16 January. 2021
Several techniques have been proposed to simulate traffic in urban environments. Many of these rely on simplified models that neglect important features of real-world situations. As part of our quest of improving realism in both simulation and visual fidelity of animated scenes, we will investigate ways to animate and/or simulate traffic using data from various sources such as cameras and GPS data. These will include methods such as Deep Learning (GANs, RNNs), Reinforcement Learning and Texture Synthesis. These kinds of simulations can then be used to improve the quality of animations for various applications such as movies, games and training environments for autonomous driving.