Re: [ros-users] Problem with canonical_scan_matcher

On Wed, Jan 5, 2011 at 10:24 PM, Zhiping <TANZ0066@e.ntu.edu.sg> wrote:

>
> Hi Eric,
>
> Were there any people moving around in the laser scan while the robot was
> stationary?
> During the experiement, i tried to isolate the robot so that the
> environment
> remained unchanged throughout the whole experiment.
>
> What does the laser scan look like? Is it noisy?
> Personally, i don't think the laser is scan is noisy. I have already
> attached the picture of the laser scan for your reference.
> http://www.youtube.com/watch?v=5NV0g0FMX6Q
>
> http://ros-users.122217.n3.nabble.com/file/n2203293/Laser_scan_information.png


Ya, it looks like what I would expect. But, as you can see, they aren't
constant, so I could understand the small variance in the position estimates
being caused by small variations in the LIDAR pings. How much of a variation
in position is reasonable due to LIDAR variance, I don't know.


>
>
> Meanwhile, i also tried to change the parameter for "sigma",
> "max_correspondence_dist" and "orientation_neighbourhood" to increase
> tolerence using these setting
>
> sigma = 1
> max_correspondence_dist = 0.4
> orientation_neighbourhood = 100
>
> What other parameters would you suggest to enhance the result ?
>

I'm not really familiar with the algorithm's parameters. We really only
looked at it for use on a system that did not have any other odometry
source, realized that it would have to interpret a person walking as
movement of the robot and therefore couldn't use it since we anticipate the
robot being used in dynamic environments.

Perhaps one of the package maintainers would have a better idea how to
adjust the parameters.

- Eric


> sorry to trouble you.
>
> With thanks,
> Zhiping
> --
> View this message in context:
> http://ros-users.122217.n3.nabble.com/Problem-with-canonical-scan-matcher-tp2197613p2203293.html
> Sent from the ROS-Users mailing list archive at Nabble.com.
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>