Patrick,

There are two things to try:

1) Record a bag file of the problem on the robot. Just record everything
since it should only take a couple of seconds to see the problem from what
you've been saying. If you send me the bag, I can try to track down what's
going on that way.

2) Reproduce the problem in simulation. If you can get this to happen in
stage in a reproducible manner, that would make it super easy to track down.
You can use the navigation_stage (http://www.ros.org/wiki/navigation_stage)
package as a starting ground, make whatever modifications you want, and post
a patch. I tried this in the morning with your configuration files to
attempt to reproduce the result, but didn't have any success.

I'm still a bit baffled about my inability to reproduce this in simulation
as I don't see what would be different between that and a real robot.
Perhaps the bag file will help. I'll also take a pass over the code again as
well to see if there's something glaring that jumps out at me.

Hope all is well,

Eitan



On Mon, Sep 27, 2010 at 2:36 PM, Patrick Goebel
<patrick@casbs.stanford.edu>wrote:

> Hi Eitan,
>
> Many thanks for the planner parameters--I already see an improvement in
> the results.  However, I am still getting velocities beyond the maximums
> I set.  I am using the latest Cturtle release (binaries) on a Ubuntu
> 10.04 LTS.  In particular, the navigation version is:
>
> ros-cturtle-navigation                 1.2.2-36~lucid
>
> I increased my max velocities to see if having them so low was a problem
> but even with max_rotational_vel set to 1.0, I just now saw 2.5 when
> asking the robot to move 1 meter in rviz.
>
> If there is any other data I can provide, please let me know.  Or
> perhaps there is another way I can check the planner velocities besides
> running the robot?
>
> Thanks,
> Patrick Goebel
>
> Behavioral Sciences
> Stanford University
>
> > Patrick,
> >
> > I tried for awhile to get the commanded velocity from the
> > base_local_planner
> > to fall outside of the maximum limits using your configuration files in
> > stage and failed. I can't get the commanded x velocity to go above 0.2
> > m/s
> > and the rotational velocity never goes above 0.15 rad/sec, both of
> > which are
> > within the specified limits. What version of navigation are you using?
> > I've
> > been testing on 1.2.2, which is the latest Cturtle release. Perhaps its a
> > problem with a previous version?
> >
> > With that said, your velocity and acceleration limits are different
> > enough
> > from the defaults that it required tuning some parameters to get
> > things to
> > run smoothly. With the parameters you provided, the robot wouldn't go
> > much
> > of anywhere. I made a few modifications, which I'll include below to get
> > things working in a reasonable manner. First, I upped the sim_time
> > parameter
> > to 2.0 seconds from the default of 1.0 second. I did this because you
> > have a
> > very low maximum linear velocity and trajectories are scored from their
> > endpoints, so simulating them a bit further helps to differentiate
> > between
> > paths. Also, since you have a really low maximum for your angular
> > velocity,
> > I felt like it might be helpful for the robot to attempt to follow the
> > global plan more closely. To this end, I upped the path_distance_bias
> > weight
> > and dropped the goal_distance_bias weight. I only spent about 2 minutes
> > tuning the parameters, so I'm sure that you'd be able to get even better
> > behavior by playing around some more (upping the sim_time even more,
> > or the
> > trade off between following the path and moving towards the goal... see:
> > http://www.ros.org/wiki/base_local_planner for parameter
> > descriptions), but
> > this at least exhibits reasonable results for me in simulation. My
> > changes
> > can be found below:
> >
> > TrajectoryPlannerROS:
> > max_vel_x: 0.2
> > min_vel_x: 0.05
> > max_rotational_vel: 0.15
> > min_in_place_rotational_vel: 0.1
> > acc_lim_th: 2.0
> > acc_lim_x: 1.0
> > acc_lim_y: 1.0
> > yaw_goal_tolerance: 0.2
> > xy_goal_tolerance: 0.2
> > controller_frequency: 5.0
> > holonomic_robot: false
> > path_distance_bias: 0.8
> > goal_distance_bias: 0.4
> > sim_time: 2.0
> >
> > Hope this helps and that all is well,
> >
> > Eitan
>
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>