| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
Hi,
I can't get my head around why to use voxel layer if at the end, collision will be checked in 2D (as robot footprint is 2D). Thus, even having 3D obstacle data in costmap is no use and adding RGBD sensor to obstacle layer should perform perfectly with appropriate max_obstacle_height parameter.
Please let me know where am I going wrong.
Thank you!
So you can raytrace and manage the world in its native coordinate systems that would then be projected into 2D for planning purposes. If you squash it to 2D and using 2D raytracing on native 3D data, its not going to work as well or accurately represent the world, necessarily.
Hence, use Voxel grid for 3D data as a standard default (or STVL or NPVL or eq.) unless you're running on such limited compute that you need to do only 2D raycasting (e.g. Raspberry Pi and below, though even an RPi4 can easily handle an instance or two of Voxel Layers).
You don't have to use a voxel layer. If you don't typically care about the third dimension, you are better off using an occupancy grid. However, if you were to make a flying (or swimming) robot, the third dimension suddenly becomes more important, and a voxel grid necessary.
Asked: 2022-07-19 02:08:11 -0500
Seen: 430 times
Last updated: Jul 20 '22
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
