“Model Predictive Control for Robust Art-Directable Fluids” by Stuyck and Dutré

  • ©Tuur Stuyck and Philip Dutré

  • ©Tuur Stuyck and Philip Dutré

  • ©Tuur Stuyck and Philip Dutré



Entry Number: 73


    Model Predictive Control for Robust Art-Directable Fluids



    Physics-based animation has become an important tool in computer graphics and is essential in recreating realistic looking natural phenomena. Researchers have been looking for tools to control passive simulations that allow artists to easily modify the simulation to best suit the artistic requirements. However, fluid motion is very hard to predict and it is very difficult, if not impossible, to achieve specific behavior just by altering the global variables. Active control of the simulation will be necessary to achieve this goal.
    We present a model predictive controller (MPC) for fluid simulations that is able to achieve control with high precision based on an optimization process. The system has the potential to be used to control fluid simulations at run-time to deal with unforeseen user-interactions by controlling a simplified simulation using a sliding window to anticipate future changes. MPC is already being used extensively for controlling massive industrial processes. Likewise, the graphics community has applied this approach for generating bipedal locomotion. In the same vein, we hope that our method will provide artists with a robust and reliable tool to orchestrate complex simulations according to artistic needs and helps to obtain physically-plausible simulations with minimal effort.


Additional Images:

©Tuur Stuyck and Philip Dutré


ACM Digital Library Publication:

Overview Page: