Possibilities of agent-based behavioral modeling in practice – ITMO University projects

We talk about the tasks that the staff of the National Center for Cognitive Development of ITMO University (NCCR) – this is an analysis of passenger flows at the Olympic Park Station, and visualization of the reconstruction project of the Stables yard in St. Petersburg.

What is agent modeling

This is one of the methods. simulation modeling. Within its framework, they simulate the behavior of autonomous agents (they can be people, cars, animals, etc.) and evaluate their impact on the state of a large system. Agent-based modeling is used to optimize logistics, manage human resources and predict the movement of human flows, and not only.

Several such projects were implemented by the interactive visualization laboratory in the NCCR.

Stables Courtyard Reconstruction Project

We performed it together with the company “START Development”. She participated in a project competition for the reconstruction of a complex of buildings on Konyushennaya Square. Engineers did not just want to prepare drawings, but make an interactive presentation of possible changes. Our task was to show how automobile traffic and pedestrian flows will be transformed. Now the area is overloaded: it has a lot of parked cars. The model “after reconstruction” shows that it becomes completely pedestrian, additional routes, including tourist ones, appear, a direct passage to the Savior on Spilled Blood arises and new public spaces are formed. But, unfortunately, the START Development project was not approved.

By the way, the reconstruction of the Konyushenny Courtyard is not the first project that we worked on with the specialists of START Development. In 2017, we developed for them, a digital model of the development scenarios of the satellite city “Yuzhny”. The system calculates the main characteristics of the urban environment in dynamics. It takes into account various scenarios for the arrangement of territories: from the development of the road network to the formation of the ecological picture (noise, CO2). The simulation results make it possible to justify the choice of infrastructure organization option.

Analysis of the Olympic Park Station

We solved this problem for the Studio 44 architectural bureau, which designed large infrastructure facilities, including the Ladoga Station and the Olympic Park. When the station was built in Adler, load calculations for passenger traffic were made general – peak loads, average density per area, etc. All this without detailed modeling accurate to individual people. Colleagues from the architectural bureau decided to check on the basis of mathematical models how much their design hypothesis is confirmed by calculations close to reality.

We considered peak loads in two versions: morning intense delivery of spectators to the territory of the Olympic Park by electric trains, when people move in a large enough crowd towards the stadiums from the station. And the second option is when in the evening all the games and events end, and people massively get on electric trains to leave.

Low level navigation in space and pedestrian avoidance evaluated using the multi-agent modeling framework Menge. But a high-level behavioral model and a traffic model were implemented already on Unreal Engine 4.

The task was to see if there are unobvious pitfalls and “bottlenecks” in the organization of space that lead to excessive crowding. The model showed that the main problem spot is the corners. People slow down on bends, so there were small crowds. But the density still did not reach five people per square meter.

“When we took a deliberately inflated stream of people, the potential for crowding was literally in a couple of places — walkways under escalators and turns. Now there are turnstiles there: due to them, the flow is calming down, and people are not crowding around the corners. ”

– Andrei Karsakov, Ph.D., Senior Researcher, National Center for Cognitive Development, Associate Professor, Faculty of Digital Transformations

In addition, we modeled the operation of the station in the current “post-Olympic” conditions. It is clear that at the end of the game the streams became smaller, more “smeared” during the day. We had no tasks to calculate absolutely extreme situations (for example, in case of fire or blocking of any exit). Nevertheless, ITMO University developments allow this to be done.

Results of work we submitted at the Venice Architecture Biennale 2018. A dynamic model was shown directly above the layout of the station itself.

More projects

Several years ago employees NCCR decided the task of building a safe route for fans from the Zenit Arena stadium during the World Cup in Russia. The model took into account various non-standard situations, for example, heavy rain or the aggressive behavior of the crowd in case of losing your favorite team. Then an action plan was worked out: how to let people out of the stadium, depending on the situation, where to set up turnstiles and arrange “snakes”.

Also in 2017, our engineers, together with colleagues from the Netherlands, the USA and Mexico, have developed PULSE program. It combines multi-agent models that describe the behavior of pedestrians during public events and emergency situations. Moreover, the actions of each virtual participant are controlled by his own artificial intelligence system. The platform is already used to study the flow of pilgrims at the largest religious festival. Kumbh Mela in India. The system helps to understand how to avoid crowding if, for example, it starts to rain and a crowd of people rushes to the temple.

Now we continue to develop tools for modeling walking mobility and traffic. One of the latest projects in this area is devoted to the historical reconstruction of urban areas based on incomplete data. Using the methods of artificial intelligence, we try to restore the picture of the life of the city, even if some part of historical information is missing.

ITMO University has already implemented a digital project “time machines“. It allows you to follow how Kronstadt developed and built up in the XVIII or XIX centuries. We have restored and digitized historical information about the lost objects of the city. Artificial intelligence systems helped generate synthetic data on the quarterly development of the early periods of the city’s development.

“We use multi-agent modeling to reproduce the dynamics of the city’s population at almost any time point in its existence – whether it be the 19th century or the 21st. Not only pedestrian mobility is taken into account, but also many different vehicles: from cars or horse-drawn carts to bicycles. I hope that soon we will be able to publicly demonstrate the first results of our work on this project and tell us more details. ”

– Andrei Karsakov, Ph.D., Senior Researcher, National Center for Cognitive Development, Associate Professor, Faculty of Digital Transformations


Andrey Karsakov spoke about his experience of AR-support of mass events in our podcast “ITMO Research_”. We published a text transcript of the release on Habré (1, 2).


Similar Posts

Leave a Reply