Michael B. Nielsen is a principal engineer in the Bifrost FX team at Autodesk. Prior to joining Autodesk he worked as an assistant professor at the University of Aarhus in Denmark, and as a researcher at Weta Digital in New Zealand where he received screen credits on Rise of the Planet of the Apes, Tintin – Secret of the Unicorn and The Hobbit. He received his PhD in computer science from the University of Aarhus.
How did you get interested in the physics-based simulation field and what does it mean to be a Principal Software Engineer at Autodesk?
I’ve had a strong interest in computer graphics ever since I started programming tiny demos in assembler code on my Amiga 500 back in the nineties when I was in high school. This interest guided many of my choices at the University and I was lucky to have the opportunity to pursue a master’s and later a PhD within the computer graphics field. It was during my PhD studies that I became particularly interested in physics-based simulations as I joined the graphics research group led by my PhD advisor Ken Museth who has a strong background in computational physics and chemistry. At Autodesk I do part research, part development. Usually I follow through on a project from the inception of an idea to final implementation in the Bifrost product. This is very fulfilling.
What is the difference between working as scientist in applied computer graphics simulation field and as a research scientist at Uni? What interested you in taking this CG/VFX development path?
I have very much enjoyed doing research both in academia and in the industry. There are pros and cons to both. I guess I have gravitated towards the industry as I find the closer interaction with technical directors and users very inspiring. I am also motivated by the path from the initial idea to the actual implementation in the final product and seeing the software getting used by talented artists. In academia you’re often getting involved with many tasks that are not directly related to your research. In the industry, at least in my experience, you’re given the time to focus on the R&D.
Do you have any advice on encouraging young people to take the CG research path?
I would let them know that CG research and development is a lot of fun and a creative mix of art and science. It’s a very exciting field to be in if you’re interested in applied math and physics and it makes use of almost all branches of computer science. Plus, given the distribution of studios and research labs all over the world, there’s a rich opportunity to travel and live in different beautiful parts of the world.
Do you run any practical physics tests, like spilling huge amount of water and setting things on fire for collecting the reference footage? 🙂
Not at the moment, though that would be a lot of fun. We do rely heavily on videos of real-world phenomena posted online as reference footage. We also consult literature published in applied physics in order to validate our results. Usually the results published are produced under very controlled conditions which makes them easier to re-produce in order to validate a mathematical model. Recently we’ve for example validated our combustion solver against real-world measurement of laminar flame speeds produced by the US Bureau of Mines back in the fifties by building a simplified digital model of their apparatus and reproducing their experiments digitally.
What is helping you making better science papers, coffee or tea?
Coffee, cola and way too much cake and chocolate are important ingredients of any paper I’ve ever written. Apart from this, what I find most inspiring is talking to artists about their work, the problems they experience in working with state-of-the-art tools and the often-ingenious solutions they come up with to work around limitations in the software. It’s at times like this that many research ideas are born.
You’ve been a returning SIGGRAPH white papers presenter since 2005. What are the benefits of attending and presenting at conferences like SIGGRAPH?
SIGGRAPH is a great venue to share new ideas and learn from your peers. I very much appreciate that it’s a mix of new research and production talks. In particular, I find the talks sessions very inspiring. This is a unique opportunity to learn from very talented technical directors and researchers. When we do simulation research in the Bifrost project we always make sure to catch up on all the latest developments that have been presented at SIGGRAPH. There are so many great ideas that we have leveraged over the years. Similarly, we highly appreciate when we are given the opportunity to contribute our own ideas back to the community by presenting at SIGGRAPH. It’s an honor every time. The bar at SIGGRAPH is high.
What are the latest developments in the physics-based simulation and what would you like to see being achieved in the future?
There are many exciting trends. Advancements are continuously being made to the Material Point Method (MPM) and the range of phenomena that can be simulated steadily grows. In Bifrost we’ve leveraged very recent papers on spatially adaptive data structures and simulation methods. This is an area of simulation which has a huge potential but making it easy and intuitive to use is not straightforward. Although we love the physics of fluids, we’re also closely monitoring progress in deep learning methods applied to simulation. Related to simulation of smoke and fire I would like to see improvements in high-level intuitive user-controls as well as in the ability to synthesize small scale detail in a physically plausible way.
What are the differences that Bifrost brings compared to other procedural dynamics simulation software?
With respect to the new Bifrost smoke and combustion solvers that I have contributed to, our focus has been on spatial adaptivity, higher numerical accuracy and on making the combustion solver more grounded in physics than is traditionally done in graphics. Spatial adaptivity refers to the ability to represent any part of the simulation domain at lower or higher resolution in order to best take advantage of the computational resources to achieve the desired level of detail in the simulation. Robert Bridson, Sr. Principal Research Scientist on the Bifrost team has developed a new improved time-integration method that allows simulations to develop more natural detail and retain energy. Finally, our combustion solver supports features such as real-world fuels and under-the-hood uses the correct physical values for properties such as specific heat, heat of combustion, laminar flame propagation speed and so on. In combination with modeling all the fundamental thermodynamic processes, including radiative heating, we envision that this will eventually make it easier for an artist to achieve physically plausible natural-looking results by default.
Weta Digital has a great name for their research for making their film looking photorealistic. What were you working on for Rise of the Planet of the Apes and Weta in general?
At Weta I was part of the research group. My focus was on volumetric data structures and tools for simulation and rendering of fluids and hair. For example, I worked on algorithms and numerical methods for controlling high resolution liquid simulations. My main contributions were to the Spielberg movie “Tintin – Secret of the Unicorn”. There were a lot of water shots in that movie. Inspired by discussions and in collaboration with artists I worked with Robert Bridson on a new method for guiding liquid simulations which was presented at the SIGGRAPH 2011 paper sessions and used in several shots of the movie. Similarly, a shot in Tintin where sand dunes turn into huge water waves inspired our paper on synthesizing waves from animated height fields which was presented at SIGGRAPH in 2013.
What are the topics of your CGA2019 Talks?
At CGA I will be talking about our physics-based approach to smoke and fire simulation in Bifrost.I will also cover our new approaches to spatial adaptivity. That is, how to focus computational effort where it is most needed, for example near the camera or in areas where the flow develops small-scale detail. Our overall philosophy behind the smoke and combustion solvers in Bifrost is that eventually they should produce physically plausible results as the default. We’re not quite there yet, but we’ve investigated state-of-the-art approaches in CFD and tailored some of these models for computer graphics where they made an improvement to the simulations visually. These improvements are now part of the Bifrost combustion solver. Turbulent combustion simulation is hard to get right and not fully understood to this day.