LARRUZ covers a paper on steering fluid and smoke simulations toward target shapes using Primal-Dual optimization. The method guides flows while keeping them natural, prevents fluid from sticking to walls, and adds fine detail to coarse simulations. Applications include smoke plumes, explosions, and breaking dams.
Dear fellow scholars, this is two minute papers with. Today we are going to talk some more about fluid simulations and fluid guiding. Hell yeah. As you know all too well, it is possible to simulate the laws of fluid motion on a computer, make a digital scene, and create absolutely beautiful videos such as the ones you see here. Newer and newer research papers show up to both extend the possible scenarios that we can simulate. And and there are also other works to speed up already existing solutions. This piece of work introduces a technique that mathematicians like to call the Primal dual optimization method. This helps us accomplish two really cool things. One is fluid guiding. Fluid guiding is a problem where we are looking to exert control over the fluid while keeping the fluid flow as natural as possible. I've written my master thesis on the very same topic and can confirm that it's one hell of a problem. The core of the problem is that if we use the laws of physics to create a fluid simulation, we get what would happen in reality as a result. However, if we wish to guide this piece of fluid towards a target shape, for instance, to form an image of our choice, we have to both retain natural fluid flows, but still create something that would be highly unlikely to happen according to the laws of physics. For instance, a splash of water is unlikely to suddenly form a human face of our choice. The proposed technique helps this ambivalent goal of exerting a bit of control over the fluid simulation while keeping the flows as natural as possible. There are already many existing applications of fluids and smoke in movies where an actor fires a gun and the fire and smoke plumes are added to the footage in post production. However, with a high quality fluid guiding technique, we could choose target shapes for these smoke plumes and explosions that best convey our artistic vision. And number two, it also accomplishes something that we call separating boundary conditions which prevents imprecisions where small fluid volumes are being stuck to walls. The guiding process is also followed by an upsampling step where we take a coarse simulation and artificially synthesize sharp high frequency details onto it. Computing the more detailed simulation would often take days without such synthesizing techniques. Kind of like with Wavelet Turbulence, which is an absolutely incredible paper that was showcased in none other than the very first 2 minute papers episode. Link is in the video description box. Don't watch it, it's quite embarrassing. And all this leads to eye poppingly beautiful solutions. Wow, I cannot get tired of this. In the paper you will find much more about breaking dams, tornado simulations and applications of the Primal dual optimization method. Normally, to remain as authentic to the source materials as possible. I don't do any kind of slow motion and other similar shenanigans, but this time I just couldn't resist it. Have a look at this and I hope you'll like the results. If you feel the alluring call of fluids, the I've put some resources in the video description, including a gentle description I wrote on the basics of fluid simulation and fluid control with source code both on the CPU and gpu, and the link to Doyeop Kim's amazing book that I'm currently reading. Highly recommended. If you also have some online tutorials and papers that helped you solidify your understanding of the topic, make sure to leave a link in the comments. I'll include the best ones in the video description. If you would like to see more episodes like this one, make sure to subscribe to 2 Minute Papers. We would be more than happy to have you along on our journey of science. Thanks for watching and for your generous support and I'll see you next time.
