Abstract
We are investigating the behavioral dynamics of visually guided locomotion in complex, dynamic environments. Previously, we developed a dynamical model of how people walk to stationary goals, avoid stationary obstacles, and intercept moving targets, based on human experiments in a virtual environment. Here we examine how people avoid moving obstacles. In the model, the direction of a goal acts as an attractor of one's heading direction, whereas the direction of an obstacle acts as a repellor of heading. With a moving target, change in the target-heading angle is nulled, creating an attractor a constant angle ahead of the target; a moving obstacle may be avoided by treating this as a repellor. The path of locomotion is the resultant of all of the forces acting on the agent at each instant. In the present experiment, participants walked to a goal while avoiding a moving obstacle, whose initial position, speed, and trajectory were varied. Testing was done in the VENLab, a 40 × 40 ft virtual environment. Participants wore a head-mounted display (60 deg H × 40 deg V), while head position was recorded with a hybrid sonic/inertial tracking system (50 ms latency). We analyzed the observed path and the time series of obstacle-heading angle to assess the conditions under which participants cut in front of or pass behind the moving obstacle. The aim of this research is to identify and model the locomotor “rules” for an individual human agent. This may allow us to predict interactions between people as well as crowd behavior in more complex situations. Locomotor paths can thus be shown to emerge on-line from the interactions between an agent and a structured environment.