Relativistic engine advice

[dontforget86]

Hi guys

Recently I've started working on a very lightweight engine for simulating motion and (hopefully) accurate collisions where bodies are moving at relativistic speeds. So far I've put together an RK4 integrator which incorporates the Lorentz factor into the timestep and this seems to be working well. The engine is being tailor made for a rather complex space strategy game in the vein of the old table top games from the 80s like Orion. Of course, everything is taken from the inertial frame of the player.

I'm looking for advice about collision detection; obviously this is an extreme case beyond the needs of most game physics engines, where distances and velocities are enormous and collisions are relatively infrequent. It occurred to me that something as tiny as a spacecraft/kinetic weapon is probably only going to collide with another spaceship if it's intentional. If you're around Julio I would definitely like to hear what you think- which approaches might work or the ones to dismiss immediately. I would also be interested in hearing your thoughts on large scale scenes and calculation precision. I'm intending on taking this whole thing to OpenCL once I have a working model but I'm worried about losing precision working on the GPU.

And if anyone knows of any projects that have implemented something like this already or any material that might be useful then please let me know.

Thanks a lot, any advice or thoughts at all are greatly appreciated.

[Stucuk]

This forum is ment to be about Newton Game Dynamics, its not a General "How do i make a physics engine" forum.

[Julio Jerez]

It has being more than 20 years since I took my last class in Classical Mechanics were I had homework using Lorentz transform and Hamiltonians.
And we did not even have computers then, everything was very theorical and analytical.
The only thing I can tell you is that televion and science fiction shows confuse people into make then think that
the Equivalence principle applied to non sub atomic particles, as if thecnology one day will make it possible traveling close to the speed of light
as if speed of light was just a barrier to break similar to speed of sound barrier, and that is not possible.

I am afraid you are out of luck with Newton if you want to use it for Analytical Mechanics; Hey Newton is not even good for Realistic Simulation that requires high degree of accuracy.
To give you an idea, if an object is traveling at one thousand of a percent of the speed of light,
that will be 300 kilometers per second, that's about 500 times the speed of sound, faster than any known object to man.
if you simulate that in a computer taking descrete time steps at 60 fps, in one time step it will move around 6 kilometers.
Believe me at that level, Lorentz Transform is the least of your problems. You have much more immidiate problems to worrie about, like float accuracy, collision and other things,
inertia of the passengers, navigation and centripelal forces generated by chnging direction, and so on.
and you are still not even going fast enought for Relativite to make any effect, as for relativity to take any significant effect you need to go at least 10 time faster than that.
I am sorry but I believe this is not the right forum for you to get an answer.
Maybe some of the Open Source Physics engines can help you a lot more that I can. I have enough with mundane Newtonian mechanics.

[dontforget86]

Float accuracy and a possible way to detect future collisions were the things I was asking about. I'm aware this will be difficult, perhaps too difficult to do in real time. If it is then I'll make it turn based, calculating a few seconds of physics ahead each turn.

I still welcome any thoughts and ideas.

Thanks for your post, Julio.