Repopulating the Colosseum

This study explores the application of artificially intelligent (AI) virtual agents to simulate the movement of spectators through a high-quality digital reconstruction of the ancient. Building on the foundational work of Gutierrez et al. in "AI and Virtual Crowds: Populating the Colosseum," this research advances the understanding of crowd dynamics within this ancient spectacular structure. The simulation began by placing nearly 8000 virtual agents into one-quarter of the reconstructed virtual environment. Analysis of agents’ behaviors as they navigate to their respective seating galleries revealed that congestion within the structure noticeably changes when the class and gender composition of the crowd is taken into account. This can be simulated by varying agent priorities, where high priority agents can physically displace lower ones. In effect, such modulation is a mode by which some of the original cultural context of the structure can be simulated, i.e., the social identity of the spectators. The simulation also generated an alternative scenario wherein the agents were instantiated over a more realistic one-hour time frame and with a more historically accurate number of agents. Despite the larger number of agents under these conditions, the crowd density in this simulation noticeably decreased, and some social factors typically present in denser crowds were mitigated. The simulation itself was generated in the popular 3D development platform, Unity, and utilized its built-in AI navigation system. This was done to both modernize the simulation environment and also increase the ease of repeatability. The AI techniques used to simulate each agent’s behavior come from this out-of-the-box Unity system which operates with, among other things, the A* algorithm and the Reciprocal Velocity Obstacles (RVO) algorithm. This article highlights the continued potential of AI-driven simulations to offer insights into ancient crowd dynamics and reconstructive archaeology and concludes by discussing the future direction for research around AI simulations for cultural heritage and its allied disciplines