PhD Thesis Proposal Defence


"Dynamic Interaction for Rigid Bodies and Deformable Surfaces"

By Mr. Sai Keung Wong


Abstract:

We aim to develop an animation system to perform animation of deformable
surfaces with rigid bodies efficiently. There are three major components
in an animation system: modeling unit, collision detection unit and
collision response unit. The modeling unit handles the modeling of
deformable surfaces either geometrically or mechanically. The collision
detection unit tells us which two regions collide. Other information such
as the collision time, the collision orientation, the collision normals,
and the contact points may be computed. In order to prevent the colliding
regions from further penetration, the corresponding collision constraints
should be satisfied. This is handled by the collision response unit.

We survey the modeling techniques for deformable surfaces and discuss
their advantages and limitations. The modeling techniques are classified
into three categories: geometrical, physical and hybrid. Geometrical
techniques are designed for capturing the static geometric appearance of
deformable surfaces. Physical techniques are designed for modeling the
physical properties of deformable surfaces. Hybrid techniques combine both
the geometrical and physical methods.

Collision detection is still a major bottleneck in the animation of
deformable objects, especially for deformable surfaces. Unlike rigid
bodies, there may be up to tens of thousands of potential colliding pairs
when deformable surfaces are in contact with other objects.

We review the collision pipeline and explore the possibility to improve
the performance further either in the broad phase or narrow phase stage.
We investigate methods to share the work load of the collision detection
process between the CPU and the conventional graphics hardware. We propose
an architecture to speedup the process performed in the bottom-most layer
- intrinsic test unit.

Finally, we propose a method to resolve all the collisions simultaneously.
We briefly present an overview of this method. By limiting the movement of
the vertices, we prevent the colliding regions from penetrating each other
further.

The preliminary results and future work will be presented at the end of
the presentation.


Date:     		Thursday, 27 November 2003

Time:                   10:00a.m.-12:00noon

Venue:                  Room 2404
			lifts 17-18

Committee Members:      Dr. Andrew Horner (Supervisor)
			Dr. George Baciu (Supervisor)
                        Dr. Siu Wing Cheng (Chairman)
			Dr. Chiew Lan Tai
			Dr. Chi Keung Tang


**** ALL are Welcome ****