Friday, July 17, 2009

One Step Forward

In regards to the previous post, I have settled on a leg/body design for my little robot "thought exercise," (since to be considered a project, you need to actually pump some effort, thought, and money into it.) Don't get me wrong, I REALLY want a pet that I don't have to feed, fear, or actually put any effort into, but the raw cost alone will be a while of saving up, and even after that, I have to put a lot of hours into learning how to make it function.

At any rate, I know which style of legs and body I wish to implement. I have decided to let go of the hamsterball idea, because I theorize that the spherical balance process will eat to much processing power to move as well. It's kind of like a Pogo-robo, I think. One point of contact with the ground, and all 5 pounds of weight precariously balanced above it. I can't expect myself to write a logarithm to balance and move. I can barely remember my trigonometry! Any thoughts or logic against this decision would be lovingly accepted and possibly implemented.

I was browsing the internet, looking for "How to build a robot" resources, when I came across a company called Lynxmotion. They are the only people I have come across so far that offer an Erector set of servos. You can't just use motors, you know. You need a motor that can go on command and stop when you need it to do so. Automatically. Anyway, Lynxmotion's hexapod sets are beautiful and pretty much exactly what I am thinking about now. You mount a Eee PC on top of it and pray for the programming sense to make it a little autonomus, and off you go! Robopet!

On an actual thought progress note, I had an idea for an add-on to the robot idea using computer science that I actually understand. Making a robot play fetch. Half-Life 2 made an NPC do it in game, why couldn't we do the same with a physical being?

Factors in mind are:
  • Item recognition (easily embraced with specific colors and good lighting)
  • Item tracking (stereo cameras and blotch tracking)
  • Item retrieval (a grabbing mechanism)
  • Item return (simply making a log of movements and performing them in reverse should do it, I think)
What I understand about electric cameras is that codecs work mostly on recording changing colors. I saw a video of a a paintball sentry gun that worked mostly on that principle. It would track, center on, and fire upon the largest moving splotch in the still image. It should be feasible to use this to find a moving splotch of specific color, and then, well, march toward it. Grabbing would be hard to program, and I still don't know jack about programming. I can turn a wrench and read though. Let's see where that gets me. Motivational video. It would seem a powerful processor isn't necessary for the motion. Maybe for the thinking, tracking, and fetching though. I don't know, this is all way too cool. To quote Freeman's Mind, "Robots are the only friends I need!"

On a related note: XKCD though it may be, this project could take me YEARS!

Sorry for the random spattering of words here. I've been thinking way too much about this. Hope you all are well.

-The0

No comments:

Post a Comment