Topology-based Motion Analysis and Synthesis


Despite the huge amount of researches that has been done in the field of robotics and computer animation, synthesizing motions that involve close contacts is still a challenging task. Previous researches suffer from a huge amount of computation for collision avoidance as path-planning is done at the level of generalized coordinates or Cartesian coordinates, which are the lowest level of state representation.

In this project, we efficiently plan such complex motions by introducing the idea of topology coordinates [Ho and Komura EG2009] which takes into account the topological relationships of the segments. In topology coordinates, writhe, which represents how much the segments twist around each other, is the main attribute of the state space. This is inspired by tangle in knot theory as in our pilot study [Ho and Komura PG2007, Ho and Komura Humanoid2007]. We can easily avoid collisions of segments by simply moving the segments along the axis of writhes in topology coordinates.

Such an approach can be used for synthesizing character-object interactions [Ho and Komura EG2009], character-character close interactions such as wrestling [Ho and Komura MIG2009, Ho and Komura CAVW2011] controlling humanoid robots [Ho et al. IROS2013] and motion retrieval [Ho and Komura IEEE TVCG2009]. The topology-based representation can further be used for constructing a collision-free human motion space for solving Inverse Kinematics (IK) efficiently [Ho et al. PG2013].

Readers are also referred to a closely related project on modelling two-person and human-object close interactions.


The Team

Dr. Edmond S. L. Ho

Senior Lecturer, University of Glasgow

Prof. Taku Komura

Professor, University of Hong Kong

Dr. Hubert P. H. Shum

Associate Professor, Durham University