Is there a robot monitor node?

asked 2016-07-27 06:48:16 -0500

Michael Stoll
71 ●6 ●8 ●12

updated 2016-07-31 05:51:10 -0500

I am looking for a node which can monitor the robots health

It should:

Monitor nodes, topics, tf frames, ...
Publish diagnostic_msgs messages
Have a configuration which describes the robot's system (like frames, nodes, etc.)

It might be similar to roswtf, but as a node and with robot specific extensions

UPDATE:

The purpose of the node is to get a operational level (as defined in REP107) of the system. This is trivial for components, which publish diagnostics messages themselves. But other components might be supervised by their external interface (topics, tf frames, ...).

edit retag flag offensive close merge delete

add a comment

5

answered 2016-07-31 16:23:16 -0500

ahendrix

47576 ●179 ●367 ●662 http://namniart.com/

There isn't a single node that can monitor all of these things, but there is a good approach you can take:

Subscribing to every topic to monitor its frequency is very bandwidth intensive. Use the diagnostic_updater library in each node, and have it monitor any topics it publishes.
Use nodes from diagnostics_common_diagnostics to monitor TF and time sync
Use libsensors_monitor to monitor CPU temperature and fan speeds
Set up the diagnostic_aggregator to validate that other diagnostics are being published. This is where you do the top-level description of the system and validate that all of the diagnostics that you expect are actually published.
Monitor the output of the diagnostic_aggregator with rqt_robot_monitor

edit flag offensive delete link

Comments

Great, the diagnostics_common_diagnostics package is pretty close to what I've been looking for.

Michael Stoll ( 2016-08-01 02:50:44 -0500 )edit

add a comment

1

answered 2016-07-31 14:05:21 -0500

Marcin Bogdanski
231 ●2 ●7 ●9

I had similar problem couple months ago. Needed to monitor robot for errors/warning, topic latency, diagnostic info etc. I tried rqt_robot_monitor but found it rather convoluted and difficult to setup. I ended up writing dedicated node to listen monitor other nodes and display easy to understand status info to the user. Unfortunately it was very project specific :| Would be great to have a proper generic tool for health monitoring.