Re: Webot contest and random results

--- In [email protected], Olivier Michel 
<Olivier.Michel@c...> wrote:
> Hi,
> 
> Sorry for my late response...
> The problems you encounter are due to the asynchrounous mode of 
> operation for robot. The simulator doesn't wait for the robot and 
runs 
> as fast as possible. In order to run your simulation in the same 
> conditions as the contest ones, I would recommand you to check 
the "run 
> real time" option in the Webots preferences. This leaves more time 
to 
> the controller process and make the simulator wait a little bit.
> 
> We are forced to use the asynchronized mode for the contest, 
otherwise 
> some robot controllers could take too much CPU time leading to 
endless 
> simulations.
> 
> -Olivier
> 
> s0172326@a... wrote:
> 
> >Hi
> >
> >We just watched the first match of our robot and noticed that it 
behaves very 
> >differently online and when using the webots evaluation (note it 
just begins to 
> >walk at every match start, using exactly the same commands):
> >
> >- starting the webots program: the robot falls most of the time
> >- pushing the 'revert' button (in real time mode): the robot 
falls in 1 of 20 
> >tries
> >- running the simulation step by step: the robot never falls down
> >- webots contest: the robot fell in 3 of 3 tries
> >
> >it seems that the simulation works differently depending on the 
resources 
> >available (as a test, we started playing mp3 files while running 
the 
> >simulation: this caused an instant fall of the robot).
> >
> >Unfortunately, this makes it almost impossible to predict the 
performance of a 
> >new controller :( we think the step-by-step-performance should 
determine the 
> >outcome of a match, as this should be independent of system 
resources. 
> >Uploading a new controller every day, with new settings, is not 
an option as 
> >this might take years to develop valid locomotion
> >
> >the Mummahad_Ai coaching team
> >
> >P.S: we have added a waiting time at match start now. not a 
pleasing solution 
> >(actually, tomorrow we're going to see if it's a solution at 
all...)
> >  
> >

I've experienced the same problem in todays match, or something like 
it... I've changed the settings in webots to 16ms per update, as it 
seems like 32ms was slower than what was happening on the server. 
maybe you could tell us which settings to use, as it seems to have 
quite an impact on robot-behavior. I think it sounds like a good 
idea to compute a match without any graphics in a mode like step-by-
step. the match doesn't really have to be transmitted directly over 
the internet? What also didn't seem to be working was my camera. 
normally sonovabic gets up straight away when he falls over.



------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/pU_rlB/TM
---------------------------------------------------------------------~->

To unsubscribe from this group, send an email to:
[email protected]

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/