Previous Abstract Return to Session C5 Next Abstract

Session C5: Vision, Radar and Lidar-Based Navigation System

Flight Test Evaluation of a LIDAR/Inertial Navigator for Autonomous Drone Operations in GNSS-Denied Urban Environments
Anne-Sophie Polz, Fabian Frickenstein, and Maarten Uijt de Haag, TU Berlin
Location: Grand Ballroom IJ
Date/Time: Thursday, May. 1, 9:20 a.m.

This paper focuses on sensor-based navigation solutions for the operation of Uncrewed Aerial Vehicles (UAV) in challenging environments, specifically, low-level flights in urban areas under GNSS-denied (or deteriorated) conditions. A navigation system, based only on high-resolution Light Detection and Ranging (LIDAR) sensors, integrated with an inertial measurement unit (IMU) to achieve accurate and reliable positioning is proposed. The paper covers three different LIDAR-navigation approaches: (1) an absolute positioning method based on Digital Scene Matching (DSM) utilizing point clouds represented by OctoMaps and an a priori known database for global lookup, (2) an odometer-like method that compares OctoMaps from consecutive LIDAR scans for incremental pose estimates without the need for a global map (3) and an odometer-like image-based local pose estimator that uses 2D grayscale heatmaps generated from LIDAR point clouds and optical flow between consecutive images. The algorithms have been tested using flight data collected onboard a heavy-payload, hexacopter UAV equipped with state-of-the-art sensors flown in urban-like environments in Schonwalde, Germany. This paper ¨ covers a description of the proposed LIDAR-based navigation methods, the flight test setup, and the evaluation of the methods using performance metrics.
Index Terms—LIDAR, inertial, navigation, UAV, urban areas, map matching, OctoMaps, heatmaps, optical flow

Previous Abstract Return to Session C5 Next Abstract