Social crowd controllers using Reinforcement Learning methods

Master thesis data
Specialization: VRI
Thesis advisor: Nuria Pelechano
Orientation: Research
Student: Not assigned

Thesis Description
The aim of this project is to create controllers for a crowd simulator that achieve multiple dynamic obstacle
avoidance. The controller is for each individual member of the crowd and considers all other members as obstacles.
The main aspect that will be studied for this project is how to parameterize the generation of controllers in order to
create a crowd with members that exhibit a variety of social behaviors using reinforcement learning to achieve
realistic crowd simulation capable of modeling real social phenomena. This project combines elements of computer
graphics, animation, reinforcement learning and intelligent avatar design.