Friday, September 19th, 2008

Looks like Stanford has their own version of MIT OCW with some neat courses up! It’s called Stanford Engineering Everywhere.

I’m just looking at the Machine Learning course now… it looks pretty fun - definitely worth checking out
SEE

Wednesday, September 17th, 2008

Recently I’ve been planning a few things that I want to post on my blog, and some other sites. Check out the list!

• Probability Intro & Bayes Theorem Tutorial
• Bayes Filter Algorithm Tutorial
• Bayes Filter Algorithm Example
• BubbleBoy Tutorial

I could extrapolate this idea into a special bloggy tutorial week where I post something each day!
Also in the works is my line following robot, and some acoustic modeling stuff. I’m pretty sure my method for the acoustic modeling will work well. Basically it will analyze the time where the waveform hits a 0, minimizes and maximizes. I’ll apply some sort of filter algorithm (not sure which one(s) yet) and then compare that result with a database of results, complete with the other mins and maxs. I can’t really explain it well yet, as it’s still in my head. After that I’ll have to implement the language model, which will choose the right word from the results of the acoustic model based on the previous words uttered and the topic!

From what I’ve seen… some recognizers are trying to skip the language model step by listening for specific phonemes in the acoustic model to try to figure out what the word is. Then, the LM’s job is to just ensure that that word makes sense in the sentence.

That process is overly complicated, all you have to hear are the words that are uttered- not all the phonemes. It is true that we don’t pronounce words the same… (nu-clee-ar vs nu-cu-lar) but essentially if one has an accent, the maxs and mins should still be comparable to the word in a “perfect” sense. (This is where the various filters and algs come in)

So yeah…

Thanks to Ron for donating $200!! It’s going to go towards a serial servo controller, a soldering iron and other things I’m so humbled by their generosity
It will definitely keep me occupied while I wait to go to university in January!

Friday, September 12th, 2008

FUDGE!
BubbleBoy did not win, nor PLACE in crabfu’s ‘Most Characterful’ competition.
I don’t understand how it didn’t… My source and schematics are even open!!
This extremely sucks as I wanted to win so badly as I don’t have the money to buy a serial servo controller Fudge!! Will MANOI forever be in a box? Talk about unethical treatment to a robot… (You wouldn’t keep a cat in a box* for a month would you?! why would a robot be any different?!)
* Schrodinger’s cat is an exception

I guess I poured too much creativity and florescent pink and lime green paint into it Grr! Either that, or it doesn’t even count as a robot since it doesn’t show its wires.
Although it is only “one contest” it was only an EXTREMELY GIGANTICOR SUPER DUPER CONTEST that I really wanted to win so I can work on my other robot so I can win other contests so I can pay for parts to keep researching robotics!!!!
sigh…

Friday, September 12th, 2008

My Friday night robotics are BACK! Throughout the week I was working on speech recognition… Friday I took a bit of a break and soldered up the Motor Shield!
The Motor Shield is pretty awesome since it has 2 H-bridges. Soldering it up was pretty fun- even with a melted soldering iron!

Once it was all soldered up… I first tried the servos. Their motion is so much stronger than the servos I’m using in BubbleBoy! I’m going to try to use ServoTimer library on BubbleBoy again (previous attempts have been extremely frustrating). Maybe he’ll be more “emo”

The DC motors scared me as I extremely didn’t expect them to work the first time! Which is sort of ironic because, even though they did work, I didn’t have them plugged in right, so they didn’t move backwards. The solution was to not over-think it, and just plug it in (Like below)

The motors went to the GearBox, externally powered with +3V. The batteries were slowly fading away, so it became kinda fun to spin the gears

I made a video of my fingers spinning the gears (rofl):

So yeah… this is great! I’m probably going to make a shield for a RBBB to attach the motor sheild. Or something like that. I need to get one of those rechargable batteries wrapped in lime green shrink wrap as all of the motors will be powered. (There’s 4 of them!) It’s going to be a really powerful robot O_O

Oh yeah! I’m not sure what I did to my camera *coughdroppeditcough* but it takes REALLY nice pictures now… it does the effect that I’ve been trying to achieve on macro for ages now- softens the background a lot, crisp in the front (you can see it in the soldering iron pic). Yay!!

