Short video of crowd simulation in the KTH-environment

Video of crowd simulation

Design of the database

Deriving the the matching function

Short explaination of how the update of the agents trajectories will occur

Pseudo code showing how a trajectory is assigned to an agent q

First xml-database prototype

A first program to write down the data to an xml-database is completed.

The picture below contains the first few lines of the database containing only 8 example situations.

The database for the first traced video textfile gives around 1800 example, takes about 15 seconds to create and is ~90MB large.

Structure of database

• To make example-based simulations a database of traced video footage is needed.
• The database will consist of example situations we have observed.
• The examples contain a subject and the influencing factors (other agents) affecting it.
• The properties of the agents are saved for every frame in the example (40 frames per example). 
• The plan is to use XmlSerializer in Unity to save down the data from each example to an xml-document. 
• The xml-document is planned to have the following structure:

Recreation of "Crowds-by-Example" dataset video