Falling through

I looked at that tutorial too, but I didn't have any problems with the collision detection. I don't have a very fast computer or a good graphics card - which the site said were two of the reasons that the tutorial was not working on some machines. I believe it has something to do with framerate that the program is running at.

Did you see this post on the front page of gametutorials.com:

Quote:

Quake3 Collision 3 Tutorial Fixed (AGAIN) We just fixed the Quake3 Collision tutorial again, since the last fix didn't work for everyone. PLEASE contact me if this is not fixed and you keep falling through the floor immediately. Apparently, many people with Radeon cards were experiencing a bug where they fell straight through the floor right away on the last collision tutorial for Quake3. (For those of you wondering what I did, I just changed camera.cpp and changed kSpeed to 200 from 300 to compensate for the 0.005f that was added to g_FrameInterval in CalculateFrameRate() to fix the floating point error/imprecision.)

Its second from bottom (back from Nov. 6), so I thought that you may not have seen it yet. I don't know if this would help you, but I thought I would show it to you if you haven't seen it yet.

A lot of falling through problems come from the check for an intersection with a plane/triangle where it is quite regular to see collision code where the collision is only tested for an actual collision. Example being - The sphere now collides with a triangle so lets react. Less often seen and implemented is the check if the sphere crossed the triangle without touching it.

ie. frame 1 sphere is in front of the triangle, frame 2 the sphere is behind the triangle. At no stage did the sphere touch the triangle as the distance travelled was to great. This is quite often missed with high frame rates as the sphere rarely travels far enough in a single frame for this to occur.

I dont know if this is the issue you are talking about, but it is quite possible.

Hey
thanks for the info so far.
I had a look at the news about the fix and changed the kSpeed. I'm still having a fall through problem. My rig is AthlonXP 2800+ and also have a radeon 9800 pro...So I think I might have to fiddle some variables.
I think its the same problem that Lubby was talking about.

If anyone else has found a fix...?

Thanks so far however.

I've not looked at that specific tutorial, but I doubt its the same issue that lubby is talking about. That happens with discrete testing of position, where as most BSP collision tutorials i've seen use ray casting and as such include the distance travelled.

What is more likely though is you are hitting precision issues, where its possible for intersections to be missed. These often get worse the higher the framerate as the distance travelled per frame becomes less, thus increasing precision problems.

Try adding extra tolerance to collision checks, and always back away slightly from the collision point (before calculating collision responce) that should help.

Also you might have more luck asking this in the maths and physics forum

I had a similar problem, when the framerate is very high. Try methods to cap the frame rate to 60 Hz, to limit the FP innacuracies.

That sounds like a good idea- capping the frame rate. Also if I add in a few more thresholds on the collision detection then it will make it less likely to fail.

Thanks for all the help. I'll put in some capping code now