
Matt Gough Blog [closed]: walking 101
Thursday, Jul 15 2004, 05:37
XSTEP is an xml-based markup language for embodied agents based on the scripting language STEP. in the words of the creators it is 'developed on the formal semantics of dynamic logic'. The implementation and design of XSTEP is indicative of many avatar / agent scripting languages, rather than scripting 'cause', 'effect' is modelled in a top down approach. when simulating movement I argue that a bottom up approach will produce more 'realistic' results.
lets examine the XSTEP walk the authors of the paper discribe it as a 'movement which exchanges two main poses'. These 'poses' are:
a pose in which the left-arm/right-leg move forwardly while the right-arm/left-leg move backward.
a pose in which the right-arm/left-leg move forwardly while the left-arm/right-leg move backward.
Although walking may appear to be a pair of poses (due to our bilateral symmetry), walking is 'active' rather than 'posed' with each individual step distinct. further, each human has an individual and distinctive gait. The lack of distinction in gait and step is the key fault in walking scripts.
whilst XSTEP allows for variable step length it does not facilitate variable step action, lets have a look at the code:
the step length variable can be scripted like this:
I believe that a scripting that works for a human avatar should relate to a human being, so lets use the first example. Can you really instigate a walk from your shoulders and hips? I don't believe so. Human walking motion is instigated from our muscles (limbs) rather than our joints. Although gait analysis may define joint rotations of the hip, knees and ankle (note: knee and ankle are missing from XSTEP) these rotations are the result of the working muscles.
step length can be defined as:
in the second part of this post we will show how the body locations and pressure point markup of xdsn can be used to script more realistic, somatically correct walk scripts.