Introducing Agent Personality in Crowd Simulation Improves Social Presence and Experienced Realism in Immersive VR

Area of Science:

• Computer Science
• Virtual Reality
• Artificial Intelligence

Background:

• Crowd behavior simulation is crucial for diverse applications like gaming, urban planning, and safety training.
• Existing simulations often lack realistic agent personalities, limiting immersion and believability.
• Immersive virtual reality (iVR) demands more sophisticated and engaging crowd dynamics.

Purpose of the Study:

• To propose a novel, lightweight mesoscopic system for personality-based crowd simulation in iVR.
• To model agent personalities using the Big Five (OCEAN) framework and enable autonomous aggregation.
• To enhance user experience in iVR by increasing social presence and perceived realism.

Main Methods:

• Developed a mesoscopic simulation system integrating the Big Five personality framework (OCEAN).
• Employed machine learning-based clustering for agent aggregation based on OCEAN traits.
• Utilized a probabilistic approach for agents to externalize personality through animations.
• Designed a flexible system using graphs for agent and map topology data.

Main Results:

• A within-subjects study with 40 users compared the personality-based system against a non-personality system.
• Results indicated that personality-based crowd simulation significantly enhances social presence in iVR.
• Users reported a higher experienced realism and a better perceived match between agents and the virtual environment.

Conclusions:

• Personality-based crowd simulation is a viable approach to improve user engagement in iVR.
• The OCEAN framework effectively models synthetic personalities for autonomous agents.
• The proposed system offers flexibility and enhances the overall immersion and realism of virtual environments.