Robotic Project

[matav]

Hey techies... i need some help regarding my robotic project.
the idea is making a robot that can detect a red-colour ball, pick it up (with the help of a robotic arm) as well as avoid obstacles.
.
can anyone give me some ideas on differetiating between a red-ball and an red coloured obstacle (like an egg).

 

[Nodsu]

You want to enter a robot contest, don't you?

The only difference between the ball and the egg would be the size/shape. Read the rules and see if the size of the ball is defined.

Colour detection is usually done by using light sources of different colors. For example, in an unrealistically ideal case, when you shine red, blue and green light on the red ball, then the ball would be visible only when using the red light source.

Or, you could just hook up a webcam and do some image processing..

If you want to build everything, then you may want to start here: http://www.societyofrobots.com/robot_tutorial.shtml

You can also buy very specialised ready-made robot parts with most of the assembly already done for you.

 

[matav]

well... it's not for any contest but for my college project.
.
we have made our basic idea but need someone to help us improve this idea.
our idea:
ball diameter : 3inch
ball colour : red
colour sensors : atleast 3, maximum 5 sensors
obstacle detector : atleast 3
robotic arm (with rotation)

placement:
object sensor: 0.5inch above ground

considering the ball is exactly in front of the robot
colour sensor:
two at 1.5inch above ground (i.e. at centre of ball on either side of ball)
& one 2.9inch above ground (i.e at top of the ball)
& one 0.2inch above ground (i.e at base of ball)
.
the problems arising are:
1. if the ground is not exactly flat, then obstacle detectors (being close to the ground may generate errors)
2. if there is a larger ball place, also red colour, then how should the robot behave
3. while finding the ball (i.e rotating), the robot may get distracted or malfunction and may not stop at the exact postion
4. we need a robotic arm working on digital input... either we'll make it ourself (as last resort) or buy it (provided we can program it the way we need) but the actual problem arising is that the robotic arm should move fluently towards the ball and should not goof-up while trying to grip it.

could u share your idea, i mean... like how you would make such a robot.

 

[Nodsu]

Heh.. I don't have the time to design stuff for you..

You don't need color sensors to get the ball diameter, do you? You need only one colour sensor to see what color the thing in front of you is. For dimension checking you don't need colour-sensitivity. Unless you have to cound for multi-colored objects like a ball that has a red spot on it of course.

Do you need a robotic arm to reach out and take the ball? Maybe you could just ride on top of the ball and grab it under your belly? This would make dimension checking easier too.

Also, the object you are interested in, the ball, can only reside on flat ground - otherwise it would roll away. Of course, it could be that your robot is on a slope and the ball is on flat ground right next to the slope.

 

[charbaby]

Good luck

I hope you have someone helping you. Or your going to be in college for a long time.

 

[matav]

@charbaby >>> lol...
.

Heh.. I don't have the time to design stuff for you..

if you dont wanna design for me, it's no prob
n did i say i need a color sensor to kno the ball diameter? thats lame. wat made u think that?
.
well... i guess i'll be using IR sensors or the Ultrasonic method.
n i'd prefer a robotic arm... it's more challenging.
.
i'm getting where i want to be, hoped on recieving ideas.
i just posted my basic idea since i dont have time to post the entire details of my project here, my group has already done huge modifications to it.

 

[Nodsu]

considering the ball is exactly in front of the robot
colour sensor:
two at 1.5inch above ground (i.e. at centre of ball on either side of ball)
& one 2.9inch above ground (i.e at top of the ball)
& one 0.2inch above ground (i.e at base of ball)

I suppose I just misunderstood that.

 

[matav]

hmm... i was foolish there, i should have explained more, besides that was the very basic idea which had many basic stupid flaws.
nywys. that has many errors so we switched to IR sensing.
i personally want to so ultrasonic detection of objects, but it's apparently tougher than using IR sensors.
we are even thinkin about the possibility of CMU Camera.
.
the programming will be a challenege since this is our first time doing such a project.:bounce:
havent decided on the language yet... maybe C, C++ or ASM or... i have no idea of any other

 

[Samstoned]

stick with C++
use camera's for both object color and size will also do vector distance
to get idea of what the camera needs to do try motion positioning search on ebay the should have a few camera's on auction this will lead you to the specific software that supports vector imaging
(random thought)I think photoshop cs3 does distance analyzing and vectoring of movies
may find more info at lego mindstorms

 

[matav]

to get idea of what the camera needs to do try motion positioning

for this, we're planning to use a CMU Camera v3.
.
one friend wants to do ASM, the rest agree on C++.

 

[Samstoned]

asm will lock you to one type of architecture ,would it not?
friend does this with Pick Processors and sure it runs stand alone ,but after building device ,if the program needs more functions you must add more processors or upgrade whole board
again same with DSP once you pick the chip you could be stuck

you can embedd asm in C++ ,ask your buddy???
PS my friend says its easyer than C++ or V C++
so its the electronics that may help judge what the code is
I do the hardware structured end
rails ,ways ,screws and motors
we are setting up mach3 software that uses camera support ,but not near the video end of it yet

 

[matav]

hmm... yeah! ASM will be difficult to redo l8r.
one guy advised us to use a software Matlab.
besides we changed our idea a bit.
the robot will have
Moving Assembly
CMU Camera
NO robotic arm. but instead just the two grippers/fingers attached to the front of the robot assembly.
.
two sensors will be place on the either finger and one on the palm, to detect the ball.
The CMUCam will lock onto the object, move towards it's centroid, meanwhile the sensors will be detecting to see when the ball comes within it's grip.