Fast and efficient traversable region extraction using quantized elevation map and 2D laser rangefinder
Conference: ISR 2016 - 47st International Symposium on Robotics
06/21/2016 - 06/22/2016 at München, Germany
Proceedings: ISR 2016
Pages: 5Language: englishTyp: PDFPersonal VDE Members are entitled to a 10% discount on this title
Lee, Lae-Kyoung; Oh, Se-Young (Department of Electrical and Electronic Engineering, Pohang University of Science and Technology (POSTECH), Pohang, Gyeongbuk, South Korea)
In this paper, we propose a fast and efficient traversable region extraction using quantized digital elevation map (Q-DEM) from the data obtained by the 2D laser rangefinder at indoor-outdoor environments. Generally, the structure of tilting a 2D laser rangefinder is a widespread strategy to acquire precise 3D point clouds. But in this research, using low-cost microcontroller-based modularization and improved tilting mechanism which controls the tilting motion for each of the scanning areas, we can not only obtain reliable and dense 3D point clouds with relative uniform distribution from single laser scans, but also enhance the accuracy of sensor measurements. Furthermore for fast computation and efficient management of the raw 3D data acquisition, we firstly adopt the modified voxel grid filtering with adaptive sampling of scalar distance fields, and then generate a grayscale reconstruction based quantized elevation map by applying a non-linear measurement model for extracted data sets. Especially through the proposed quantized elevation map representation, we have a relatively simple and fast data processing operation by leveraging the advantages of existing image processing techniques. Finally, for the implementation of the stable traversable region extraction, we mainly divided into two main categories, "traversable" and "non-traversable regions" with histogram and edge/texture information of Q-DEM, and then more detailed distinction for terrain classification (flat region, slope, stair, and obstacle) is performed according to the characteristics of each terrain. The experimental results show that our proposed method has a stable terrain classification performance(Avg. 83%) based on the fast map generation(Avg. 10ms) with an effective mapping capability, regardless of the variety of environmental characteristics. As a result, our proposed method was able to make a more stable path generation and utilization for mapping and navigation in given environment.