The dynamics of perception and action: of bouncing
babies and legged locomotion.
Where does the organization in behavior come from?
I will develop the view that stable behavior emerges
from the interaction between agent and environment, exploiting
physical and informational constraints. This interaction
can be formalized as a nonlinear dynamical system, called
the behavioral dynamics. Two case studies illustrate
this approach.
First, how do six-month olds learn to bounce in a "jolly
jumper"? A longitudinal study reveals that infants suddenly
discover the solution of driving the system at its natural
frequency. This can be modeled as a forced mass-spring
system whose parameters are specified by perceptual information.
Second, how do adults visually guide locomotion through
a complex, changing environment? Locomotor behavior can
be decomposed into four basic components: (a) steering
toward a stationary goal, (b) avoiding a stationary obstacle,
(c) intercepting a moving target, and (d) avoiding a
moving obstacle. We use a virtual environment to study
each behavior and model it as a dynamical system. By
combining these components, we can predict locomotor
paths in more complex environments.
Organized behavior can thus be understood as emerging
on-line from the interaction between a structured environment
and an agent with simple control laws, making explicit
planning unnecessary.