I'll let the replay speak for itself. One lap at the dragstrip, not long. Hit sidewall, spin like a top for 15 seconds. I know this is common, but this one is pretty severe.

We have already seen worse ones, like the one on the oval, where the car kept spinning for some minutes

It's the nature of realtime physics that when you feed in extreme circumstances (collisions) outside of the normal operation of the physics engine (racing) that it just can't cope with the values involved because you are suddenly operating outside of normal parameters.

To solve this LFS would need to massively increase the number of physics calculations per car which would raise the CPU requirements of the simulation immensely, and you can't have an option that "turns this on for high spec computers" because the car needs to process the same on all machines playing the game.

The solution is twofold. Firstly LFS handles collision intensity by measuring the overlap of the 3d objects, this doesn't work well because if you travel at 20mph and are 19 'mph' away from an object, the next phase would see you overlapping by 1mph. If however you are 1 'mph' away from an object the next phase you would overlap by 19mph... An impact 19 times as intense despite you travelling at 20mph in both examples.

The example above is quite extreme because actually LFS is dealing in fractions of a mph per physics cycle - but when you consider the force of a hard impact it's easy to see how such a system just falls apart at the knees and how two collisions at the same speed can have massively different results.

This collision system obviously needs to be done away with for one that works on the speed and mass of objects, but it is infinately more complex to integrate into the car physics system - so either we have to wait a long time for it, or LFS needs to run a parallel physics system that comes into pay when a collision is detected.

Quote from Becky Rose :

It's the nature of realtime physics that when you feed in extreme circumstances (collisions) outside of the normal operation of the physics engine (racing) that it just can't cope with the values involved because you are suddenly operating outside of normal parameters.

To solve this LFS would need to massively increase the number of physics calculations per car which would raise the CPU requirements of the simulation immensely, and you can't have an option that "turns this on for high spec computers" because the car needs to process the same on all machines playing the game.

The solution is twofold. Firstly LFS handles collision intensity by measuring the overlap of the 3d objects, this doesn't work well because if you travel at 20mph and are 19 'mph' away from an object, the next phase would see you overlapping by 1mph. If however you are 1 'mph' away from an object the next phase you would overlap by 19mph... An impact 19 times as intense despite you travelling at 20mph in both examples.

The example above is quite extreme because actually LFS is dealing in fractions of a mph per physics cycle - but when you consider the force of a hard impact it's easy to see how such a system just falls apart at the knees and how two collisions at the same speed can have massively different results.

This collision system obviously needs to be done away with for one that works on the speed and mass of objects, but it is infinately more complex to integrate into the car physics system - so either we have to wait a long time for it, or LFS needs to run a parallel physics system that comes into pay when a collision is detected.

How did you get to know all this stuff Becks?

Mostly it's common sense, on top of which a bit of reading.

You can tell she's not an engineer though

Quote :

You can tell she's not an engineer though

Nope, engineers are below me, i'm a [n underpaid] senior research & support technician .

Quote from Becky Rose :

Nope, engineers are below me, i'm a [n underpaid] senior research & support technician .

Lucky engineers!

I think this replay is a wonderful example how great the physik engine works (of course except the crash). After the crash the cars has a very high angular momentum and you can see how the car finds a stable axis by minimizing the torque of inertia. (first it rotats around the z axis then around the x axis). Nutation! Then it acts like a gyroscope. Precession! You can see the angular momentum preservation.

For me this is one of the worse flaws of the physics engine by far, whatever is causing it. Its just unbelievably unreal (and annoying) to end up like this just because the side of the car scrapped the side of the wall

Quote from Gentlefoot :

Lucky engineers!

i.m.h.o LFS's physics engine flaws aren't at the top of the TODO-list. In normal racing the physics work, and I think the way cars collide with other cars in normal racing situations is believabe and realistic. On the other hand, if I smash a wall at 100km/h it really doesn't bother me if my car bounces to 7th floor - it would anyway be a total wreck if the damage model was complete.

(It's late, i'm tired, and i'm day dreaming but...) I could just picture we login tomorrow and somehow our LFS has had a proper damage model put into the game and we all end up totalling our cars. The amount of replays posted on here saying "look at my crash!" would go through the roof .....

ahhh we can all dream....

Yurgh! chocolate pudding's disgusting, i'm not eating that piece of ass again...