suggestions for ugrad research

Other suggestions (I don't know how hard or acceptable these would be):

aritificial "life": study behavior of ants or something simplistic, and try coming up with a small algorithm that would be adaptable to whatever's thrown at it. If you've read Michael Crichton's Prey, you'll know what I'm talking about.

object recognition: code a program that can detect humans in a picture or film, or some other object.

distance calculator: using stereo vision, calculate the distance to an object.

Those three things are what I'd be most interested in right now, if I were working on a robot.

memetics (e.g., memetic ai.)

brief introduction to memetics

anyone here been through the grad school experience and want to share?

Looking through the Peterson's guide for graduate programs, Carnegie Mellon listed under their financial aid that "all graduate students recieve full tuition + $1775/mo stipend".

holy shit, that's a lot of money.

Quote:

Original post by Adraeus memetics (e.g., memetic ai.) brief introduction to memetics

that's pretty cool. I had just been discussing with my AI professor about the fact that humans are capable of learning without experience, and that I thought that would be an integral part of truely intelligent artifice.

Quote:

Original post by capn_midnight Looking through the Peterson's guide for graduate programs, Carnegie Mellon listed under their financial aid that "all graduate students recieve full tuition + $1775/mo stipend". holy shit, that's a lot of money.

Hmm, wait. Usually that's only for the PhD students.

Quote:

Original post by capn_midnight that's pretty cool. I had just been discussing with my AI professor about the fact that humans are capable of learning without experience, and that I thought that would be an integral part of truely intelligent artifice.

Human cognition is limited to experiential education due to perception's singularity. That is true if a general definition ("participation in an event") is applied to the term experience.

I did an undergraduate Computer Science project on mesh parameterization last year (mesh parameterization == automatically assigning texture co-ordinates). It was hard work, but interesting. Some parts were very mathematical, which impressed my supervisor, and it would seem the examiners were impressed as well because I got a good mark in the end.

(by the way, if anyone's interested in mesh parameterization I could give you a few pointers)

Assuming you're studying Computer Science (implied but not stated) and you can cope with a bit of maths then I'd recommend making sure your project involves some reasonably heavy stuff, basically so that you can show off. I'm not sure all of your ideas lend themselves to this (though the component assembly one certainly does).

Having said I went to Oxford University, and the focus of Computer Science there is very mathematical. Feel free to ignore me if attitudes where you study are different. In any case, talk to your tutors / lecturers about your ideas and they'll quickly tell you which ones are potentially good projects and which ones are not.

Another bit of advice I'll offer is this - make sure you can cut back your plans if your project turns out to be more difficult than you expected, and make sure you can expand them if it turns out to be straightforward. You *will* hit unexpected difficulties, and you will have to adapt your intentions as you go.

Quote:

Original post by Adraeus Quote: Original post by capn_midnight that's pretty cool. I had just been discussing with my AI professor about the fact that humans are capable of learning without experience, and that I thought that would be an integral part of truely intelligent artifice. Human cognition is limited to experiential education due to perception's singularity. That is true if a general definition ("participation in an event") is applied to the term experience.

I think by 'experience', he meant direct experience - the kind where you experience consequences depending on your actions (pain, hunger, reward). Humans are capable of learning without this - for example I've never been shot but I know to be afraid of guns because I've learned from others. Point a loaded gun at a dog, and it couldn't care less.

IMHO, the 3D Desktop sounds like an awesome idea. Make it like a FPS view, where you can see you're files hanging around; their size in "space" would be proportionate to their size on disk. Yeah, I'd buy that...

Some sort of super-sexy 3d desktop would be very l33t. Pointless, but l33t.

Alternately, re-implement Cube/Sauerbraten's "Editable Map Geometry" system into an open engine (like Irrlicht or Ogre) - then your project would not only be neat and get you credits, but also useful to an OS community.

Quote:

Original post by capn_midnight I've heard that completing an undergrad research project significantly enhances your chances of being accepted into a graduate program (any thoughts on this?). So, I've decided to do an undergraduate research project, at least just for shits and grins, and I'm looking for ideas.

To answer the first question, it probably does help your chances (at least thats what I'm banking on), because the majority of time you spend in grad school will be spent on a research project.

