SpirOps Crowd

Populate your 3d decors with thousands of virtual pedestrians in a few hours.

One Artificial Intelligence for each pedestrian

SpirOps has been developing since 2007 a crowd simulator in which each agent has their own desires, their personal characteristics and a subjective knowledge of the world.

We use sociological research to analyse pedestrian behaviours. We reproduce their motivations to ensure that our behaviours are as realistic and repeatable as possible in various environments.

Our AI Engine allows us to improve and enrich the behaviours depending on our partners' needs.


To reach their destination, agents don't simply follow the shortest path but also:

  • don't hug the walls,
  • avoid obstacles (other agents, poles),
  • free their field of view (by slightly shifting to the side for instance),
  • pay attention to their personal space.

Friends and families

A crowd isn't solely composed of single individuals. That's why we also simulate behaviours of groups.

  • the members of the same group mutually wait for each other,
  • each member of the group moves to stay within talking distance to other members of the group,
  • individuals of different groups respect politeness rules : they don't disturb a group of friends talking or don't separate children from their parents.

Interactions with the environment

Pedestrians don't simply navigate, they also interact with their environment. Thus, they can:

  • wait in an organised or disorganised queue (respectively ticket booth / movie theater entrance),
  • planify a route requiring resources (train tickets, etc.),
  • satisfy needs (hunger, toilets, etc.),
  • spectate events (street performances, incidents, etc.)

Specific Structures

An environment can contain conveyor belts, trains, escalators or elevators.

For our partners, we have implemented some of these complexe structures like the crowded Eiffel Tower elevators or a double floor train in which people can sit.

It is possible to modify in real time multiple parameters (speed, doors opening time, etc).

  • How to create a simulation?

    Our optional integration into Autodesk Maya makes it easy to create and run simulations.

  • Import your raw 3D decor

    We support any 3D file format that Maya can open.
    We automatically compute where pedestrians can walk in your decor.

  • Annotate your environment

    Our Maya tools allow:
    - to place waiting lines, doors and signs,
    - describe the pedestrians (characteristics, entry and exit points).

  • Visualize your crowd in 3D

    Our 3D viewer allows to visualize and interact with the crowd:
    - from any points of view,
    - in real time, fast-forward, backwards,
    - with information on each pedestrian.

  • Export and Analyse

    It is possible to export data per pedestrian, interest points (waiting lines, etc.) or by zone in order to analyse them.

Train station simulation

Our tools allow us to simulate a large-scale train station at rush hour.
You can easily change the train station parameters and get a large amount of datas.


  • travelers entries rates,
  • boarders destinations,
  • type of vehicles,
  • train schedule (depart and arriving time, mission),
  • loads for each wagon and for each train,
  • elevator direction,
  • localisation of ticket machines,
  • number of gate lines.


  • number of boarders per carriage,
  • density maps (based on Fruin Level of Service),
  • use map,
  • train/platform exchange time,
  • entering and exiting flow rates for each:
    - gate lines,
    - escalators,
    - stairs.

Specific train behaviours

On top of their generic behaviours, a simulated traveler is capable of:

  • choosing a spot on the platform according to different strategies,
  • getting on and off a train,
  • sitting down in the wagons according to their preferences,
  • running to not miss his train.

Thesis (in French): http://www.theses.fr/2019CNAM1227

Positioning on the platform

The positioning on the platform of the travelers is extremely important if we wish to study the density on the platform, the safety, or how long it takes for travelers to get on and off the train.

Sociological studies show that this positioning on the platform relies on several parameters such as:

  • the platform width,
  • how escalators and stairs are placed,
  • train stations exits of the travelers destinations,

SpirOps implemented and calibrated the most common positioning strategies through real life observations. The behaviours were implemented in order to be repeatable between different train stations.

Train/platform exchange

The same work was done for the exchange between the train and platform which also relies on numerous adjustable parameters:

  • the vehicles' load,
  • the amount of people going on and off the train,
  • the type of vehicles (door width, amount of seats),
  • the politeness of travelers.

Urban Simulation

In order to simulate pedestrians and vehicles in a bustling city, we use a large amount of open data such as:

  • a HD Map for roads,
  • sidewalks / crosswalks shapes,
  • traffic lights / road signs,
  • subway entrances / shops,
  • urban furniture (benches, trees, poles, street lamps, etc.).

Specific behaviours for pedestrians have been implemented in order to:

  • favour sidewalks and crosswalks,
  • pay attention at traffic lights,
  • avoid vehicles obstructing their way.

Integration in Gazebo

Gazebo is an open source robotic simulation software.

Our plugin allows our partners to create in Gazebo a crowd capable of interacting with simulated robots.

Our partners can thus improve the behaviour of their robots in a densely populated environment.



Ile-de-France Mobilités IDFM
Cité des sciences