First published July 1 2016
Support Vector Machine and Spiking Neural Networks for Data Driven prediction of crowd character movement
Israel Tabarez-Paz, Isaac Rudomin, and Hugo Prez
Microscopic crowd simulation usually uses ad-hoc models. While these have been proven to be useful, they are difficult to calibrate and do not always reflect real behaviour. For this reason we propose a machine learning approach using neural networks. The main contribution of the project is a first exploration of prediction of agent trajectories using two specific types of neural networks, Support Vector Machine (SVM) and Spiking Neural Networks (SNN).