As for ideas, you should probably take a look at ongoing research projects at your uni, and see if any of them stands out as particularly interesting to you.

Independent research is probably not bad either, but being involved in a large-scale project will give you a better idea of what it's really like, plus it increases the chances of faculty interaction (think recommendation letters). Plus, you can get paid a fair amount of money for doing research (more than you'd make at an average internship).

From personal experience on a large scale research project, I can say the experience is pretty rewarding; and most university research projects are necessarily pushing forward the state of the industry.

Good Luck!!

as g pointed out, I failed to qualify myself.

I'm a senior studying Computer Science with a concentration on computer graphics. I've completed 3 levels of calculus (though only got transfer credits for 2), so mathematics shouldn't be a problem. I grew up on physics.

As for the 3D desktop, the FPS idea was exactly the reason why I considered this to be a topic, in that I DID NOT want to implement it as an FPS. There are a number of other 3D desktop projects, and they all use the FPS model, and they all pretty much suck. In order for a true 3D desktop to actually work and be more efficient than 2D, it's necessary to rethink how we use 3D, and explicitly avoid that "real world" model.

If you have an concentration on computer graphics, and you want an impressive research project, stay away from anything that is just grunt work no matter how complex it may be. It needs to have a solid theoritical background, either physics or math, or something else... Research into what you plan to do, so you aren't doing exactly what someone is doing. It's unlikely you'll come up with something someone isn't doing to some degree, so just try not to repeat someone else's work [smile].

For example, 3d desktop user intefaces is grunt work, unless you want to go "into" UI design and stuff: then the goal is finding good 3d interfaces, and the coding is not that important. You could probably use something already existing or quickly code up something and do a paper on say non FPS interfaces. or something.

Destructable terrain sounds promising. Similarily maybe something in fire simulation or something like that. You could do a theory part on the theory of it, your specific approximations and so on, possibly new applications of algorithms to your specific problem, so on. And then you could have pretty graphics if you really wanted it, but that's not all that important. Further, you can then "test" your simulation on real world situations: for example, given this valley, with this rainfall and this and that, what is the result - mudslide or not?

However I'm pretty heavily biased against most graphical things, so take my thoughts with a grain of salt.

This thread might be useful: http://www.gamedev.net/community/forums/topic.asp?topic_id=258256&whichpage=1�

That's an interesting idea, the fire thing. We've seen destructable terrain in games like Red Faction, but it's been of the type "Hit with bomb, reform with explosion shaped chunk, change texture to burnt texture." With fire, there is an opportunity to do something different, to model actual consumption.

Also, another problem with the Red Faction model, the blown away terrain just dissappears, there is no conservation of mass.

I wonder if there is currently any work looking at cutting objects, like a knife through bread or a shovel through dirt. In this case, you're not actually destroying the terrain as much as you are reforming it.

There's been a lot of research into 3D desktops/file systems. There's also been a good amount of work done on deformable terrain. Does your research need to be original or not?

Your idea on the deformable terrain side seems to be moving away from the graphics side, towards the physics side. Have a look into fragmentation meshing - it may be the sort of thing you are looking for.

Major projects I did at an undergrad level:
Procedural Libraries for Cartoon Animation (group proj) - it generated stylised effects for use with hand drawn animation (fire, rain, etc) rather than just looping a sort anim.
Volumetric Mesh Generation (individual) - it took scanline images of slices of the human body & created a volumetric mesh with them, which could then be run through a FEM.

Post grad:
Assassin (group) - 3rd person game written using RenderWare graphics.

And I'm currently meant to be writting a thesis on Procedural Weather Generation for Games in my spare time, but I don't have much of that at the moment.

capn-midnight -- have you played both Worms 2 and Worms 3D?

If not, you should. It's interesting to note the differences in environment destruction between the 2 games.


Oh, and Worms 2 rocks.

I'm fairly certain it does not have to be original. I had a friend that basically did a report on network intrusion detection systems last year (though that might have just been an independent study credit).

Undergrad research is rarely, if ever, original. The only time you really have to do something original is if your'e writing a PhD. thesis. In fact, there are plenty of Master's students who perform research in well-understood areas... so it's pretty unlikely that the constraints on UGrad research would be more severe than for most graduate students.