View Full Version : control of a robotic arm using neuro fuzzy system.

01-08-2005, 02:03 AM
hi all,
this is my first post in this forum.I am doing B.Tech in Mech.Engg in nitc.ac.in.i want to do a project related to robotics.the problem is,i dont have much idea about robotics,i am only starting to learn things.as a bigining,i had made a biped with 4 dc motors and interfaced it thru parallel port.the dc motors were cheap ones,with i myself adding the reduction.there was no feedback at all.
now i want to do a better project.i have some basic knowledge of ANN Fuzzy Systems,so i was thinking about fabricating a robotic arm and doing some control using ANN-Fuzzy.I have 3 stepper motors(1.8 degree resolution)..can someone suggest any good project;like whether to build a (movable)platform and place the arm on it etc etc and what should i control,obstacle avoidance of the arm or the platform etce etc
I have some crude encoders made from those coming with scroll mouses.i added a belt and pulley system, so that it has got pretty good resolution.
i have a budget of abt 6000 rs max.
Thanks in adv,

01-08-2005, 11:06 AM
Using crude things u wont achive much precision . Also for an arm you will need feedback like the servo motors . For arms SErvos are always recommended .

There are lots of projects with the arm that you could do ... But will have to figure out IKS first .

My advice build a fixed arm . Get the basic movement working and depending on the precision you get decide on a project .

01-08-2005, 04:00 PM
hi vikas,
thank you wery much for your advice.
i certainly want to do a better project, but money is a huge constraint!!but not having money cant justify not doing a project and thats why i built all these from scrap.
the 'crude' encoder has got pretty good resolution.the one coming with mouse had 36 holes and i added a 1:10 reduction so i am getting abt 360 ppr.
my doubt is,if i builld a 3D 4dof arm will i be able to do the (neuro-fuzzy) control ; i have seen quite a few articles/papers on this, but most (or all?) of them were just 2D arms.
so whether we should do the moving platform thing with say a 3 dof planar arm and do the control of the movement of the platform instead.
all sorts of advices are welcome.
thanks in adv,

01-09-2005, 04:50 PM
If it is an arm you are trying, why not try a simple thing like a drafter,
the tool tip would be costly but that would probably be the only costly component.

01-11-2005, 11:20 AM
hi arun...

basically i m just an amateur and i dont hav much of an idea ....i m just an undergrad student
but since u ask i guess if u plan to incorporate NN and fuzzy u are looking for some control system that will be robust and learn frm its data...noisy data

A robotic mechanism that can pick up objects frm one area and place them in another area may be the kind of problem u are looking to solve...

the parameters that u can hav which will help train ur NN may be

1. varying the size and shape of the objects.....so now u ll hav to code to clasp the object accordingly and place it also accordingly....e.g. a smaller object will be realeased at a lesser height frm ground than a bigger one.

2. locating the object more and more accurately by using sensors

3. finding locations to place the objects appropriately

this is wat i can think of as of now.....

pls tell me if it helps ....

01-11-2005, 05:16 PM
Zaphal and Kaushal,thanks for replying.
Zaphod,to be sincere,I didnt get what you were trying to convey.Please be kind enough to help me out.Were you trying to tell about building a drafter with motor at joints...
Kaushal,your reply was certainly helpful,

a smaller object will be realeased at a lesser height frm ground than a bigger one.
excellent.Let me have a chat with my team mates.
About identifying the object using sensors,what kind of sensor can be baught within our budget say 5000 bucks?
Again,All kinds of suggestions are welcome.
[ http://arunsdiary.blogspot.com ]