Scratch

I'm sure it is possible...let's look at the second part first.  You could have hidden sensor sprites that surround the sprite and give it information about obstacles around it.  Archmage has a project that shows how this is done here:

http://scratch.mit.edu/projects/archmage/101212


  As for walking around objects, I don't think that would be hard to do either.  Here is a very crude attempt that I threw together very quickly:

http://scratch.mit.edu/projects/Paddle2SeeFixIt/186501

I should make one of these, since they working AI seems to be a popular topic now 

Yeah, these are fun...thanks for bringing it up lestrade333!  I'm working on a more sophisticated version now that may or may not actually work.  I'll let you know if it leads to anything useful.

you're making me feel like a tinker...

Paddle2See wrote:

lestrade3333 wrote:

you're making me feel like a tinker...

I'm sorry, I'm not familiar with that saying...how does a tinker feel?

I think he means "thinker" 

lestrade3333 wrote:

no, I think it's a reference from Paganini (or someone of the sort) when he heard Brahm's third symphony, in other words, you make me feel stupid.

No need to feel stupid!  You've raised up one of the more challenging and interesting problems that you can attempt to solve with Scratch.  Problems of this caliber tend to attract the more advanced programmers.  You can use as much or as little of what they produce as you wish.  If you have questions on a particular technique, I'm sure the author would be happy to answer them.  Thanks for keeping Scratch interesting!

dbal wrote:

deatheater wrote:

I made one here    http://scratch.mit.edu/projects/deatheater/187379

Great programs are made to be broken ):

Click inside the elbow very close to the black material and the cat will go into an endless loop trying to find a way to solve the problem.

I applaud the AI's persistence.   
What if a door opened to the solution, or the obstacle moved.
The lazy AI would not get it.
Now if the AI would hurt himself or others by continuing forever- and not find food for example... then that's a bad thing.  It's just being obedient.

Last edited by AddZero (2008-06-14 12:30:15)

AddZero wrote:

dbal wrote:

deatheater wrote:

I made one here    http://scratch.mit.edu/projects/deatheater/187379

Great programs are made to be broken ):

Click inside the elbow very close to the black material and the cat will go into an endless loop trying to find a way to solve the problem.

I applaud the AI's persistence.   
What if a door opened to the solution, or the obstacle moved.
The lazy AI would not get it.
Now if the AI would hurt himself or others by continuing forever- and not find food for example... then that's a bad thing.  It's just being obedient.

No if a door opened it would choose a path to get in. it tells the cat to choose a path in, because it's so far in the cat can't reach it, if it had an opening the cat would try to get to the target anyway possible without touching the black 

lestrade3333 wrote:

If anyone is interested in an advancement of the matter, here is another question:

(first see the second demonstrative project by Paddle2See)
He shows three black balls which the cat avoids if it hits them.
He does that by a hidden sensor dish, which is also circular.
His programming takes advantage of the fact that when the two circles hit each other, the curve gradient of the two arc should just be able to make the cat look like it skirts at the edge.

But the problems are thus:
1) what if the obstacles are not circular?
2) If two circles are placed near each other, they act like a messed-up gravimetrical field (or a magnetic field, at any rate) and the cat would curve into one of the circles.

I am trying to make multiple blocks in a maze-like (but less complex) arrangement.
Oval obstacles have failed. Blocks don't work. So, maestroes of the Scratchworld, another dilemma to add to your list.

My remix get's rid of that problem 

Rurigok, that's exactly what I need, is there a link?
By the way does the project you refer to chooses the logically shortest path to the target?
If you have the link, I would be really grateful.

Here's the link to AddZero's two probe approach:

http://scratch.mit.edu/projects/AddZero/186736

i will try improving addzero's project, see what i can do