Finally, Deformation Simulation... in Real Time! 🚗

Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. Today, with the power of computer  graphics research, we can use our   computers to run fluid simulations, simulate  immersing a selection of objects into jelly,   or tear meat in a way that much like in  reality, it tears along the muscle fibers. If we look at the abstract of this amazing new  paper, we see this, quoting: “This allows us to   trace high speed motion of objects colliding  against curved geometry, to reduce the number   of constraints, to increase the robustness of  the simulation, and to simplify the formulation   of the solver. ” What! This sounds impossible,  but at the very least, outrageously good. Let’s   look at three examples of what it can do and see  for ourselves if it can live up to its promise. One. It can simulate a steering  mechanism full of joints and contact.    Yup, an entire servo steering mechanism is  simulated with a prescribed mass ratio. Loving   it. I hereby declare that it passes inspection,  and now, we can take off for some off-roading.    All of the movement is simulated really well,  and… wait a minute. Hold on to your papers!    Are you seeing what I am seeing? Look. Even the  tire deformations are part of the simulation!    Beautiful. And now, let’s do a stress test,  and race through a bunch of obstacles and see   how well those tires can take it. At the end of  the video, I will tell you how much time it takes   to simulate all this…and note that I had to look  three times because I could not believe my eyes. Two. Restitution! Or in other words, we can smash  an independent marble into a bunch of others   and their combined velocity will be correctly  computed. We know for a fact that the computations   are correct, because when I stop the video  here, you can see that the marbles themselves   are smiling. The joys of curved geometry and  specular reflections. Of course, this is not true,   because if we attempt to do the same with  a classical earlier technique by the name   position based dynamics, this would happen.   Yes, the velocities become erroneously large   and the marbles jump off of the wire. And  they still appear to be very happy about it.    Of course, with the new technique, the  simulation is much more stable and realistic. Talking about stability. Is it stable  only in a small-scale simulation,   or can it take a huge scene with lots  of interactions? Would it still work?    Let’s run a stress test and find out.   Ha-haa! This animation can run all day long   and not one thing appears to  behave incorrectly. Loving this. Three. It can also simulate these beautiful  high-frequency rolls that we often experience   when we drop a coin on a table. This kind of  interaction is very challenging to simulate   correctly because of the high-frequency nature of  the motion and the curved geometry that interacts   with the table. I would love to see a technique  that algorithmically generates the sound for this.    I could almost hear its sound  in my head! Believe it or not,   this should be possible and is subject to  some research attention in computer graphics.    The animation was given, but the sounds  were algorithmically generated. Listen!    Let me know in the comments if you are one  of our OG Fellow Scholars who were there   when this episode was published  hundreds of videos ago. So, how much do we have to wait to simulate  all of these crazy physical interactions?    We mentioned that the tires are stiff and take a  great deal of computation to simulate properly.    So, as always… all nighters, right? Nope! Look at  that! Holy mother of papers! The car example takes   only 18 milliseconds to compute per frame,  which means 55 frames per second. Goodness!    Not only do we not need an all-nighter, we  don’t even need to leave for a coffee break!    And the rolling marbles took even less, and,  wo-hoo! The high-frequency coin example needs   only one third of a millisecond, which means that  we can generate more than 3000 frames with it per   second. We not only don’t need an all nighter or  a coffee break, we don’t even need to wait at all! Now, at the start of the video, I noted that the  claim in the abstract sounds almost outrageous.    It is because it promises to be able to do more  than previous techniques, simplify the simulation   algorithm itself, make it more robust, and do all  this while being blazing fast. If someone told me   that there is a work that does all this at the  same time, I would say that give me that paper   immediately because I do not believe a word of  it. And yet, it really lives up to its promise. Typically, as a research field matures, we see new  techniques that can do more than previous methods,

but the price to be paid for it is in the form  of complexity. The algorithms get more and more   involved over time, and with that, they often  get slower and less robust. The engineers in the   industry have to decide how much complexity they  are willing to shoulder to be able to simulate all   of these beautiful interactions. Don’t forget,  these code bases have to be maintained and   improved for many-many years so choosing a simple  base algorithm is of utmost importance. But here,   none of these factors need to be considered,  because there is nearly no tradeoff here:   it is simpler, more robust,  and better at the same time.    It really feels like we are living in a  science fiction world. What a time to be alive! Huge congratulations to scientists at NVIDIA  and the university of Copenhagen for this.    Don’t forget, they could have kept the results  for themselves, but they chose to share the   details of this algorithm with everyone, free  of charge. Thank you so much for doing this. Thanks for watching and for your generous  support, and I'll see you next time!