English Deutsch Français Italiano Español Português 繁體中文 Bahasa Indonesia Tiếng Việt ภาษาไทย
All categories

I have an older pc which my dad's work used to use for simple automation of instruments. Now, I want to take that old purpose and take it further: I want to turn it into a real fully automated robot with moderate AI. I have random other components that would be easy to add to the robot such as a small camera, etc. What is the best way to go about this. I simply want the robot to move around independently and perform simple tasks such as picking something up. I realize object recognition isnt easy, but how would I go about even some AI. What is the best way to give motors and otherwise inedependent hardware input from the pc. Thanks!

2007-07-21 03:44:38 · 4 answers · asked by Anonymous in Computers & Internet Programming & Design

To the "top contributor": Your stance baffles me. This type of thing is not only possible but it is commercially marketed. There are packages such as the one by microsoft described in the other answer to my question, and there are hardware packages for just the idea I wish to fullfill. Thanks for the friendly input, but I believe that you are woefully wrong!

2007-07-21 03:54:42 · update #1

Im not looking for a Data, or even r2d2, just a simple robot.

2007-07-21 03:57:06 · update #2

To "contributor": Ahh, the arrogance. My friend, from your now testy comments, it is apparent that you far outclass me in this area. However, it seems you either mistake my intentions, or you lack the ability to think creatively. I recognize your prowess and apparent skill. However, while you have, no doubt, been involved in much more complicated than I, I have worked on some robotic projects too. As I said, however, they were far less ambitious than this or others you have participated in, and required moderate amounts of human input to func. In this past, my team and I set up a crude loran system for location. For depth perception, we set up two lasers (red/green), one pointing ahead, the other at an angle, on either side of the cam. When the green(R) laser was visible to the right of the red(L) laser, it was apparent a collision was imminent. I am looking for ideas on how to make the computer recognize and act on these parameters.

2007-07-21 04:44:01 · update #3

cant add more details due to length constaint...

2007-07-21 04:44:30 · update #4

Given these comments, please just give me an idea of whether or not the idea is possible.

2007-07-21 04:46:10 · update #5

Thanks for your input. It gave me some things to think about. I realize how vague my question is, I have just found that when I ask too specific a question, I tend to get no answers. My philosophy is ask a broad quetsion, get a broad answer, and hope to off on the right track. Thanks!

2007-07-21 06:25:06 · update #6

You are right in that simulated AI is the proper term. And yes, while this computer probably wouldnt run any complex cutting edge simulated AI, it does have the processing power to make some very simple decisions. Great links, thanks.

2007-07-21 12:53:33 · update #7

4 answers

Sorry mate, looks like you're in the wrong time period. Set your time machine to about a thousand years in the future, jump into the time warp, and ask again.

Edit:
At questioner: I highly doubt you have the knowledge and resources to be able to take an arbitrary PC and make a robot out of it. Sure, there are shrink-wrapped robot packages like the one marketed by Lego I think, and others. But what you’re trying to do is build up an entire system from scratch. You’ll have to write all the programming by yourself, do all the hardware work by yourself, build out the materials by yourself.

I worked on robotic teams myself. It’s complicated, and a lot of cutting edge stuff there. You realize object recognition isn’t easy? Uhh, you have no idea. Try taking a cam that can give you raw real time data. Forget about a robot. Move that cam around, and try to write an algorithm that can detect, from real time image data, whether a collision is imminent. No seriously. That’s one of the things your robot will have to do, so might as well work on this mini project.

Can you write filters for image data? How much programming experience do you even have? How much comp. sci knowledge do you have? Connecting a PC to independent parts means exploiting ports like USB ports and serial ports. Have you done any reading on how these ports work? What sending data along them is like? How good are you with a breadboard and soldering?

In fact, aside from asking this question, what effort have you made to understand what it takes to have an arbitrary autonomous robot that can move around even slightly?

Edit2:

You can do this. It will take you time though. Lots of it.

If you have experience, then the first thing you'll want to do is determine constraints and design requirements. Any programming, and electrical work has to be within these constraints. Is this a desktop or laptop PC? How will you deliver power to it? This is the first question you need to ask. If you don't have power, you can't do much of anything. You'll have to work around these power constraints then. How will you transmit data and commands to it? Wireless? Use a USB port? A normal PC can transmit data through network ports (wireless or wired) or through USB and serial ports. You could probably hack something through the audio and video ports as well. Remember, all of this draws power, so you must figure out how to deliver power first.

Motors can either be controlled through direct PC signals through a port, or they can come from an intermediate source like an FPGA. Effectively, your FPGA negotiates high level commands from the PC, and then sends the appropriate signals to the motors. This comes down to soldering type work. I don't know how big the PC is, how heavy it is, and so on. So the type of motor you need is governed by that. Controllable motors should have some sort of input, specified by the manufacturer.

I'm not being testy. Nor am I trying to put you down. Everything I've said revealed various things you should know. You will have to work on the camera/collision detection independently. You will have to test your motors independently. Your question is horribly vague. Look at what you asked. How do you go about AI? How do you control motors? I have absolutely no information, other than a vague topic.

Figure out your design first.

2007-07-21 03:49:08 · answer #1 · answered by csanon 6 · 0 0

yes there are a lot of obstacles to overcome, but if you want to there is nothing to stop from trying.
really you probably need to hook up with a local robotics club and retailer for hobby robotics. where in the US are you. I'll try and point you to some.
as for machine vision, check out 'roborealm', www.roborealm.com . its free and gets you a lot closer to the goal in one minimal outlay step.
as for AI, that is currently not possible. The state of the art is simulated AI which allows for a broad list of possible responses and weighted random decision making. That won't run on a 'older' computer. But you could certainly program a task, like search a room for an object, pick it up and return to start.

2007-07-21 13:54:10 · answer #2 · answered by Piglet O 6 · 0 0

the thing u r trying to attain is rather complex and advanced robotics will be required to attain the thing u r lookin for a lot of research is goin on over it well u can still build it up at it will cost a lot and advanced computing and hardware is required but just a mere software wont be able to do what u r lookin for ! try searchin on google about robotics i hope u will get something !! good luck

2007-07-21 11:06:17 · answer #3 · answered by roxy 2 · 0 0

I'm not sure about the hardware part. But Microsoft released a robotics studio for the software. Of course knowing Microsoft, it will crash and your robot will end up killing you. But you can read about it here

http://www.microsoft.com/presspass/press/2006/jun06/06-20MSRoboticsStudioPR.mspx

2007-07-21 10:50:23 · answer #4 · answered by WebDev 3 · 0 0

fedest.com, questions and answers