Chief Delphi

Chief Delphi (http://www.chiefdelphi.com/forums/index.php)
-   Robotics Education and Curriculum (http://www.chiefdelphi.com/forums/forumdisplay.php?f=66)
-   -   Learning robots (http://www.chiefdelphi.com/forums/showthread.php?t=39416)

John Gutmann 24-08-2005 21:32

Learning robots
 
well i am gonna get more involved with programming and sensors and i have a couples of ideas for robots i will make to practice coding but, i only have 2 ideas:
a wall hugger
and a light/dark seeking robot

can anybody help me with any more that can use inexpensive sensors?
or atleast sensors that are reasonably priced?

NoodleKnight 24-08-2005 21:34

Re: Learning robots
 
line following =)

jdiwnab 24-08-2005 21:54

Re: Learning robots
 
Oh, I thought this thread was going to be about robots that learned... That would be cool to see a student implement. lol :rolleyes:

John Gutmann 24-08-2005 22:06

Re: Learning robots
 
Quote:

Originally Posted by NoodleKnight
line following =)

o yea i had that too but i 4got bout it

John Gutmann 24-08-2005 22:10

Re: Learning robots
 
Quote:

Originally Posted by jdiwnab
Oh, I thought this thread was going to be about robots that learned... That would be cool to see a student implement. lol :rolleyes:

eventually and hopefully, i have a theory of a basic one to learn somewhat of its surroundings

have a distance sensor and have it on a servo so it will turn when it starts up to like scan a room, then it will keep it in its memory and its position, then as it moves it will have encodes on its wheels and then it will use a math formula to determine the distance it moved from the counts on the encoder then plots its position, but then to make it better you can have something like on a optical mouse encase of somehting like sand or ice where ur wheels will turn and you wont move and a stall sensor wont be efficient

jdiwnab 24-08-2005 22:34

Re: Learning robots
 
Sorry for hijacking your thread. To make up for it, have you checked out All Electronics? They have lots of different electronic stuff (as the name implies).

John Gutmann 24-08-2005 22:50

Re: Learning robots
 
yea i love that place except for the 6$ shipping for on any order evenm if it is like a buck :(

John Gutmann 24-08-2005 22:53

Re: Learning robots
 
Quote:

Originally Posted by jdiwnab
Oh, I thought this thread was going to be about robots that learned... That would be cool to see a student implement. lol :rolleyes:

but that is a good point now that u mention it, in what ways can you make a very simple inexpensive robot learn?

but the harder question is to ask learn what?

sanddrag 24-08-2005 23:11

Re: Learning robots
 
Quote:

Originally Posted by sparksandtabs
but that is a good point now that u mention it, in what ways can you make a very simple inexpensive robot learn?

but the harder question is to ask learn what?

You can make a robot that follows a line and has wheel encoders. It follows the line and "learns" what path to take. It stores the values from the wheel encoders and is then able to replicate the same path without any line. Would that count as learning?

jdiwnab 24-08-2005 23:16

Re: Learning robots
 
Quote:

Originally Posted by sparksandtabs
but that is a good point now that u mention it, in what ways can you make a very simple inexpensive robot learn?

but the harder question is to ask learn what?

Learning the field is good and helpful, but not much more learning than line following or using encoders on a known field. Learning, I think would have to do with adapting itself to changing situations to better it's position. A truly smart robot would be able to play by itself the whole match. It would see what the other robots are doing, and be able to accomplish tasks not directly specified in responce to it's own input. As simple as line following is, it is close to what it takes, if only it could know more that what is direcly underneith it. While I think a full sized computer drivin robot might be able to accomplish some of this, I don't think those little contolers would be able to process that much. Behold the power of the human mind! 100MHz, several billion parellel processors with fuzzy logic! YEAH:D

We are working now on a way to basically program a map into the robot so it can know where it is, where it is facing, where it wants to go, and can figure out how to get there. It can tell movement based on encoders, can tell which direction it is facing with encoders, can even tell where it needs to go, but to find out how to get there with obsticals, that is hard.

mechanicalbrain 24-08-2005 23:25

Re: Learning robots
 
Definitely something that implements accelerometers. How about a robot that hugs a wall and uses the data from two accelerometer (or a 2 axis) to map outer boundaries of a room. You could have it even explore the inside if the room once it knows the boundaries. :D Slightly complex but it gives you a great understanding of the uses of accelerometers.

