Controlling steering behavior for small groups of pedestrians in virtual urban environments.

摘要

The central focus of this thesis is exploring pedestrian steering behaviors for small groups in virtual urban environments. I investigate how individuals and groups can be modeled in order to produce plausible collective walking behavior, including moving and standing formations. Control of walking behavior is distributed among a set of independent, goal-oriented processes, each responsible for some aspect of performance. I propose a framework that integrates these constraints to simulate the walking behavior of individual pedestrians and small groups of pedestrians. This control system is extended to include group navigation and the behavior of pedestrian groups standing around a pathway.; A control system that combines group walking behavior with group standing behavior, including transitions between the two behaviors, is tested with scenarios that demonstrate two-dimensional simulations of pedestrian behaviors on pathways. The experiments begin with simple path following and gradually add complexity in the form of obstacles, pathway constrictions, interactions with other pedestrians, and transitions between behaviors. In these tests simulated pedestrians move along a pathway, while avoiding obstacles and each other. Members of a group maintain a side-by-side formation while walking, and stop together in arc and ring configurations. In some cases results are improved by introducing state to temporarily suspend a behavioral constraint. State is also used to select situationally appropriate constraint sets when pedestrian behavior switches between walking and standing.