[ros-users] Nav Stack with nonholonomic robots

Eitan Marder-Eppstein eitan at willowgarage.com
Tue Sep 7 18:39:18 UTC 2010


Christian,

A couple of things:

* The carrot_planner is, most likely, not what you want. All it does is seed
the local planner with a point that lies on the straight line between the
robot and the specified goal point that is in free space. This can make the
local planner susceptible to local minima and will often fail to produce
feasible plans. We use the carrot planner for things like people following
where you are guaranteed that your goal point is very close to the robot. If
you want a planner that will produce global plans that can be followed
exactly (or at least close to exactly), I'd suggest checking out the
sbpl_lattice_planner package as Sachin mentioned. The caveat here is that
the code is still experimental, so you should expect the same level of
documentation and testing that something like navfn has.

* As I mentioned to Eric, if you want a local planner that just attempts to
follow a global path, you can check out the pose_follower package. I've only
used it with a holonomic robot, so there's no guarantee that it'll do what
you want, but it would be a quick thing to check.

* The local planner exists for two reasons. The first is that navfn only
uses the inscribed circle of the robot to create a plan, which means that
the global plan may not actually be feasible for the robot, and additional
planning that takes the full footprint into account is required. The second
is that global planning can take some time and a local planner is needed to
make sure that the robot doesn't hit anything at a fairly high rate. If you
had a global planner that was very fast and took the kinematics and dynamics
of the robot into account, only a very simplistic local planner would be
needed. Unfortunately, navfn is a planner which requires a local planner
that has more freedom to break from the global path for things to work well.

* I'll try to take a look at the simulation you set up sometime in the next
day or two.

Hope all is well,

Eitan

On Tue, Sep 7, 2010 at 9:41 AM, Christian Verbeek <
verbeek at servicerobotics.eu> wrote:

>  Eitan,
>
> After some more hours playing I found that all the pain comes from the
> base_local_planner. This is a powerful building block, but it does not
> address my needs.
>
> The plan produced by the carrot_planner is looking good all the time and
> what I need is not a local planner but something just eating up the
> global plan. If there is an obstacle along the global plan than the
> robot should stop and after a while with the obstacle in the global
> costmap replan and follow the new global plan. After all I am missing
> the idea behind the local planner with a perfect global plan and
> obstacles also be handled within the global costmap.
>
> Regards
> Christian
>
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20100907/01321a3d/attachment-0003.html>


More information about the ros-users mailing list