Problem simulating kinect in gazebo

asked 2011-11-29 17:50:04 -0600

Laurie
98 ●4 ●4 ●10

updated 2016-10-24 09:03:24 -0600

ngrennan
1 ●1 ●1

Hi,

I am trying to simulate the turtlebot in gazebo. I am using the turtlebot_gazebo and stack for that (using robot.launch).

However, the depth data coming from the (simulated) kinect is all wrong. Every point in the point cloud published by the /camera/depth/points topic has a depth of 1.0. This is when the light source in the .world file is of the directional kind. When I use a point or spot light source, the bottom half of the pointcloud has depth of 1.0 and the top half either NaN or 1.0 (depending on whether something is within range or not). This is preventing me to use any SLAM algorithm in gazebo. I get the camera rgb image through correctly.

I am using Ubuntu 11.04 and electric.

Thanks a lot.

edit retag flag offensive close merge delete

add a comment

2

answered 2011-12-12 05:05:36 -0600

tfoote

58457 ●128 ●550 ●526 http://www.ros.org/

@Laurie, In general you want a discreet graphics card to run the simulations. There's a list of user tested cards at http://www.ros.org/wiki/simulator_gazebo/SystemRequirements

edit flag offensive delete link

Comments

I am using a nVidia card now, and the depth data is correctly simulated. Thanks.

Laurie ( 2011-12-21 22:58:50 -0600 )edit

add a comment

2

answered 2011-12-01 15:22:50 -0600

JonW

586 ●11 ●18 ●24

I had a similar issue with a depth camera in Gazrbo. From memory it was caused by the use of texture shadows. Setting shadows to none in the world file resolved the issue for me.