Timing for Physics vs Rendering
Author
I want to maintain consistent physics as frame rate changes. In my current game attempt, movement of objects is scaled by time between frames, but the physics still change as frame rate changes.
How do I base physics on a constant rate, while frame rate is changing?
Thanks.
You can force you physics into a minimal time step by looping as needed based on the elapsed time since the last physics calculation. ie, if if .1 seconds passed and you want your steps to be no larger than .01 seconds you would loop 10 times with a .01 second step. Any remainder can be used as the final step because it will be under the minimal time step requirements. This is not the best way but it's a good hack.
This is a great article about how to sync your frame time.
http://www.gaffer.org/articles/Timestep.html
Hope that helps,
Oscar
http://www.gaffer.org/articles/Timestep.html
Hope that helps,
Oscar
Author
Thanks iequalshane and others.
It's been a while, but this morning I implemented the physics timing into my game according to iequalshane's comments. It works perfectly, and my code turned out to be very much like the article from the other posts.
Is this method really considered a hack? If so, then what would be the "sophisticated" way to go about it?
Thanks all.
It's been a while, but this morning I implemented the physics timing into my game according to iequalshane's comments. It works perfectly, and my code turned out to be very much like the article from the other posts.
Is this method really considered a hack? If so, then what would be the "sophisticated" way to go about it?
Thanks all.
Quote:Original post by Vectorg
Is this method really considered a hack? If so, then what would be the "sophisticated" way to go about it?
I guess the part where the leftover time is not accumulated is what he is calling a hack.
The problem is likely that you're using a euler integrator, which isn't very accurate. The common solution for the simple physics used in games is to use something like a runge-kutta integrator. RK4 would probably fit your needs.
Runge Kutta on mathworld
Runge Kutta on wikipedia
If you want the physics to be entirely framerate independant (usually not required for games), you'll need to use a fixed physics timestep. With RK4, though, you'll be able to use a larger fixed timestep which means you can run the physics simulation less often.
Also, instead of doing as previously suggested and running the physics until your timestep is 0, you should do something like
Runge Kutta on mathworld
Runge Kutta on wikipedia
If you want the physics to be entirely framerate independant (usually not required for games), you'll need to use a fixed physics timestep. With RK4, though, you'll be able to use a larger fixed timestep which means you can run the physics simulation less often.
Also, instead of doing as previously suggested and running the physics until your timestep is 0, you should do something like
while(DeltaTime >= MIN_TIMESTEP){ DoPhysics(MIN_TIMESTEP); DeltaTime -= MIN_TIMESTEP;}/*...*/DeltaTime += ElapsedTime();
Author
Yes, Extrarius, my code is functionally identical to your sample. My physics now remain constant for any reasonable frame rate.
Euler integrators and runge-kutta? These topics are beyond my current game application right now, but thanks for giving me even more stuff to study... and I have finals coming up.
Thanks.
Euler integrators and runge-kutta? These topics are beyond my current game application right now, but thanks for giving me even more stuff to study... and I have finals coming up.
Thanks.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
