[ros-users] Holonomic motion with move_base

[Eitan Marder-Eppstein]

Christian and Florian,

The default local planner that move_base uses, TrajectoryPlannerROS, follows
a heuristic where it will only exploit a holonomic base in situations where
it cannot navigate in a conventional manner. For example, the local planner
will strafe when the robot finds itself too close to a wall, but won't do so
in open space. The reason this heuristic was put in has to do with sensor
placement on the PR2, which has all of its 3D sensors facing forward. This
means it is preferable to move in a manner that keeps sensors facing in the
direction of motion for safety reasons. There is potential to write a local
planner that fully utilizes a holonomic base, but it would take a little bit
of work and we haven't had a real need for it. We tend to use move_base to
get from place to place around the office and then switch to more
specialized controllers/planners for doing things like moving the base
sideways along a table.

As far as "smooth motions" are concerned, what exactly do you mean? Is the
robot stopping/jerking a lot? Oscillating? Overshooting its goal? Etc. A bit
more information will give more of an idea about what might be causing your
problem. One thing I've noticed from looking at your configuration file is
that you've set your maximum velocity and acceleration limits to be pretty
low and are using a very long simulation time. The robot will navigate with
these settings, but I'd expect it to be pretty conservative... especially
with sim_time set to 5 seconds. This means that any velocity command that
would result in the robot hitting something if executed for 5 seconds won't
be considered valid and, around obstacles, your robot will probably run very
slowly.

Hope all is well,

Eitan

On Wed, Apr 7, 2010 at 10:42 AM, Weißhardt, Florian <
Florian.Weisshardt at ipa.fraunhofer.de> wrote:

>  Hi together,
>
>
>
> we are currently exploring the ROS navigation stack on Care-O-bot 3.
> Gmapping and amcl is already up and running yet we are now running in some
> problems with the move_base package. We are having troubles getting smooth
> motions and exploiting the holonomic base. E.g. the planner does not command
> any sidewards motions although we set the holonomic parameter in the local
> planner to true. Does anyone has experienced similar problems and knows
> which parameter to adapt? For reference we appended the configuration files
> for the base_local_planner.
>
>
>
> TrajectoryPlannerROS:
>
>   max_vel_x: 0.2
>
>   min_vel_x: 0.01
>
>   max_rotational_vel: 0.5
>
>   min_in_place_rotational_vel: 0.1
>
>
>
>   acc_limit_th: 0.5
>
>   acc_limit_x: 0.1
>
>   acc_limit_y: 0.1
>
>
>
>   holonomic_robot: true
>
>
>
>   sim_time: 5
>
>   sim_granularity: 0.01
>
>   xy_goal_tolerance: 0.3
>
>   yaw_goal_tolerance: 0.1
>
>
>
> We would be grateful for any help.
>
>
>
> Best regards from Stuttgart
>
> Christian and Florian
>
>
>
> ---
> Dipl.-Ing. Florian Weißhardt
>
> Fraunhofer IPA, Abt. 324/Robotersysteme
> Nobelstrasse 12, D-70569 Stuttgart (Germany)
> Phone +49(0)711-970-1046,
> Fax +49(0)711-970-1008
> http://www.ipa.fraunhofer.de/
>
>
>
> _______________________________________________
> 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/20100407/8dd9bb29/attachment-0003.html>