Efficient Yarn-based Cloth Simulations | Two Minute Papers #166

Key Takeaways

Dear fellow scholars, this is two minute papers with Kohaa Eher. This paper is about creating stunning cloth simulations that are rich in yarntoarn contact. Normally this is a challenging problem because finding and simulating all the possible contacts between tens of thousands of interlin pieces of geometry is a prohibitively long process. Also, due to the many different kinds of possible loop configurations, these contacts can take an awful lot of different shapes, which all need to be taken into consideration. Since we are so used to see these garments moving about in real life, if someone writes a simulator that is off just by a tiny bit, we'll immediately spot the difference. I think it is now easy to see why this is a highly challenging problem. This technique optimizes this process by only computing some of the forces that emerge from these yarns pulling each other and only trying to approximate the rest. The good news is that this approximation is carried out with temporal coherence. This means that these contact models are retained through time and are only rebuilt when it is absolutely necessary. The regions marked with red in these simulations show the domains that are found to be undergoing significant deformation. Therefore, we need to focus most of our efforts in rebuilding the simulation model for these regions. Look at these results. This is unbelievable. There's so much detail in these simulations. And all this was done 7 years ago. In research and technology, this is an eternity. This just blows my mind. The results are also compared against the expensive reference technique as well. And you can see that the differences are minuscule. But the new improved technique offers a four to five times speed up over that. For my research project, I also run many of these simulations myself. And many of these tasks take several allnighters to compute. If someone would say that each of my all-nighters would now count as five, I'd be absolutely delighted. If you haven't subscribed to the series, please make sure to do so. And please also click the bell icon to never miss an episode. We have tons of awesome papers to come in the next few episodes. Looking forward to seeing you there. Thanks for watching and for your generous support. And I see you next time. [Music]