Quote :

Im not game developer, maybe its not so easy..

I'm a [hobbyist] game developer, and it isnt so easy However, it is possible .

Quote from schofei :

Online you only get 6 packets/sec, so it's even worse, although there might be some interpolation in between.

Yes, there is prediction in between, otherwise driving would be impossible

maybe im wrong but i remember in s1 the collision system was much better , I remember the pushing wich people moaned abou or "bump drafting" but now when u tap another car in the back its a lottery to what will happen same with tapping people on the sides , I do not remember this problem before maybe im mad

I remember when the S2 demo was first released, I was under the impression that collisions had taken a step backwards from S1. Gentle pushing and light contact gave a much more violent sensation than before. TBH, I'm used to it as it is now, and I can't actually say for certain if anything did change at all, in fact.
I think for the most part it's very good, 95% of the time collisions have the effect you'd expect.

maybe the system hasn't changed but kevs right, its not the same as it used to be.... if someone touches the back of your car at all (since S2 came out) it bounces all over the place.

This is most drastic in the UF1k.... illepall

This collision system combined with the even more ridiculous "crashing" sound is the most annoying thing in LFS. That sound just breaks my ears and headphones. I mean that sound is really annoying, when racing close I'm more afraid of hearing than sound that actually getting crashed.

And you can't even turn it off, not even by turning volume to 0% from ScrapeSoft.raw and ScrapeHard.raw files. Afaik that sound is generated from those files.

Quote from Becky Rose :

As computers get faster Scawen can increase the number of iterations the physics engine does to improve accuracy.

wont work with the low amount of packets sent during online play ... try crashing in sp orbit flights rarely happen with ais

if im not mistaken improved collision detections effectively create samplepoints in between the usual sampling grid

Quote from Honey :

the problem is that Poisson teach us that in such case it is 99.9% (or more) probable that a collision happens in the meantime

theres a poission theorem on sampling ?

Quote from schofei :

Difficult to understand is, that a collision is handled differently, one car gets tossed around just like by titanic forces whereas the other driver very often not even realizes that his car was hit (and in fact, his F10 doesn't show any damage).

afaik physics are calculated on the clients with the server doing some basic sanity checking to prevent cheats (maybe it doesnt)
so one car might lag into your own sending you flying and until your position packets reach the other client your car is already barely touching his

Quote from kevomcenzy :

maybe im wrong but i remember in s1 the collision system was much better , I remember the pushing wich people moaned abou or "bump drafting" but now when u tap another car in the back its a lottery to what will happen same with tapping people on the sides , I do not remember this problem before maybe im mad

when you look at crashes in slowmo it appears as if the forces caused by that collision lag behind the damage model
maybe something like the overlap increasing while the damage is calculated

Quote from Shotglass :

theres a poission theorem on sampling ?

no, i meant that if collisions are calculated at t1, t2, t3, t4, t5, etc. the probability that two objects start to collide at exactly t1, or t2, etc is about zero, thus 99.9% of times objects collide and since collision detection callback function still wasn't called, object will continue their movements and thus meaning object will overlap -> overlap is the source of the problem.

in few words, i used poisson name to explain it's no lfs flaw for object overlap, it's a mathematical issue.

Before realistic collision physics, realistic modifications to physics due to a crash would be better. Specifically, aero physics. Its amazing how I can have a completely destroyed front wing on the FOX with no negative handling impact (assuming the front wheels are still pointed straight). And the GT cars don't seem to lose any front downforce or having any engine cooling issues after coming up into somebody's rear end.

Modifications to a cars regular drving physics because of a crash would definitely at realism and maybe have the added effect of cleaner driving.

Quote from Honey :

no, i meant that if collisions are calculated at t1, t2, t3, t4, t5, etc. the probability that two objects start to collide at exactly t1, or t2, etc is about zero, thus 99.9% of times objects collide and since collision detection callback function still wasn't called, object will continue their movements and thus meaning object will overlap -> overlap is the source of the problem.

in few words, i used poisson name to explain it's no lfs flaw for object overlap, it's a mathematical issue.

The problem is that there are two simulations trying to simulate the same thing, but they are not in sync because of network latency, so bad things can happen when one simulation thinks you are somewhere you are not and vice versa. Some work could probably be done to cut down on the sending you into orbit collision responses while playing online, but I'm sure Scawen has plenty to stay busy for a long time.

