Dynamic field of view as collision detection for autonomous multi-agent

Abstract

In this paper we present a hybrid collision detection technique in crowd simulation by constructing a field of vision for each agent. The field of vision is represented as a bounding volume that is dependent to the agent's locomotion variation that results in a variation of its length and angle. The variation will hence create more dynamic execution and selective toward agent's collision testing between object and its field of vision. In many crowd simulation behaviors such as flocking, agents are considered as a whole summation of force value that determine the agent's behavior and decision. This is similar to what happens in the real world; human and animals behave and react upon what they perceive. This research presents agents' unique perception based on their own speed variation thus producing more dynamic and selective collision response execution. This technique also gives more possibility to design conceptual bounding volume that represent agent's field of vision based onlinear line intersection.