🚀 go-pugleaf

feng@srvr1 (Liqiang Feng, ) writes:

> 
> I knew that one of the big problems with the tracked vehicle is severe 
> slippage especially during the turning. I do not have any experience 
> with tracked vehicle, I have several questions and like to hear from 
> knownledgable netters:
> 
> 1. If I instruct a tracked vehicle to turn 90 degrees, what kind error 
> in orientation I should expect? say on concrete floor? 5 degrees, 10 degrees 
> or more ? In other words, is dead reckoning still a feasible option for 
> navigation? 
Here's what I've experience in the ISC (Intelligent Sensing and Control)
lab in the AI dep of the Uni of Edinburgh,Scotland,last summer:
You can't really ever rely on the accuracy of the machine when turning or
even moving, you ALWAYS gotta program it assuming that it turns 45 instead
of 90 degrees,maybe 120,depending on the floor structure,contact of tyres
with ground,actual acceleration (which is a BIG prob,as it changes the
friction- at least with 3-tyred vehicles),down to mechanical spikes in the
gear box,down to different battery power while the vehicle is in operation.
You can really clog up your vehicle with more sensor's,like IR,ultra sound,
compass,gyro..but we experienced there,that more data does always mean
more pain to the programmer,in fact you have TOO MUCH data with which you
dont know what to do with and wgich isn't always usefull to archive the
goal. (we had goals of navigating through a lab,finding back,reacting to
light sources/killing all of them and finding solutions NEVER to get stuck)
> 
> 2. Is this error consistent or more random in nature? Seems to me it is 
> random. 
Things like failing motors/mechanical structure are quiete consistent,but
once you have dealed with them programming the machine and you
actually change the set up (the machine itself) (which you do a lot! because
some kind of setups are a lot better to solve certain probs/environments,
and you generally improve the tasks solution by changing the machine,both
mechanically and sensors setup etc.) ,you start all over again,because
the software can't deal properly with the new setup/new probs arise.
Other things like floor's friction ,light conditions,battery power sucks
are more random-like,but who said that humans,which are the copied
'machines' in the end,can't deal with this (outside the lab,too- ha!).
> 
> My E-mail: feng@caen.engin.umich.edu
> 
> Thanks a lot.
bye,Kai.

--------------------------------------------------------------- 
Tel.:  +49 421 556803 (voice)
E-mail: kai@nicedel.north.de (private UUCP site)
       j08i@dhbrrz41.bitnet (not valid;gets forwarded to rs550)
       j08i@rs550.zfn.uni-bremen.de (University)
---------------------------------------------------------------