Our research focus is shifting from the perception of optic flow to the visual control laws for human locomotion. The general program is to determine the visual information used to guide walking toward (a) goals and obstacles that are either (b) stationary or moving. We try to model steering behavior with simple dynamical systems.

Our results thus far support a "mixed" model in which redundant information is linearly combined in task-specific control laws of the form:

where x is a control variable and the i's are redundant informational variables that depend on the current state of x. This insures robust behavior under varying environmental conditions. These control relations can be embedded in a larger dynamical model of steering, obstacle avoidance, and route selection.

For example, we find that optic flow is a dominant type of information for walking toward a stationary goal, although other variables (e.g. egocentric direction) also contribute. Motion parallax (and the strategy of nulling motion parallax) appears to be central to a range of tasks, and its neural support should be pursued. On the other hand, walking toward a moving target seems to be controlled by its egocentric direction with respect to the locomotor axis. We plan to investigate the specific flow variables (FOE, motion parallax, flow speeds) that control steering behavior, and how such task-specific control strategies interact in more complex situations.