[Bf-committers] NZ & Game Engine Physics
j.merritt at pgrad.unimelb.edu.au
Sat Nov 27 07:19:42 CET 2004
I'll see if I can make it to Wellington - I'll keep in touch closer to
I should take a good look at the game engine physics code sometime.
Just out of interest, how easy do you think it would be to incorporate a
specific mechanical model into the rest of the physics system? Is there
a good level of abstraction / modularization that might allow this? I'm
interested mostly because it would be interesting to be able to
incorporate a specific mechanical sub-system into the more general
system. An example of a specific sub-system might be a driven kinematic
chain representing a musculoskeletal system (or a robot, for that matter
For my PhD, I'm working on resolving muscle forces in the equine
forelimb. The basic stages are as follows:
1. Measure motion of the horse (kinematics), and ground reaction forces.
2. Apply inverse dynamics to find the moments applied to each joint
of a rigid-body model of the limb.
3. Measure muscle architecture parameters and anatomical origins and
4. Use static optimization to solve for muscle coordination
according to various criteria (eg: minization of the sum of muscle
5. Solve for joint reaction forces.
6. For selected bones, predict strains induced during locomotion,
and measure these strains using gauges implanted in-vivo for comparison
Step 4 is required because the horse (like virtually every mammal I can
think of - and I've seen inside quite a few!) has more muscles in its
forelimb (esp. in the shoulder) than are actually necessary to achieve
any specific motion. To find out how the force is distributed among the
various muscles, it's necessary to introduce this extra
physiologically-based (ie: non-mechanical) condition.
So, none of my PhD directly involves so-called "forward simulation",
like the game engine is trying to achieve. However, I think it would be
extremely cool to be able to take the results of my work and drive a
model of the horse's limb, with different muscles perhaps hilighted in
different colours depending on the force they are applying. However,
there's no way that a simple engine like the one you describe could
possibly be used drive the model accurately - I think the numerical
errors would make the exercise quite pointless. So, I'd require some
way of being able to incorporate an extra basic element (the
musculoskeletal model) that can perform its own internal calculations,
but also interact with the remainder of the system somehow.
I've been thinking about trying to achieve most of the above using
Python within the main part of Blender, but it would just be interesting
to see if it could be done in the game engine. It doesn't add anything
to my PhD (I have the original kinematic data afterall, so why do I need
to drive the limb to see how it moves?! :-), but I'm interested in the
>I haven't got any plans to go down south, but I can meet up if you're coming
>up to Wellington at all.
>The physics response is still quite simple:
>1. Integrate forces
>2. Integrate momentum
>3. Collision response
>4. Contact Response
>The integration is all Euler and the collision response is impulse based. The
>contact response simply moves intersecting objects that shouldn't.
>On Friday 26 November 2004 15:32, j.merritt at pgrad.unimelb.edu.au wrote:
>>Just a quick email to find out if you're going to be on the South Island of
>>NZ any time during December? I'm going to be at various places
>>down there from the 12th to the 31st, and wondered if you'd like to meet
>>up sometime so that I can congratulate you on how much work you're
>>doing in the game engine? :-)
>>I'm a qualified Mechanical Engineer (BE/BSc) from Melbourne Uni, and
>>am currently doing a PhD in biomechanics. Most of my work has very
>>different requirements to the physics simulation in the game engine,
>>but I'd gladly help out if you have any queries that you think might be
>>btw - Do you know if power plugs have the same geometry in NZ as
>>they do in Australia? It only just occurred to me to wonder about that.
>>Bf-committers mailing list
>>Bf-committers at projects.blender.org
More information about the Bf-committers