Automated Crowd Simulation with Artificial Intelligence

Descripción del proyecto

«Software» de simulación de multitudes para el transporte público

El transporte público afronta retos: pasar del transporte individual al público y cumplir los objetivos climáticos. La ampliación de estaciones es vital, pero renovar las existentes resulta complejo y propenso a errores. Aunque combinar la simulación de multitudes con el modelado de información para la edificación evita costosos errores y requiere más potencia de cálculo, lo cual limita las opciones de diseño. El equipo del proyecto ACSAI, financiado con fondos europeos, emplea modelos de simulación y aprendizaje automático para crear un nuevo modelo. La inteligencia artificial evalúa rápidamente la viabilidad del diseño, ahorrando tiempo gracias a la eliminación temprana de procesos inviables. El equipo del proyecto pretende integrar perfectamente esta solución en el «software» de simulación de multitudes para obtener un producto listo para el mercado. Dicha empresa, fundada por mujeres, se esfuerza por ser un modelo a seguir en una industria dominada por los hombres.

Objetivo

With the transformation of transportation from individual to public transport and the new climate goals of the European Green Deal, public transportation faces major challenges in the coming years. The number of passengers is expected to more than double. Stations must be quickly expanded to accommodate the increasing number of passengers. Remodeling existing stations is more complex than building new ones, is error prone and quickly lead to costs and time explosions.
New methods, technologies, and processes are needed to solve this. Our project Automated Crowd Simulation with Artificial Intelligence (ACSAI) focuses on exploring and combining existing deep tech solutions into a new one.
BIM in combination with crowd simulation enables planners to digitally prototype buildings at early planning stages to prevent costly mistakes. Although BIM speeds the process of data preparation, a high computational effort is needed for the execution. As a result, only a few design variants or scenarios can be analysed as of today.
To overcome this issue, we combine simulation models with machine learning methods: AI Models shall be developed and trained with the help of precalculated simulation data. The trained NN can give answers at the push of a button, whether a design is compliant or not. This accelerates the process enormously since it quickly sorts out designs that are not feasible. Thus, only pre-checked design variants will be simulated for final confirmation and optimisation. The whole approach is a quantum leap in the planning process since it speeds it up and reduces the risk of planning mistakes enormously.
We have tested the approach for railway stations in a PoC environment in the research project BEYOND. The next step is to implement it in our crowd simulation software into a ready-to-market product. As a small company, we cannot afford such big R&D projects on our own. And as a female-founded company we hope to be a role model in a male-dominated sector.