ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

How stable is ROS for long-term operations?

asked 2012-10-24 03:34:59 -0600

Roba gravatar image

updated 2014-04-20 14:09:29 -0600

ngrennan gravatar image


We are considering using ROS for an industrial robot application. The application requires that we need to run the robot for more than 100 days continuously without errors (software crash, etc) and human intervention. The use of ROS with its topics/services/etc. will considerably reduce development time for our application, but how will ROS perform for long term testing?

Anyone who has experience with long-term testing of systems running on ROS (e.g., > 10 days continuous operations)?

(Note: the "Office Marathon" by Willow Garage is perhaps one example of relatively long-term testing, but that was still "only" for 30 hours operation)

edit retag flag offensive close merge delete


I think that the answers to this question will depend heavily on the particular packages (and libraries) being used, whether nodes are listed as required, and what hardware is being used. You should probably specify ones that you may be interested in..

SL Remy gravatar image SL Remy  ( 2012-10-24 04:35:40 -0600 )edit

Thanks for the comment trinighost. We are planning on mostly writing our own nodes/packages, but we are considering making use of ROS-industrial, the OROCOS toolchain for realtime operations, and the BFL. HW is to be decided on. My main concern is the main ROS functions such as message passing.

Roba gravatar image Roba  ( 2012-10-24 04:53:54 -0600 )edit

If you're worried about the core services and want to know how robost ROS is in an industrial env I would suggest taking a cut and branching from there. So you can throughtly test what you need to work. Then move forward with that. Factor back in updates as you need them.

MarkyMark2012 gravatar image MarkyMark2012  ( 2012-10-24 05:37:34 -0600 )edit

You might have a look at Baxter from rethink robotics. It's pretty early in its life, but they're marketing it, so I'd be surprised if it wasn't fairly reliable. (I have no experience with it btw.)

thebyohazard gravatar image thebyohazard  ( 2012-10-24 09:02:53 -0600 )edit

2 Answers

Sort by ยป oldest newest most voted

answered 2012-10-26 08:03:07 -0600

tfoote gravatar image

At Willow we've done long duration experiments 10+ days: Related Blog post

edit flag offensive delete link more

answered 2012-10-24 07:37:49 -0600

SL Remy gravatar image

I'm assuming that you've already joined the googlegroup

And I'd also suggest taking a look at the multimaster SIG.

Personally I haven't used any of these stacks/packages/toolchains.. but I've done just under 10 days at a time with the same roscore (diamondback from debs with custom nodes creating and processing data) and "simple" message passing (std_msgs and sensor_msgs). My aim at that time wasn't to explore reliability, but it probably will be changing soon.

Others can also comment, but I do not think message passing is an issue at this stage, the challenge is going to surround what your rosgraph does with the messages that are being passed. How well is your activity specified, does your code meet specs, do you have an accurate spec for the code/packages that you depend on etc..

edit flag offensive delete link more

Question Tools



Asked: 2012-10-24 03:34:59 -0600

Seen: 2,054 times

Last updated: Oct 26 '12