More realistic, flexible, and expressive social crowds using transactional analysis

Abstract

Recent algorithms have been able to simulate "social crowds" that allow agents to interact socially as opposed to only treating other agents as obstacles. Unfortunately, past social crowd algorithms lack realism and flexibility because they do not allow agents to move in and out of different and repeated social interactions, are built around a specific obstacle avoidance algorithm, or are tuned only for a specific social setting and do not allow for artist directed changes. We propose a new, simplified social crowd algorithm that focuses on the evolving social needs of agents and allows each agent to join and leave different encounters as desired. Our algorithm is based on the psychology research area of transactional analysis, does not require a specific obstacle avoidance algorithm, and allows for easy artist direction for determining the precise social environment being simulated. Our algorithm runs in real-time with 3,000 to 4,000 agents without the restrictions of previous research. © 2012 Springer-Verlag.