View Chapter

Chapter 45 — World Modeling

Wolfram Burgard, Martial Hebert and Maren Bennewitz

In this chapter we describe popular ways to represent the environment of a mobile robot. For indoor environments, which are often stored using two-dimensional representations, we discuss occupancy grids, line maps, topologicalmaps, and landmark-based representations. Each of these techniques has its own advantages and disadvantages. Whilst occupancy grid maps allow for quick access and can efficiently be updated, line maps are more compact. Also landmark-basedmaps can efficiently be updated and maintained, however, they do not readily support navigation tasks such as path planning like topological representations do.

Additionally, we discuss approaches suited for outdoor terrain modeling. In outdoor environments, the flat-surface assumption underling many mapping techniques for indoor environments is no longer valid. A very popular approach in this context are elevation and variants maps, which store the surface of the terrain over a regularly spaced grid. Alternatives to such maps are point clouds, meshes, or three-dimensional grids, which provide a greater flexibility but have higher storage demands.

OctoMap visualization

Author  Maren Bennewitz, Wolfram Burgard, Armin Hornung, Cyrill Stachniss, Kai Wurm

Video ID : 79

This video shows the Freiburg Computer Science campus in a 3-D OctoMap. Note that free space is also encoded, although not shown in the video. The map covers an area of 292 x 167 x 28 m^3 and requires only 130 MB in memory at 20 cm resolution.

3-D textured model of urban environments

Author  Michael Maurer

Video ID : 269

In this video, a micro aerial vehicle developed by the Institute for Computer Graphics and Vision, Graz Univ. of Technology, flies to predefined points and captures images for building a 3-D textured model of an urban environment. The video contains a nice description of the different steps necessary to generate a precise model by fusing the areal images with public geographic data.

Service-robot navigation in urban environments

Author  Christian Siagian

Video ID : 270

This video presents the navigation system of the Beobot service robot of the iLab, University of Southern California (USC). Beobot's task is to fulfill services in urban-like environments, especially those involving long-range travel. The robot uses a topological map for global localization based on acquired images.

Learning navigation cost grids

Author  John Rebula

Video ID : 271

The video shows how the LittleDog robot of IHMC learns a terrain cost map based on several surface parameters. The map is then used for determining foot placements for the robot to enable it to traverse rough terrain.