Semantic crowds

Recent advances in crowd simulation techniques have led to increasingly realistic agent and group behavior. As many crowd simulation solutions typically target only specific types of environments and scenarios, numerous special-purpose methods and systems have emerged that are unsuitable for other contexts. Solving this situation demands a higher-level approach that takes re-use and re-configuration of crowds as a priority, for adequate application in a broad variety of scenarios, virtual environments and interaction with the entities present in that environment. In this article, we propose semantic crowds, a novel approach that allows one to specify and re-use the same crowds for virtually any environment, and have them use the objects available in it in a meaningful manner. To have the agents autonomously interact within any virtual world, we avoid in them explicit object-related information. Instead, this knowledge is stored in the objects themselves, which can then be queried, according to an agent’s needs. To facilitate creating such crowds, we developed an interactive crowd editor that provides high-level editing parameters for defining crowd templates. We illustrate the flexibility of semantic crowds by means of three cases, in which we let the same crowd populate quite differently configured airport terminal environments. These examples also highlight that this modular approach easily combines with your custom implementations of agent behavior model and/or motion planner.