presents MilliNavigator, the first system to use millimeter wave radar and anchor-aware path planning to achieve high accuracy, 6DOF, online localization. By generating a localization precision score map from known anchor deployments, the system jointly optimizes travel distance and localization performance. We implemented and evaluated MilliNavigator on a drone built with commercial, off-the-shelf parts. We ran over 165 successful missions across 7 different tag deployments. Our system successfully achieved 7.9cm overall median error and had a 90th percentile error of less than 21cm.