The physic system should have atleast an update rate of 60 times/s and 30 in multiplayer, that would make the incredibly drive right through barries impossible.

Quote from InSaneTK :

The physic system should have atleast an update rate of 60 times/s and 30 in multiplayer, that would make the incredibly drive right through barries impossible.

The collision physics already run 100 times a second both in single and multiplayer.

Vain

The collision detection LFS has atm, in my personal oppinion is good. Racing means NOT going 200mph into someone whos sitting still. I've never experianced the effects that people have stated (when you are inside the other car and go flying) i am occasionaly door to door with a few people, especialy when overtaking on a tight bend and i still dont get it, neiter does the other person to my knowledge.

Whats there atm does its job. It allows contact and not "ghost" racing whereby their is no collisions and people can pass thought eachother.

I've been thinking about those small hits for long. In real racing, it's quite normal to bump into someone's rear now and then, not hard, but just a little touch. But in LFS, when I get close to someone, I'm scared, because if I brake just a little bit too late (what would cause the small bump), the other car or me, could be sent flying, and that's not what I want! So I just keep backing off untill this gets better, but it does kinda turn down the fun of close racing for me.

online lag is the problem and it wont go away. When cars start to pinball on contact you know the server is overloaded or the pings times are too high.
Its less noticable with the slower cars.

In any kind of racing bumping the back of another car when it is already close to the edge of it's abilities can have very bad consequences(losing control, etc)

Quote from -wes- :

online lag is the problem and it wont go away. When cars start to pinball on contact you know the server is overloaded or the pings times are too high.
Its less noticable with the slower cars.

it's not an online issue, online is only a little worse, but it's pretty bad even on single player, this issue has been discussed many times and Scawen also explained a bit why, he knows quite well where the problem resides, he only have to plan when to work on it, to find a solution and code it.

What you'd have to do on collisions is a rudimentary sanity check. The easiest way to do this is probably just storing a small position history of each car, so when two cars, or a car and a object collide, you can look at the history and see the relative speed that the objects had to each other. So instead of taking the intersecting "deepness" to calculate the force, you can just extrapolate from the position data that the objects were closing in to eachother with, say, 2 mph and calculate the forces from there.

Single player + autocross + barriers = car 50 feet in the air

Quote from Bob Smith :

Single player + autocross + barriers = car 50 feet in the air

haha yea, a friend of mine where going 20km/h into a barrier and did 4 backflips.

for some inexplicable reason barriers send you to the moon and back a lot more often than cars
eg in banger racing it was a lot worse to hit a barrier than to hit another car even on a laggy server

Quote from Honey :

snip

The op was talking about the car to car collisions so was I.
The other stuff is pretty funny. Doesn't it happen more often on slower pc's?
I've not had many moon trips myself

Quote from -wes- :

The op was talking about the car to car collisions so was I.
The other stuff is pretty funny. Doesn't it happen more often on slower pc's?
I've not had many moon trips myself

I don't have moon trips either... It's just that when comming into contact with someone, the collision is often much more severe than it should be. I don't know how much my ping is, and how laggy I am to other people, but I do notice sometimes that I crash a lot harder than I should have.

Quote from Vain :

LFS's collision system looks like it is based on a simple rule: No two objects may intersect at any given time.
That means when two objects intersect because of lag there has to be a force that is so big that by the next calculation step, that means within 100th of a second, a legit state of objects is achieved. That means if a car is stuck 1 meter into a red/white barrier LFS will generate a force that will rocket the car away fast enough so that by the next calculation step the car doesn't intersect anymore. In other words: It is accelerated enough to move 1 meter within a hundreth of a second. That's equals more than 300 km/h (according to Orions calculation far above).

This is obviously a placeholder and should be fixed as soon as possible. Good we're approaching S2 beta.

Vain

awesome explanation - are you repeating what you heard or just figured it out? It makes perfect sense! (damn those red/white barriers!)

I think the collision is excellent - in that you dont bounce of objects you havn't hit ... it's a problem in some games but LFS has very neet collision zones, the problem is what happens after the actual collision - this could be handled by an independent impact system - rather than within the racing physics system which handles it now.