Revision history [back]

We indeed were wondering if ROS nodes could run fast enough. (<1ms)

Well I'm not saying it's impossible.

What is impossible is to make blanket statements about "ROS" (in quotes, as what would be possible is almost completely determined by how things are used, so saying something "about ROS" in general is rather difficult).

So depending on how you use ROS infrastructure it's possible to "run nodes" at kHz frequencies. It gets difficult when node<->node communication gets involved. Not impossible (see MoveIt's servo component fi), but difficult, as determinism is gone, and jitter will increase.

The way to go then here is use RT Linux for all 'RT' control systems and use ROS nodes for higher level planning, management and steering.

That is how this is done often in ROS 1, yes.

If you have RT requirements, I would suggest to take a look at ROS 2. The real-time working group and safety working group may be interesting and perhaps even interested.