Real-time Finite Element Based deformations
We have developed a matrix clustering method for speeding up finite element computations for non-rigid object animation. The method increases the efficiency of computing deformation dynamics through a compression scheme that decomposes the large force-displacement matrix into clusters of smaller matrices in order to facilitate the run-time computations of linear finite element based deformations. The deformation results are compared with the results produced by using modal analysis method and the standard linear finite element algorithm. We demonstrate that the proposed method is stable with comparative computational speed to modal analysis method.
A hierarchical skeleton-based system is also
implemented to add constraints to material nodes. Thus, real-time
deformations can be directed by motion captured data sets or key-framed
animations (the jelley man
images).
(comparsion of our algorithm with standarad FEM and modal analysis, right: a bunny deformation force distrbution)
Our recently developed new real-time finite element soft tissue model:
Main technical strengthes of the new model are: extrmely fast, i.e real-time performance up to 100 fps on 8 cors computers; handles complex structures large deformations, this enables us to work on needle insertion and cut simulations for a range of surgual traing applications. Here are some examples:
Motion synthesis
We
have developed a novel compression algorithm for reducing the size of
motion captured datasets for example, data captured by Vicon mocap system,
upto 70%. The algorithm works for a character skeleton
with 62 degree of freedoms at the movement. Apart from the good
compression rate, one of advantages of the algorithm is also the
sensitivity of the algorithm in terms of detecting fine features
of different movements within a motion sequence, for example we are
able to classify a set movement that share similar features into
one group .
2. Constraint based IK
solver:
We have developed a constraint
based real-time IK algorithm, which works for multiple IK
chains . The algorithm is based on a real-time optimization algorithm
applied on a set of constrained dynamic system. We are able to
solve four IK chains for character animation in real-time
.
3. Learning algorithm for virtual
characters:
We have developed an animation system that simulate adaptive motion behaviours for virtual character animation. the system is based on a set of learning and evolution algorithms, such as genetic algorithm, Q-learning, and classifier system
( A synthesized walk motion )
(A synthesized martial Art motion)
(real-time constraint based optimastion for four IK chains)
(Learning by experience)
|
|
Real-time simulation guidewire and catheter for interventional radiology training
Our paper on guidewire and Catherter modelling and simulation has been selected as one of the 70 best papers out of 305 submissions to be published in the internal jounral The Visual Computer, and to be presented at the internation conference " Computer Graphics International 2010" in June in SINGAPORE.
W Tang, P. Lagadec, D. Gould, T. R. Wan, J. Zhai, and H. Thien (2010) " A Realistic Elastic Rod Model for Real-time Simulation of Minimally Invasive Vascular Interventions" to be published in The Visual Computer, international journal of computer graphics
(guidewire deformation) (insde a blood vessel) ( an overview )
(to view
the videos, right mouse click to save the video onto your desk top
)
video 1
Wen Tang last update Sept. 2009 copyright@