What I've found can make a big difference in the Create is dropping the rate
of the local planner. Reducing controller_frequency to something in the
range of 5-10hz (instead of the default 20) made a huge difference in how
well the Create followed the path (even using the default acceleration
limits).

This is especially true if you aren't carrying around much computing power
(most of the time, it's a netbook or two on top of our lab's Creates), but
even seems to be applicable when you have plenty of cycles to burn.

-Mike

On Wed, Sep 1, 2010 at 10:11 PM, Steven Martin <
s34.martin@connect.qut.edu.au> wrote:

>  Hi Guys,
>
> I have the Nav stack running on a couple of irobot creates we have in our
> lab and I have implemented my own global path planner using some optimal
> control stuff I have been working on. However the local DWA path planning
> really doesn't do a very good job at following the global path. I have had a
> bit of a look through the source and the paper on DWA and it should work
> fine for the nonholonomic robots.
>
> I was wondering if anyone could suggest a set of parameters for the local
> planner? And can anyone else comment on the performance of the Nav Stack on
> the creates?
>
>
> Cheers
>
> Steven Martin
> steven.martin@qut.edu.au
>
> PhD Student
> Queensland University of Technology
> Gardens Point, S1107
>
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>