I need help with circle to circle collision response

[MrPenguin9]

I need help with circle to circle collision response. I have two circles, both moving. I know how to find if they are colliding (something like this)

float dist = sqrt(((circle1.pos.x - circle2.pos.x) * (circle1.pos.x - circle2.pos.x)) + ((circle1.pos.y - circle2.pos.y) * (circle1.pos.y - circle2.pos.y)) );
if(dist < 64){
//do collision
}

But I have no idea how to do the response to the collision. All I know is the position of the circles, (circle1.pos.x, circle1.pos.y, circle2.pos.x, circle2.pos.y) velocity of the circles, (circle1.vel.x, circle1.vel.y, circle2.vel.x, circle2.vel.y) and the diameter of both of the circles is 64. (radius is 32)

Thanks!

[JasonR]

I googled "Elastic Collision Game Programming" and came up with several references on how to handle the collision, as long as you are assuming the circles do not deform during the collision.

[ClarityGames]

Hi Mr. Penguin. I've done something like this before and have a couple of pointers.

1. You can do away with the square root function.
Square root functions are notoriously slow and cpu intensive, and should be avoided. On the face of it, though, it's an essential last step in calculating the distance between two points. But if you take advantage of the fact that if

a = squrt(b)

then

a squared = b

you can transform your CD calculation to

float dist = ((circle1.pos.x - circle2.pos.x) * (circle1.pos.x - circle2.pos.x)) + ((circle1.pos.y - circle2.pos.y) * (circle1.pos.y - circle2.pos.y));
if(dist < 4096){
//do collision
}

4096 being 64 squared. Try it, it works!

2. The collision response for two circles colliding can be based on the vector between the two centers. Basically the vector created between circle1 and circle2 represents the line of kinetic force that circle1 applies to circle2, so the negative of it would represent the reactive 'push' of the collision.

e.g. if circle1 - circle2 = (4,4) then applying a force on circle1 of (-4,-4) would be a realistic response.

Of course the 'applying the force' bit is the tricky part of the deal and I'm a little hazy on that side of things, but I'm pretty sure that the whole process can be handled by vector arithmetic.

You could try something like this as a simple start:-

Circle1VecX = (circle2.pos.x - circle1.pos.x)/64

and then the same for Y to see the circles respond with the correct lines of force away from each other. This won't give you the correct direction or speed of response but it'll at least show the basis of a proper 'reaction', i.e. the two circles should always respond by heading away from each other in exactly opposite directions. After that it's just a matter of applying the correct magnitude to the reactive force and adding it to the original vector of the circle and volia! Realistic response!

[MrPenguin9]

Originally Posted by ClarityGames

Hi Mr. Penguin. I've done something like this before and have a couple of pointers.

1. You can do away with the square root function.
Square root functions are notoriously slow and cpu intensive, and should be avoided. On the face of it, though, it's an essential last step in calculating the distance between two points. But if you take advantage of the fact that if

a = squrt(b)

then

a squared = b

you can transform your CD calculation to

float dist = ((circle1.pos.x - circle2.pos.x) * (circle1.pos.x - circle2.pos.x)) + ((circle1.pos.y - circle2.pos.y) * (circle1.pos.y - circle2.pos.y));
if(dist < 4096){
//do collision
}

4096 being 64 squared. Try it, it works!

2. The collision response for two circles colliding can be based on the vector between the two centers. Basically the vector created between circle1 and circle2 represents the line of kinetic force that circle1 applies to circle2, so the negative of it would represent the reactive 'push' of the collision.

e.g. if circle1 - circle2 = (4,4) then applying a force on circle1 of (-4,-4) would be a realistic response.

Of course the 'applying the force' bit is the tricky part of the deal and I'm a little hazy on that side of things, but I'm pretty sure that the whole process can be handled by vector arithmetic.

You could try something like this as a simple start:-

Circle1VecX = (circle2.pos.x - circle1.pos.x)/64

and then the same for Y to see the circles respond with the correct lines of force away from each other. This won't give you the correct direction or speed of response but it'll at least show the basis of a proper 'reaction', i.e. the two circles should always respond by heading away from each other in exactly opposite directions. After that it's just a matter of applying the correct magnitude to the reactive force and adding it to the original vector of the circle and volia! Realistic response!

Great idea! Thank you!

Originally Posted by JasonR

I googled "Elastic Collision Game Programming" and came up with several references on how to handle the collision, as long as you are assuming the circles do not deform during the collision.

Thank you so much! I was googleing "Circle to circle collision response" and I couldn't find anything, but "Elastic Collision Game Programming" work perfectly!

Thank you both very much!