Well, just a mistaken guy, who thinks it is the first game which has soft-body physics. Maybe I get something wrong. If so, what do you mean?
I don't believe this is real time simulation. They said it took 18 seconds to render(?) the car scene. I'm not smart enough to understand what is actually being shown, so I could be totally wrong.
The advantage with this is that is firstly isn't based on beams and nodes but on real physics, also it handles high frequency oscillations better, stuff like light stiff cars don't work well in beam
but it has to have some connections between all the body panels etc. for them to work properly, otherwise they would just fall off or be "glued" for good
This is very cool but incomparable to the concept we know and love over here Also it seems none of it has quite left the lab for end user purposes yet. But it seems open source if I understood it correctly. No, they actually state that it takes 18ms/frame to real-time-render the complete ac buggy and running it on 55 frames/sec. --- Post updated --- True, some vectoring should be involved. There's also no mentioning of a damage model, maybe currently all those components bounce back perfectly but what if the buggy hits a wall ? Will there be permanent dents ? I suspect not, I don't think any damage model is involved.
Well, they stated that it runs at about 55 fps. Yes and no... Essentially what is going on here is that it is dynamically simulating collisions of mesh objects in real time and accurately enough to enable fluid motion of joints. So basically, unlike beam where to simulate a ball joint, we just connect a beam to a node, and then limit that nodes motion through other connections, what they are doing here is physically creating the geometry of the ball joint... and then just letting the simulation run with it. What comes out is natural, realistic (if jittery at the moment... I am guessing their tolerances have to be rather large at the moment still for this to work), proper, physically simulated motion. Just from geometry. Now this isn't the first time something like this has existed. I literally have a tool just like that in Autodesk Inventor called Dynamic Simulation. The difference here though is that in Inventor, the motion CAN NOT run in real time... or at least on my work computer, it's not even close. The previous algorithms used to compute the geometry collisions is VERY slow. I would say it runs at about 3 FPS on my computer. So this is a massive improvement.
what that video shows is literally just collision triangles in beamng it's nothing new. That channel shows things that are "new" but are a thing since like 2009 lol
Yes and no, while collision simulation isn't something new, the problem with is that it's traditionally bad at complex shapes, because you'll need exorbitant amounts of points to check for (if your simulation also has to run at a reasonable frequency, beamngs 2500/sek isn't actually that high, the workload multiplies exponentially). The cloth in the video is a notoriously difficult scenario, as you can see traditional method result in the cloth getting stuck in the spikes, something similar happens with small objects in beamng (guard rails, traffic cones), this as far as I'm concerned isn't a huge problem, but still something that could be improved... Yeah probably not, the methods of the paper are most likely not compatible with the legacy beamng tech, but the cool thing about this one is it's far higher fidelity combined a 30 faster simulation time. Sure beamng physics calculations are more than just collision, but I think the future lies in faster code rather than more power
Dear fellow skahlors, dis is too minet paypers with docter kahlo johai faher Jokes aside I love this dude I watch him constantly what we need in BeamNG is fracture simulation, so car accidents are like, actually super realistic.