Robotics StackExchange | Archived questions

Gazebo performance for very large point clouds

I am at the beginning of my bachelors university project, where I aim to achieve autonomous 3D mapping of caves using flying drones.

I need to build a simulation environment to test my algorithms. For now, I am thinking about ROS+Gazebo. I've got several Point Cloud maps of caves, which I want to use as a world model and equip the simulated robot with LiDAR sensor.

The point clouds are very large (>=2GB) and I am worried Gazebo would not be able to render this efficiently. What is the reality?

Is there a way I could achieve a realistic performance? If this will not work, do you have any idea which tool might work for me? I am thinking also about e.g. ROS+Unity.

Thanks for suggestions, it's very important for me to start on the right track.

Asked by Milos9304 on 2019-03-05 17:49:19 UTC

Comments

you could filter your pointcloud first http://pointclouds.org/documentation/tutorials/voxel_grid.php and then try to convert it into a mesh that you can import in Gazebo. https://github.com/PointCloudLibrary/pcl/blob/master/io/tools/converter.cpp . I am not sure about the converter thing, did not have the chance to try it.

Asked by mehdi on 2019-03-14 08:37:49 UTC

Answers