We have about 50us of jitter in our realtime loop, and it isn't a problem for us, but our robot is also statically stable.

We're currently using the 3.0.6-rt17 kernel, which is a bit old at this point, but it has proven more stable than the previous PR2 kernel. My configuration is set up with a minimal number of drivers built into the kernel, and with a large number of USB device drivers built as modules. Since we use NFS on the PR2, there are a number of NFS-related things enabled; and probably a few other obscure things we use that most other users don't need.

Before you worry too much about the remaining jitter in your system, I would try to determine how much jitter is acceptable. This seems to have some ideas on where to start in that process: http://www.cs.st-andrews.ac.uk/~pn/establishing-timing-reqs-mm02.pdf

If you find that you need lower jitter, I would ask on Linux RT users mailing list: linux-rt-users@vger.kernel.org

Regards,
-Austin

On 04/10/2012 01:35 PM, Soo-Hyun Yoo wrote:
Hello Austin,

Would you be willing to share your PREEMPT_RT kernel configuration file for the PR2? I have tried disabling all unnecessary (in my judgment, which could be wrong!) features for kernel v3.0.25 patched with PREEMPT_RT v44, but the EtherLab Master node still suffers from a jitter of around 60 us.

Does that sounds reasonable? The high jitter could just be due to our use of an Intel Atom processor compared to the PR2's Core i7's, but I don't know how (if at all) that would affect the jitter. What kind of jitter do you get on the PR2?

Soo-Hyun Yoo

On Wed, Jan 11, 2012 at 11:56 AM, Austin Hendrix <legotown@aol.com> wrote:
This sounds nearly identical to the system we use here at Willow Garage on the PR2.

We're using a PREEMPT_RT kernel (2.6.31+, moving to 3.0) with a 1kHz control loop in user space.

-Austin

Jonathan Hurst <jonathan.hurst@oregonstate.edu> wrote:

>Dear ROS Users,
>
>We would like to use Robot Operating System on ATRIAS, the bipedal
>robot we are building
>(http://mime.oregonstate.edu/research/drl/Site%20Pages/Atrias.html),
>but we must have reliable real-time control at 1kHz with communication
>over an EtherCAT bus.  We have assembled some existing software tools
>to enable this, including the EtherLab driver and Orocos RTT to tie
>things all together in real time.  If you are interested in this
>discussion, please feel free to join the Email list - it has a brief
>history with one or two summaries of working systems, but I hope we
>will have much more detailed discussions, or receive pointers to other
>working systems.  You can sign up for the list here:
>https://secure.engr.oregonstate.edu/mailman/listinfo/ethercat-ros
>
>Jonathan
>--
>*****************************************************************************
>Jonathan W. Hurst
>Assistant Professor
>School of Mechanical, Industrial, and Manufacturing Engineering
>Oregon State University
>204 Rogers Hall
>Corvallis, OR 97331-6001
>
>jonathan.hurst@oregonstate.edu
>Voice: (541) 737-7010
>Fax: (541) 737-2600
>http://mime.oregonstate.edu/research/jhurst/
>******************************************************************************
>_______________________________________________
>ros-users mailing list
>ros-users@code.ros.org
>https://code.ros.org/mailman/listinfo/ros-users
_______________________________________________
ros-users mailing list
ros-users@code.ros.org
https://code.ros.org/mailman/listinfo/ros-users