Automatic Animation Between Yoga Poses
Jessica K. Miller and Yael F. Schwartzman
Objective
To compare transitions between hierarchical model  positions using Catmull-Rom curves and the A* search algorithm.
Methods

A* search algorithm

 

Uses a  good monotinically increasing objective function that complies with physics rules (such as maintaining center of gravity and avoiding intersections) and area-specific heuristics(such as keeping the spine straight in yoga). Dynamically generates a graph that traverses the search space using this objective function starting from the start position until the end position is reached.

 
Catmull-Rom curves
 This method guarantees a smooth transition between positions but it doesn’t guarantee “correctness of move” given physical and area-specific objectives (such as maintaining center of balance or intersecting other objects).
Results
 

Proof of Concept
Calculating Intersections
Animating Positions

A* search

 

 

Catmull-Rom curve interpolation

 

Future directions
  • Improve search heuristics. One way in which to improve it would be by allowing each state in the search space to be able to explore more than one degree of freedom at a time. This would hopefully take care of the "choppiness" of our current algorithm.
  • Use center of gravity and other physics principles to guide the search
  • Author and animate more positions
  • Include more degrees of freedom into A*'s search space
Acknowledgements

Karen Liu

Brett Allen

Adrien Treuille

Michele Banko