I modeled an Omni wheel in gazebo using friction coefs. It works quiet
well. I don't think it's feasible to model the wheel using geoms, due to
the number of geoms and the high rotational velocities involved (For our
wheels I'd need 26 geoms per wheel per robot).
The trick lies in two modifications to gazebo:
1) The CollisionCallback function in ODEPyhsics needs to compute the
primary friction direction based on the surface description of the
colliding objects and their rotation.
2) The same function needs to query the colliding geoms for surface
parameters depending on the contact point and normal, since an Omni
Wheel (and other physical elements) have different friction parameters
depending on whether the collision was on the wheel's shell or side.
In order to implement this, I added a <surface>-tag to the XML
description of geoms, holding collision parameters and a positional flag.
If you are interested, I can try and make a patch out of the modifications.
cheers,
hendrik.
Sawyer wrote:
> Chris,
> The ability to adjust the ODE parameters for the friction coefs (esp axis
> specific) is not currently implemented in Gazebo world files. I was wondering
> how you would go about modeling an omni-wheel when this question was first
> proposed and I think you are onto an elegant solution.
>
> Jordi, Nate,
> Where in the svn code base are the ODE parameters for wheels adjusted? Where
> are the XML parameters read? I am trying to give Chris an idea of where the
> current system would need to be patched to allow this behavior.
>
> Sawyer
>
> (Sorry I can't be of more help, I'm currently still working with the older
> Gazebo platform)
>
> On Monday 18 February 2008 1:57:09 pm chgans wrote:
>> Hi all,
>>
>> Sorry not to reply in the thread, but i've just subscribed, and so i
>> don't have the original email.
>>
>> I'm very interesting as well in modeling omni-directional wheel and
>> holonomous robot in Gazebo.
>> I used to work on an open source project which try to develop such a
>> platform[1] (as well as an ODE-based simulator).
>> At that time we used a "normal" wheel model, but the simple trick was to
>> define the ODE friction coef along the wheel axis to be very small
>> compared to the one along the wheel path, I'm not sure gazebo allow to
>> access these type of parameters from the xml file.
>>
>> I'm really new to player/stage/gazebo, I start having a look on
>> Saturday! Did a few build, trys (and errors), and finally get everything
>> running (last player/stage (pre-)release, and a fresh svn for gazebo).
>> Try playercam, joystick, laser, sonar, i am still exploring... and i
>> enjoy! Thank's to you all!
>>
>> I'm so enthusiastic, that i would like to resuscitate the WISIRC project
>> that i start with some of my friends in 2004, but actually what i
>> realized is that all we wanted to do is something like
>> player/stage/gazebo, but the main difference is p/s/g do it for real! :-)
>> Having all this framework running, will allow me to concentrate more on
>> HW, meca, drivers, sensors, etc of the robot (i'm actually more on HW,
>> low-level SW side).
>> For this holonomic platform we developped a specific PC104 motor control
>> board[2], based on 3 HCTL1100 ICs(z filter), each of these was accessed
>> from userspace by the "motion control" daemon, which in turn was
>> accessed by the path planner. The goal was to implement the same "motion
>> control" interface within the simulator, and we almost did, almost...
>> Then we decide to start working with a DSP running the motion control
>> algorithm (2 PIDs, one for controlling the linear position/velocity, the
>> second one for controlling angular position/velocity), which has
>> odometry directly connected (wheel encoder + laser motion sensors, such
>> has found in laser mouse for gamer), and then the project start slowing
>> down and down and ... dies. (snif! ;))
>> Ah, and i didn't tell you about our plan for having DSP-based (TI C6711)
>> camera imagery , etc... ;)
>> Well, the big difference between p/s/g and wisirc, is that our goal was
>> to build completely from scratch (HW, Meca, SW) an 99.99% open-source
>> holonomic platform. As hobbyist, the fun is in the do-it-youself! (And i
>> don't have the money to buy all these very expensives ready-to-uses
>> sensors)
>>
>> I'm really happy to see all the features that p/s/g offers, very
>> exciting. So for now, i have to create these configuration/model files
>> for my bot, and start digging into the player's drivers/interfaces model.
>>
>> Have fun!
>> Chris.
>>
>>
>> [1] http://home.gna.org/wisirc/images/wizbot1-20040901/
>> http://download.gna.org/wisirc/MecaWiz/wizbot-20060819/chassis_1.png
>>
>> http://download.gna.org/wisirc/MecaWiz/wizbot-20060819/new_wheel_1.png [2]
>> http://home.gna.org/wisirc/projects/WizHard
>> PS: this site is really out of date, unfortunately our newer wisirc.org
>> web site disapeared recently...
>>
>> -------------------------------------------------------------------------
>> This SF.net email is sponsored by: Microsoft
>> Defy all challenges. Microsoft(R) Visual Studio 2008.
>> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
>> _______________________________________________
>> Playerstage-gazebo mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/playerstage-gazebo
>
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Playerstage-gazebo mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/playerstage-gazebo
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Playerstage-gazebo mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/playerstage-gazebo
