Abstract
Question: Do steering adjustments depend on the perceived distance or time-to-contact with an obstacle? Our locomotor dynamics model (Fajen & Warren, JEP:HPP, 2003) currently assumes distance as an input variable but could alternatively use time-to-contact, specified by the relative rate of optical expansion. To dissociate these hypotheses we independently vary physical walking speed and the visual gain in a virtual environment. The manipulations produce similar optical speeds but only the physical walking speed affects the inertial properties of the body.
Design: The experiment is performed in an ambulatory virtual environment (10m x10m) with a head-mounted display (60° H x 40° V) and a sonic/inertial tracking system (50–70ms latency). Participants are instructed to walk at three speeds (2/3, 1 or 3/2 of preferred walking speed). These are partially crossed with three visual gains (2/3, 1, 3/2), such that the optical motion in the display is slower, matched, or faster than the actual walking speed. Participants walk to a stationary goal (6m straight ahead) around an obstacle whose position varies across trials (initial distance of 3 or 5m, 3° to the left or right of the straight path).
Results: Using distance as the control variable, the model predicts that the agent will turn later as speed increases, whereas using time-to-contact it predicts the reverse. The dependence of steering behavior on the body's inertial properties can also be assessed. Model predictions are evaluated against the human data to empirically determine whether human obstacle avoidance is controlled by distance or time-to-contact.