Society of Robots - Robot Forum

Software => Software => Topic started by: ranzdelfin on September 19, 2009, 08:11:51 AM

Title: visual sensor
Post by: ranzdelfin on September 19, 2009, 08:11:51 AM
hi my name is ranz and i would like to ask your help about a certain problem. I want to build a robot that can automatically sense and detect a certain object like for example a crumpled paper or a soda bottle. The reason why the examples were like trash is because i want o create a trash robot just like wall-e for my future thesis. please help me. THANKS!!
Title: Re: visual sensor
Post by: Miketronix on September 19, 2009, 05:05:25 PM
You may want to narrow down your thesis a bit.  What you are trying to do is near the state of the art for small robots.  The robot you are talking about will also need expensive hardware capabilities and a lot of processing.  Think about all the things that you have implied your robot will do:

1)  Search for an object -- this involves recording how much you have moved from your starting point and the location of any obstacles that obscure the line of sight or block your path.  You will need obstacle detection sensors, or you could use a simple motion base like the iRobot Create where you can just bump into obstacles and record their location.

2)  Detect a specific object -- this can be done but may require expensive sensors.  Your best bet is to pick a simple shape like a soda can that is a single contrasting color to your environment and use CMUCam to find it in relation to your robot.

3)  Move to the object -- this involves estimating the location of the desired object repeatedly and using it to plan to motion commands.  You will have to figure out what to do when you hit an obstacle on the way.

4)  Grab the object -- How do you position your gripper so that it is exactly positioned around the object, without bumping the object out of place in the process?

5)  Lift the object

6)  Search for the "trash bin".  Even if you knew where this was in relation to your starting point, most navigation sensors are inaccurate enough that you will need to do a little "search" to relocate the trash bin.

7)  Move to the trash bin.  If you don't locate the object correctly with respect to the bin, it may fall on the floor.

8)  Release the object into the trash bin.  That part was easy!

9)  Repeat

So depending on what hardware you have in your lab, you might want to work on just on piece of this problem rather than the whole thing.

Good luck!
Title: Re: visual sensor
Post by: ranzdelfin on September 20, 2009, 06:07:24 AM
thanks man!! i'll take note of what u've said..
Title: Re: visual sensor
Post by: gaurav.p on September 20, 2009, 06:59:21 AM
for  differentiating of objects u can use objects of diff colour and diff, using colour sensor and ir sensor to detect the obj and lastly trial and error is very important in this project of yours