Abstract
Our steering dynamics model can accurately simulate human locomotor behavior with respect to environmental objects: stationary goals and obstacles (Fajen & Warren, 2003), moving targets (Fajen & Warren, 2007) and moving obstacles (Cohen, Bruggeman, & Warren, VSS 2005). Thus far, these objects have been non-intentional physical entities moving on fixed trajectories. The aim of the present research is to extend our investigation to intentional, interactive agents. Specifically, we seek to model situations in which two intentional agents interact in a range of pursuit and evasion scenarios. Ten pairs of participants interacted with one another in 4 different scenarios: (a) target interception, (b) moving obstacle avoidance, (c) mutual evasion, and (d) pursuit-evasion. In general, one participant acted as a pursuer while the other acted as an evader en route to a goal; in the mutual evasion scenario both participants avoided one another while en route to separate goals. Head positions were recorded using an inertial/ultrasonic tracking system (IS-900). Surprisingly, the steering dynamics model generalized to the interactive scenarios simply by linearly combining existing components for target interception and moving obstacle avoidance. This finding demonstrates the generality of the constant bearing strategy that underlies both components, and indicates that evasion does not require a qualitatively different strategy. Small parametric adjustments improved the model fit, shedding light on subtle differences in locomotor behavior elicited by intentional and non-intentional objects. Future research will investigate this difference by manipulating the intentional properties of targets and obstacles in both real and virtual environments.
Supported by NIH EY10923.