Ask Your Question
1

rviz increase network traffic on lo interface

asked 2013-09-02 23:39:27 -0500

BenMa gravatar image

updated 2013-09-04 15:46:42 -0500

joq gravatar image

Hi, there,

I have a robot with Velodyn 32E installed. This laser outputs data at around 18Mbps. In my configuration, the robot is configured as ROS master, a base station computer is configured as a client. When I start the launch file to launch the velodyne laser driver node at the robot side, I also run the rviz at my base station side. Now, I have an interesting observation, once I start rviz, even without subscribe to anything, the loopback (lo) network interface at my base station computer shoots up to 18Mbps. What happened here? Thanks a lot.

UPDATE:

<launch> 
  <arg name="pcap" default=""/> 
  <arg name="calibration" default="$(find velodyne_pointcloud)/params/32db.yaml"/> 
  <include file="$(find velodyne_driver)/launch/nodelet_manager.launch"> 
    <arg name="model" value="32E"/> 
    <arg name="pcap" value="$(arg pcap)"/> 
  </include> 
  <include file="$(find velodyne_pointcloud)/launch/cloud_nodelet.launch"> 
    <arg name="calibration" value="$(arg calibration)"/> 
  </include> 
</launch>
edit retag flag offensive close merge delete

Comments

Are you running the node or nodelet versions of the driver and pointcloud conversion packages?

joq gravatar imagejoq ( 2013-09-03 06:19:11 -0500 )edit

I am running the nodelet from the velodyne driver package.

BenMa gravatar imageBenMa ( 2013-09-03 14:45:18 -0500 )edit

Please update your question to add the launch script you use.

joq gravatar imagejoq ( 2013-09-03 17:31:34 -0500 )edit

I simply run the sample launch file, <launch> <arg name="pcap" default=""/> <arg name="calibration" default="$(find velodyne_pointcloud)/params/32db.yaml"/> <include file="$(find velodyne_driver)/launch/nodelet_manager.launch"> <arg name="model" value="32E"/> <arg name="pcap" value="$(arg pcap)"/> </include> <include file="$(find velodyne_pointcloud)/launch/cloud_nodelet.launch"> <arg name="calibration" value="$(arg calibration)"/> </include> </launch>

BenMa gravatar imageBenMa ( 2013-09-04 15:38:53 -0500 )edit

What version of the driver are you running? What ROS distro?

joq gravatar imagejoq ( 2013-09-04 15:47:13 -0500 )edit

ROS is fuerte, Ubuntu12.04. I installed the stack by running sudo apt-get install ros-fuerte-velodyne

BenMa gravatar imageBenMa ( 2013-09-04 16:30:10 -0500 )edit

Since you are running velodyne_driver and velodyne_pointcloud on the same nodelet manager process, I would not expect that to generate loopback traffic by itself. Once rviz subscribes, the extra traffic makes sense. Is rviz already subscribed due to an earlier invocation?

joq gravatar imagejoq ( 2013-09-04 16:38:05 -0500 )edit

I deleted all the rviz subscriptions and use bmon to check the lo traffic. It is still around 18 Mbps. and this slows down a lot of other coms

BenMa gravatar imageBenMa ( 2013-09-04 17:00:57 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted
0

answered 2013-09-05 07:44:53 -0500

joq gravatar image

I don't yet understand what is happening. I am only writing this as an "answer" for easier formatting.

You may get some insights from running this command on the robot:

$ rostopic bw /velodyne_packets

For your 32E, I'd expect that to report about 2MB/s (bytes, not bits). But, as long as velodyne_pointcloud is the only subscriber, I would not expect that bandwidth to appear on loopback. While rostopic is running, I would.

$ rostopic bw /velodyne_points

The corresponding point cloud published by velodyne_pointcloud will be considerably larger. It should not appear on the network unless something external to the nodelet process (like rostopic or rviz) is subscribing.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower

Stats

Asked: 2013-09-02 23:39:27 -0500

Seen: 255 times

Last updated: Sep 05 '13