Crowd simulation is concerned with creating models of how groups of people move. This Dissertation is focused on creating a model of dense crowd flow through deep reinforcement learning. At the time of writing, to the best of the authors knowledge, no one has simulated dense crowd flow with deep reinforcement learning. The model was developed in Unity using the package ML-Agents.
This project was submitted in partial fulfilment of the requirements for the degree of MComp (Hons) Computer Science at the University of Sheffield by William R Shaw. The paper is available here (FILL).