Simple Circle physics

In case of circles, you do not have to include torque, it's just a case of getting velocities right.

The key is to decompose velocities with regards to collision normal(¹).
V = Vo + Vp;
where Vo is orthogonal to collison normal and Vp is the rest (abs(dot(Vp,normal))==length(Vp)) - Vp is perpendicular to the normal vector.
In case of circles, the normal is always normalised difference of circles' centers.

Then the orthogonal component can be left alone, and you deal only with the perpendicular components - and the problem becomes 1-dimensional!
In case of bodies having same mass, all you have to do is to swap 1D velocities.

So, having:
N - collision normal.
V1 = V1o + V1p;
V2 = V2o + V2p;
We get after the collsision:
V1' = V1o + V2p;
V2' = V2o + V1p;

---

(¹)
Decomposision proceeds as follows:
Having:
V,N
we calculate:
Vo = V - N*dot(V,N)
Vp = V - Vo

Thanks a lot! Will try to implement it today and we will see how it goes :)

If your problem becomes a little more complex (i.e. different masses), you might want to look at my post on a previous thread about this topic. For objects of equal mass it simplifies to the simple component-wise reduction that Deffer gives.