Wednesday, September 10th, 2008

Today is a pretty happy day! It’s sunny out, and it’s my birthday! I didn’t get any presents, but that’s okay I did something that I wanted to do for a while now… I still haven’t finished blogging all of my pics from the Stanford EPGY SI AI trip!

The last place I went to while in California was the JPL! It was really amazing! Our tourer works on Deep Space. Uber cool!

Click continue reading to see the pics
More

Monday, September 8th, 2008

Even more BubbleBoy awesomeness

And here’s the circuit diagram!

I hope I entered crabfu’s most characterful robot contest correctly. The deadline said: 09/08/08 at midnight — but it almost sounds as if it’s already closed and so now I’m so worried that I didn’t even enter and I won’t even be considered so therefore I wouldn’t stand a chance :’( I HATE TIMEZONES!!! :( Let’s just all hope that it is entered in the contest…
Update: Everything is okay. I’m entered in the contest!

Thursday, September 4th, 2008

Yay! BubbleBoy is finished (version 1). Briefly, BubbleBoy was created to demonstrate emotions and their applications in robotics. BubbleBoy has a food and water level, and when averaged out, a health level. The mood is determined by the health level. There are 10 moods, each of which BubbleBoy has a little ‘action set’ with. This is to try to get more food! After the ‘action set’ is over, you can feed it or give it water. When the food/water button is pressed, it increases the food/water level by a random int in between a lower bound and upper bound. Also, when BubbleBoy finishes three ‘action sets’, it becomes more hungry and thirsty - thereby decreasing the food/water levels. So, BubbleBoy is similar to a ‘tamigotchi’. It was fun to make- BB has 2 servos, 2 DC motors, 2 buttons, 4 LEDs, 1 LDR, and 1 LCD. All of these are controlled by the Arduino board.

The source code is here! There are no comments in the code yet (they’re all in my brain at the moment, haha).

Although I said (on Wednesday) that this was going to be posted tomorrow… it’s actually early Friday. BUT, since I haven’t slept yet, it’s still Thursday to me :] I had so much trouble getting the footage from the tape recorder onto the mac! It seemed to be like technical difficulty after technical difficulty the entire time. I’ll post more details tomorrow about the schematics… to be honest I forgot if I used pull up resistors or not! So I have to check…

All the awesome pictures! The plastic was painted green, and the push buttons were created

There’s BubbleBoy!
The future versions will employ ‘expectations’. For instance, when BB is fed, it will record the LDR value. If after three action sets and the LDR value is the same… it will expect to be fed! If it’s not fed, then it will become more hungry or thirsty since it was somewhat ‘disappointed’ because it wasn’t fed.
Schematics and fancy vocabularized blurb tomorrow

Wednesday, September 3rd, 2008

Tomorrow I’ll be finishing off BubbleBoy, posting the schematics, the code, pictures, and videos! Woohoo!

I might even write a fancy shmancy few paragraphs about how scientific it is.

Monday, September 1st, 2008

Recently I’ve been thinking about swarm robotics and the potential they could have on some of our other planets. Do we really know everything there is to know about the Moon? We have lots of data… but are there underlying details that are still hidden?

Instead of spending trillions of dollars on a gigantor research project, a few billion dollars could send some robots to the moon. The robots could have the ability to drill, evaluate the dirt, and ‘laser beam’ rocks like the MER. It will also have cameras on it I’m kind of envisioning about… 2.5 times the size of Sojourner.

About 12 of these robots could be placed on the Moon. When one of the robots find results that are ‘interesting’, a few robots that are close to it will become closer, evaluating where else the interesting results lie. Each of the 12 robots could have different ‘accessories’ on it too, so that they’re all not evaluating the same thing.

Course, it could also be applied to other planets- like Mars, or maybe they could be sent out to Triton. That would be kinda cool!

It would be simple, and autonomous. The results obtained could potentially trigger new information, which could lead to a more precise robot, which could lead to new discoveries.
Sounds like I have even more research to do!

Friday, August 29th, 2008

Dormant blogging kills the robot star
BubbleBoy is doing good- and looks EXACTLY like a pink snowman with the scarf, which is what I was aiming for.
The coding is going good… it’s all really simple Check out this video of cycling the health levels. For health >= 90 there is an action, but the others there isn’t yet because I haven’t programmed it yet teeheehee.

The video is really slow… as the delay between cycles was 10 seconds At the end you’ll see BubbleBoy’s ‘behaviour’ for health >= 90.