Keyframe Control of Smoke Simulations SIGGRAPH 2003
Overview ► Introduction ► Basis equation ► Proposed method ► Results ► Future work
Introduction ► Goal: Control of smoke simulation ► Difficulties Complexity Non-linearity ► Proposed method: Control the simulation by given density and velocity
Basis Equations ► Navier-Stoke Equation: Velocity diffusion Velocity advection External forces Smoke density advection
General procedure Add forceAdvect Diffuse Project
Framework ► State consists of of densities of velocity vector ► Initial state: ► State at time t: ► Simulation:
Control ► A set of keyframes that the smoke should achieve Specifies the density should match at time t Specifies the constraint on ► A set of parameterized forces Amount/direction
Matching Keyframes ► Goal Match the user-specified keyframe Use as little force as possible Solve for the equation
Computing Derivatives ► Calculating derivatives by simulating the entire process in a space consisting of A density and velocity field Their derivatives ► Initial state: ► State at time t:
Computing Derivatives ► Standard solver process: Mass preservation step Advects the smoke density Projects the resulting field Performs diffusion Advects the velocity External forces Calculating S
Computing Derivatives ► Calculating Each operation induces a operation Ex: And similarly for Therefore,
Derivatives ► Projection ► Diffusion
Derivatives ► Advection
Derivatives ► Mass Preservation ► Forces
Control Parameters ► Two types: Wind forces Vortex forces
Wind forces ► A single control vector scaled by a Gaussian falloff function ► Derivative
Vortex Forces ► Using Gaussian falloff approach ► Derivatives
Objective Function ► Smoothness Derivatives
Objective function ► Keyframe-matching Straightforward method Proposed method Derivatives
Results
Future Work ► Drawbacks: Computationally prohibitive with fine-grained control Optimization might be caught in local minimum ► To paradigms other than keyframes