Programming Balls #2 Circles V Edges Collisions C++ 

Hi JD DK, thanks buddy, Its a great pleasure to hear the videos being appreciated!

Thanks Dankbender! :D

Thanks Daniel - I will!

Hi Chad, thanks so much. It means a lot to hear that!

No problem! Its a simple simulation but useful for platform games.

Yes! They look like conveyor belts. I used to love that game!

Hi iii, Thanks man! I believe people who are privileged/fortunate enough to have been educated well, are obliged to educate and help others.

It is!

Thanks John!

Thanks leapingbear! I will likely do a video on multi-threading soon (and you've put using balls as an example in my head now), but probably not using the GPU - It's a fascinating topic, but difficult to put into context in a single video, since I try to show everything.

javidx9 hehe, glad to do a bit of brain worm job there. Well, Any video on multi threads would be awesome, so i will look forward to that!!!

The equation he is using preserves momentum. Gravity is adding momentum to the system even with ground added you will get constant bounce without gravity nothing will come to a stop on the surfaces. What you need is a inelastic collision rule. Id say off the top of my head when you check for collisions between balls and the walls or ground ect. You should probably check the relative velocity's between the balls to each other if it is a very small difference and they are very close only apply the static check then set both ball or particles velocitys to be equal to each other. Ignore the bounce effect application. A more real inelastic collision is done by using momentum two objects combining into one (or sticking together will both have the velocity after collision for ball1 and ball2 = (v1 * m1 + v2 * m2) / (m1 + m2) so for example if you have ball1 moving with 1000 mass at 9 pixels per second = 9000 momentum + ball2 with mass 2000 at 0 pixels per second then 2000 * 0 = 0 so 9000 = 9000+0 then 3 pixels per second is found by 9000 total momentum divided by 3000 total mass and both objects will have that total speed in the direction of travel.

I was wondering the same thing. I think that it's because it gets normalized by dividing by the same value in the next line of code to get it between 0 and 1. If you have two vectors that start at (0,0) and your edge goes to (6,0) and your ball is at (3,2), the dot product is (6*3)+(0*2) = 18. Dividing by the line length (6) gives you (3), the point that the ball casts onto the edge, but dividing by 6^2 or 36 gives a normalized value of the total length of the line, (0.5). Hope that made sense.

Hi Joe, this is a good idea - I would like to do a video about angular velocity and momentum probably starting with balls, but moving to convex polygons.

@@javidx9 thanks that would be so helpful. I'd finally be able to fix the rotation issues I have.

I suppose a 2d roller coaster should be quite possible yes.

You may need to have many iterations of collision if your dt is too large. At all costs, you must avoid the solid bodies overlapping each other. This may take several passes of the static collision part first. Also double check your displacement, ideally you want to displace balls so they just about touch. Ive seen when the simulation gets out of control, but consider that this is the result of positive feedback in your simulation, i.e. there must be something there that makes it worse each frame.

Yeh, I found a better solution which was to do exactly as you said to iterate collision many times. However I found the main problem which was that when they pile on top of each other, the bottom ones keep intersecting despite the amount of iterations of the collsion method i put in. The bottom ones also seem to sink through the floor and pop back into the screen. All methods work if the balls are not on top of each other.

lol, thanks, and congratulations! you win the award for fastest youtube comment ever!

Dammit! Don’t you have to watch the vid first? :D

i dont get it. Is it nice or noise

For the most part, it is simply adding a third component Z. Detecting collisions between spheres is easy enough, but the deflection angle can take a bit of thinking about

Setting vx and vy to 0 when creating the fake ball should solve it.

Is this a good comment? or a bad comment? :D

In University, you are master of Physic?

Hey Huy võ lê thanks, but no. The physics here are pretty basic, turning the physics into code is harder. I like that obeying the simple laws of physics yields such a rich and complex simulation in this instance.

Hi John, Cheers! AI is a big topic. I think I would need an AI project over several videos to even explain the basics - I actually have a background in AI so maybe I should do something...

Okay, Thanks!

Your videos have opened my eyes to so many problems I've had programming. Especially with Audio and Video Games.

Im self taught. Started when i was 9. Not many mentors around then. Lol 😂

@@javidx9 i figured awesome.

That's a really big question. There's a reason GTA takes 3-4 years development with a team of 100+ people and a budget of many millions of dollars. If we scale your question back a bit, say "How do we make a 3D game?", that's a lot more achievable. I will be making a video about 3D this year, but ultimately you need to understand vectors and matrices, some physics, and to make it look really pretty, some image processing. In short, 3D games require knowledge in several disciplines, before you even start writing code.

Hi Nolan, i dont have yet but it is coming very soon. Not next video but maybe one after.

Hi akarsh, dont forget that ball vs ball check is incredibly simple to do, so in this unique scenario its quite effective. There are as you mention spatial partitioning systems, but the overhead of searching may work against you for a simple setup such as this.

javidx9 Alright, thanks for the great videos and insight you have. Please do upload/ recommend a video that utilizes quad-tree, spatial HashMap implementations etc so that we can almost make our own box2D

No worries. Quad trees are definitely coming up!

If i wrapped up a vector in a type, I'd still need to interact with the type? I think I dont understand. :(

I mean, math vector, not std::vector. You could have reused some code then, instead of writing those long lines for distance computation, when you could just write distance(ball.p, target.p);

You're correct, I should wrap up a vector, and would do if I was working on a project, (i like glm as it happens). But I know there will be people watching this video that may benefit from seeing how the maths is implemented, and how it interacts with the algorithms.

its less to type and achieves the same goal?

K Parks Non-warning: please do it high haha.

ага)))