This package contains ROS libraries and interfaces for measuring network statistics without introduction additional network overhead.

To see the library in action, using TF, run:

roslaunch lcsr_nettools tf_statistics.launch

The following topics will be published in addition to /tf:

• /statistics/tf_frames/robot/header_samples
  • This contains samples of the header message from the /robot tf frame.
  • It is published each time the tf_statistics node receives a tf message.
• /topic_statistics
  • This contains the computed statistics over a window of 10 seconds, and since the start of the node.
  • It is published every 5 seconds.

One of the main libraries in this package is liblcsr_nettools which provides an API for measuring the packet loss and min/max/avg/var latency of a topic with messages with ROS Headers.

This class is used to compute statistics for a given topic. It is intended to be used from within code which is already subscribed to a given topic, so that additional socket connections are not needed to gather information about a given data stream.

It can compute the folllowing min/max/average/var over a window and for all time:

• Message latency
• Message frequency

Usage of the class involves primarily two functions:

• StatisticsTracker::sample(...) Sample a header data point
• StatisticsTracker::publish(...) Publish the topic statistics on the topic_statistics topic.

You can view the statistics with rqt_plot or rxplot. In order to record the statistics, you can use rosbag to record the messages published on the topic_statistics topic.

See the StatisticsTracker API for more information.

If you don't mind making another topic connection, you can run the statistics_node which subscribes to the header_samples topic and instantiates a StatisticsTracker. You can publish to header_samples from any other kind of ROS node, and these samples will be used to compute the statistics.

See examples/tf_statistics.launch for a usage example with a python node wich publishes to header_samples.

This package also includes RViz plugins which incorporate liblcsr_nettools and report diagnostics on visualization data.

These plugins include:

• lcsr_nettools/Camera: Standard rviz/Camera plugin with statistics

These plugins are loaded automatically when RViz starts. To use them, add them like you would any other display or panel element.