Social Crowds Using Transactional Analysis

More and more applications are relying on simulated crowds to populate films, games, and architecture. Decades of work in this area have produced agents that deftly avoid collisions, but the crowds still look stiff and false because agents do not socialize naturally with each other. On the other hand, ours is a new, expressive algorithm for adding social dynamics to crowds that breathes a new dimension of realism into simulations. Unlike previous approaches, our work allows agents to have multiple social encounters with other agents. We correctly allow interactions to evolve as time passes using the psychological area of transactional analysis. Additionally, we break from previous paradigms since we do not tie our approach to a specific obstacle avoidance algorithm. Instead our algorithm has a flexible architecture that will run with almost any obstacle avoidance algorithm. Finally, we allow for artist direction in our simulations, including bi-modal crowds and social environments that can be changed in real-time. Our results show that our social crowd algorithm runs in real-time with up to 4,000 agents with far more realistic behaviors than previously simulated.