[ODE] Scientific modelling of a runner

[Marcus Brubaker]

jnilson_99 at yahoo.com wrote:

>Hi Martin,
>
>You might want to talk to Jason Malios,
>jmallios at cs.brown.edu , who has tried to do a similar
>thing. That is take key frame data and mix it with
>physics based inverse kinimetics (ode). Search his
>name on the pipermail usergroup for ode.
>
>As far as your basic question, which is, can a physics
>engine model human movement, well much has been
>written about that :-).
>
>
>Forgive me if I suggest you are naive, but you must
>understand the considerable difficulty in such
>modelling. NO ONE has done it before, otherwise we
>would see animotronic humans roaming the earth. Yes a
>few people have simulated bipedal movement, see Russ
>Smith PHD thesis on the Fox Controller. However SMOOTH
>simulation of human bipedal movement is far in the
>future I am affraid.
>
>I only say this in cautioning you to the complexity of
>the problem you seem eager to undertake.
>  
>

Not as far in the future as you think.  See the PhD thesis of Jerry 
Pratt from MIT who has successfully created a "2D" walking kneed and 
ankled robot and simulated a 3D version, both of which have extremely 
natural looking gaits.  Based on some of his ideas, I myself have 
created a stochastic walking controller for a 2D kneed biped using ODE 
and am currently working to extend it to an ankled model.

The problem of determining forces based on mocap data has been recently 
published about at SIGGRAPH 2005, see 
http://grail.cs.washington.edu/projects/charanim/phys-style.html  They 
use a Lagragian dynamics formulation which allows them to easily infer 
the external forces.  I've spoken with one of the authors about this 
problem (as it's something I've looked into) and was told that they were 
forced to write their own kinematic procecssing of the raw mocap data 
because the kinematics output from the Vicon system induced a lot of 
noise in the computed forces.  It is not a trivial thing but it has been 
done.  However, a simulator like ODE will generally be of no help.

Regards,
Marcus