Previous Talks

Abstract An central element of embodied cognition is the body itself, yet we still are without a complete picture of how movements are organized.
The main reason for this state of affairs is that modeling the generation of movement can be extraordinarily difficult owing to the complexity of the underlying musculoskeletal system. The dynamics equations are non-linear and have many degrees of freedom, making them all but intractable to solve directly. We show that this system can be greatly simplified if its movements are modeled as a set of basis movements that can be adapted to varying conditions. We show that the command torques for such movements can be read out by a dynamic model that is constrained to follow human movements as measured by a motion capture system. The resultant torques can be replayed in the model to reproduce the original movements provided they are augmented by linear stabilizing terms. The overall system provides a novel method of modeling human dynamics and provides a new perspective on the way the body formulates and solves the general problem of controlling its movements.

• Reference PDF Article (Rao, R.P.N. and D.H. Ballard,(1999), Predictive coding in the visual cortex: A functional interpretation of some extra-classical receptive-field effects, Nature Neuroscience 2, 1, 79)
• Slideshow Click here to view
• Lab Website http://www.cs.utexas.edu/~dana/