Workshop on the dynamics of
perception and action

Prof Bill Warren will lead a 3-hour workshop on Thursday 27th (2-5 pm) prior to the Welcome Party. The workshop will offer attendees the chance to gain valuable hands-on insight into research on perception-action coupling, and how this links in with the dynamics of movement. All attendees are welcome to participate in this Workshop at no added cost.

Workshop on the dynamics of perception and action:
William H. Warren

Department of Cognitive & Linguistic Sciences
Brown University,
Providence, Rhode Island, USA

How are perception and action organized to yield adaptive behavior? This workshop will discuss the organization of action, the information available for control, and models of perceptual regulation. We will review several basic models, including the cognitive approach emphasizing representation and planning, the cybernetic approach emphasizing linear closed-loop control, and the dynamical approach emphasizing stability and nonlinearity. The issue of separate "what/how" systems for perceptual recognition and visually controlled action will also be raised. We will discuss the organization of locomotion in the form of stable gaits and gait transitions, with free parameters for regulation. The information available in optic flow for locomotion and interception (e.g. catching) will be examined in some depth, including a hands-on computer demonstration of flow patterns.

I will bring these threads together by developing the view that adaptive behavior emerges from the interaction between an organized agent and a structured environment, represented as a pair of coupled dynamical systems. On this view, action is a function of information, according to laws of control, and reciprocally, information is a function of action, according to laws of ecological optics. The resulting behavior corresponds to attractors in the dynamics of the coupled system. Thus, to understand how perception and action produce adaptive behavior, the task is to characterize the system-level dynamics and show how the coupling laws give rise to them. Several case studies will be presented, including infant bouncing in a "jolly jumper," visually controlled braking, and walking to a goal. This view suggests that, rather than acting like a centralized controller, the nervous system exploits informational and physical regularities to stabilize the dynamics of the agent-environment system.