Hi everybody, i am starting work on very interesting part of my little engine. I am going to implement collisions of solids and would like to know your opinion on this problem.I am not sure that my way is the right way;my way is following:every solids consist of set of mass points and in every loop i compute following main tasks for every solid: 1.For every mass point in solid i find forces which i apply on it. 2.i compute the resultant force(valid for decision point) and resultant torque taken to decision point.decision point is in(0,0,0) 3.from force and torque i compute new position and orientation of system where solid is located this described steps working without problems and i think i won´t have to change it. Problem is where and how to find reaction FORCE for every mass point which collide with some object and how to determine its size and direction. I will give you an example: I have solid,dumbell for example,which consist of two mass points.it is located somewhere in world scene and i "turn on" gravity,so the dumbell start fall.In a few moments the left part of dumbell begin collide with the plane of table. What i have to do now?Have i compute reaction force,which will have the same direction like normal of plane of table and which size will depend on distance from plane?(and then process steps 123) Or is there some other(better) way to do it? Thanks for replies P.S.:Try to use simple sentences,my english skills are not so goodd...

IT WORKS,impuls forces must depend on velocity too...
I am happy now and I would like to know if some comparing test of physics engines exist.I would like to compare mine with some havok,or what is its name
So if anybody know about some setting of problem,let me know ,please.
Thanks Alex

There was an article on Gamasutra, comparing Havok and MathEngine, with many different tests. I''m sure that you can find it.

Otherwise, to find the "classical" way of doing collisions, Google for "Baraff", and click on the first link that comes up.

Cédric

thanks,
have you any tips on "non-classical" ways of doing collisions?
I can''t find it,everything what i found is done via "coefficent of restitution e" :-((