John Gutmann 25-08-2005 00:20

Re: Learning robots
 
Quote:

Originally Posted by sanddrag
You can make a robot that follows a line and has wheel encoders. It follows the line and "learns" what path to take. It stores the values from the wheel encoders and is then able to replicate the same path without any line. Would that count as learning?

deffinately, btu like how interms of programming could i do it? could i have have it record the inputs and just store it as one string out as like multiple bytes or words and then just read them all simutaniously,

that is an awesome idea, i deffinately will try it when i get the money for 2 encoders!

Quote:

Originally Posted by jdiwnab
Learning the field is good and helpful, but not much more learning than line following or using encoders on a known field. Learning, I think would have to do with adapting itself to changing situations to better it's position. A truly smart robot would be able to play by itself the whole match. It would see what the other robots are doing, and be able to accomplish tasks not directly specified in responce to it's own input. As simple as line following is, it is close to what it takes, if only it could know more that what is direcly underneith it. While I think a full sized computer drivin robot might be able to accomplish some of this, I don't think those little contolers would be able to process that much. Behold the power of the human mind! 100MHz, several billion parellel processors with fuzzy logic! YEAH:D

We are working now on a way to basically program a map into the robot so it can know where it is, where it is facing, where it wants to go, and can figure out how to get there. It can tell movement based on encoders, can tell which direction it is facing with encoders, can even tell where it needs to go, but to find out how to get there with obsticals, that is hard.

i was thinking of something where it would have to go to a point a certain distance away at a paticular angle, so what it would do is based on the wheel size it would determine the amount of impulses needed to get there in a straight line, but then i would throw in and obstactle and maybe with an ir rang finder if the closest thing was outside of X cm it woudl go that direction or which ever way that had the obstacle the farthest away, then it would record how many impulses off course it was and readjust.

but then of course if i wanted to do it a little different i could have it determine the distance in a straight line, then when it comes to and obstacle it would determine the angle it veered of at and update the straight line distance from its new position, and of coure it could record the position of the obstacle

Quote:

Originally Posted by mechanicalbrain
Definitely something that implements accelerometers. How about a robot that hugs a wall and uses the data from two accelerometer (or a 2 axis) to map outer boundaries of a room. You could have it even explore the inside if the room once it knows the boundaries. :D Slightly complex but it gives you a great understanding of the uses of accelerometers.

what would i implement the accerometers for?

also something i was wondering about is how i would program like a physical area or boundry into a program without using solely a variable counter with encoders, like if the boundry would show up on a control screen and where its current position it be like a dot or something, i know how i COULD do it a couple of ways, one could be just to count impulses from an entry point, or to use a rangefinder with an ir beacon to detect distance, but is there any way to plot out where the robot is in a code?

sciguy125 25-08-2005 00:42

Re: Learning robots
 
Quote:

Originally Posted by sparksandtabs
that is an awesome idea, i deffinately will try it when i get the money for 2 encoders!

If you want to be really cheap, you can use an IR light sensor and an IR LED (or a photointerrupter - same thing but both are in a neat little plastic package). Get a plastic disk, then use some sort of IR opaque material to put stripes on it. Or, you could start with an IR opaque disk, then cut slots in it.

Depending on how advanced you want to go, you might want to try out PID. If you have wheel encoders, it'd be interesting to get it to go through an obstical course (probably just ramps or sticky stuff if you don't have any other sensors) at a constant speed.

mechanicalbrain 25-08-2005 00:58

Re: Learning robots
 
The accelerometer would measure the g-force in a direction. You can run this through a simple equation and you would know the distance you have traveled in a direction (im working on this because it's more accurate than encoders). Im not that great of a programmer but you could save this and your robot would be able to travel around and make maps of a room. You actually see this type of thing on rockets but it has some cool aplications in robotics.

John Wanninger 25-08-2005 01:00

Re: Learning robots
 
Quote:

Originally Posted by sparksandtabs
that is an awesome idea, i deffinately will try it when i get the money for 2 encoders!

Supposedly Radio Shack will soon be releasing part #276-2156, 100 pulse/rev optical encoder good for up to 1020 rpm, or about 1700 pulses/sec, plugs right into VEX controller; $20 for package of 2.


All times are GMT -5. The time now is 06:16.

Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.
Copyright © Chief Delphi