Humanoid Robot in MSRS

Hi all!

I was invited to present my current project in the sandbox. So I'm pleased to do so.
I'm a CS student at the University of Dortmund. Currently I'm in a project group at the Institute of Robotics, and we are preparing there a humanoid robot to participate at robocup in Atlanta (we are qualified ). Our team is called "DoH!Bots" and our webpage can be found at www.dohbots.com.

My part in team is the simulator. Currently, we are changing our simulator-plattform to MSRS.
As the first task, I have modelled the robot. It took me about one week, to assebly all the meshes, physical bodies and joints. At the beginning I had some problems with the stability (or better - elasticity) of the joints, but finally, I managed the robot to keep standing upright
More screenhots you can find here: http://www.dohbots.com/dev_simulator
Unfortunately, I haven't managed yet to get the robot shaded - had no time for it - so It's a bit dark.

I cannot provide the source by now, but I'll make it, when I clean it up

Greets,
Robert

Wow, it looks promissing! So, it is a model of a real robot, right? Are you planning to use the same control code for both the real robot and the MSRS simulated one?

I can understand that the simulation of the NXT differential drive gives you enough realistic feedback to develop code that will fairly work in the real life, but this kind of humanoid robot.. Do you think the model and the simulated physics are that much accurate to provide useful experimentations??

Regards,
Raul

Hi Raul,

array2001 wrote:

Wow, it looks promissing! So, it is a model of a real robot, right? Are you planning to use the same control code for both the real robot and the MSRS simulated one?

Yes, the model is based on our real robot and - yes - there will be also the same robot code running on the real bot and for the simulated one (but not necessary the same walking/cognition etc. solutions)

array2001 wrote:


I can understand that the simulation of the NXT differential drive gives you enough realistic feedback to develop code that will fairly work in the real life, but this kind of humanoid robot.. Do you think the model and the simulated physics are that much accurate to provide useful experimentations??

Regards,
Raul

You are right - the simulated humanoid and the real one aren't really comparable and compatible. The walking-engine of the real one won't work well on the simulated robot and the other way round. But there are still a lot of things, that you can develop or at least design on the simulator first and then adapt to the real robot. Begining on the "drafts" of - for example - kicking move trajectories, through image processing: f.e. self-locator (with an omnivision camera, this topic is a bit tricky on a simulator, but not impossible), and last but not least: behavior.
If there is a team of about 10 people that want to make a briefly test of their work and you have two robots - there is a massive need of a simulator - and it hasn't to be as exact as the reality (or better: as noisy as the reality ), since the fine-tuning has to be made in any case on the real robot.

Greets,
Robert
www.dohbots.com/dev_simulator

Wow ... it is very exciting to see this robot ... i have also assembled a lynxmotion humanoid without all your electronic stuff and i'd love to see a tutorial on how you create it in MSRS (or maybe modify yours if you put your source code here [A]) ...

Good job

Hi!

Walking, kicking, standing up movies are available!
Check out: http://www.dohbots.com/dev_simulator and comment how you like them.

Greets,
Robert

Hi, I have seen the videos, nice job.
Your robot make complex movimets like walking.
Could you give me some suggestion to move the Kuka's arm through angles given by file?

tnx

irotech wrote:

Hi, I have seen the videos, nice job.
Your robot make complex movimets like walking.
Could you give me some suggestion to move the Kuka's arm through angles given by file?

tnx

Hi,
You can take a look at Laurent Lessieux's webpage:
http://www.lessieux.com/mywiki/index.php?title=Microsoft_Robotics_SDK_Work
There you can find there a Kondo Projectfile (under "18 September 2006"), where some animations are read from a file and then replayed.

After RoboCup (mid July), we will write a team report, where I will write a detailed description on howto create an humanoid robot in MSRS, how the interface to our robot-code is done and so on.

Ciao,
Robert