Planning for and managing crowds is critical to successful design and operation of public spaces. A great deal of time and effort is spent understanding and predicting how crowds navigate in their environment. SpirOps AI R&D is involved with researchers and companies to build a full collaborative crowd simulation workflow.
This tool will allow designers, architects and engineers to collaborate on a common platform and help them take into account future users in the design of environments. Crowd simulation tools will offer the possibility to explore and test multiple configurations before committing the resources to change the physical environment, and ensures a more effective use of construction spending.
Applications include layout design for train stations, airports and transportation hubs, exit strategies for stadiums, emergency evacuation analysis, outdoor entertainment management, etc.
The key idea of our collaboration is that each member shares the results of its research in order to build a common generic platform and set of behaviors. Companies and researchers interested in this project and its applications are welcome. Please contact us for more information.
CURRENT R&D COLLABORATORS
- A famous worldwide theme park company; they presented the simulation platform at the CASA 2010 Crowd Simulation workshop (paper).
- RATP, the major transit operator responsible for public transportation in Paris and its surroundings
- NURC, one of the three research and technology organisations in NATO
- SpirOps, a Private Scientific Research Lab focused on Artificial Intelligence
- Accurately reproduce real world crowds behaviors using autonomous agents
- Predict human behaviors in new scenarios
- 3D Real-time simulations for large crowds
- User-friendly interface
- A common platform for engineers, researchers, designers, architects and managers
CURRENT TOOL STRUCTURE AND FEATURES
AUTONOMOUS AGENT DESIGN
- An agent perceives its environment and reacts to it depending on its own logic and personality
- Crowd patterns emerge from the combined behavior of many agents
- Existing behaviors: navigation, static and dynamic collision avoidance, queuing, families, etc.
- Current research: advanced density detection, realistic perceptions, etc.
- Using the SpirOps AI Editor, behaviors are designed independently from each other,
facilitating addition of new ones
- Create and modify environments in a user-friendly interface
- Populate and tag the different elements
- Initialize simulation settings
- Accessible areas, pathfinding informations, elevations, etc.
are automatically computed by the SpirOps Path Generator
- Real-time OpenGL 3D rendering of the simulation
- 1000+ characters on a single core
- Go back in time and replay situations (deterministic)
- Display and edit agent and target attributes in real-time
- Export data for further analysis (.csv)