Synthetic Simulated Environments

Abstract

In this chapter, we proceed from datasets of static synthetic images, either prerendered or procedurally generated, to entire simulated environments that can be used either to generate synthetic datasets on the fly or provide learning environments for reinforcement learning agents. We discuss datasets and simulations for outdoor environments (mostly for autonomous driving), indoor environments, and physics-based simulations for robotics. We also make a special case study of datasets for unmanned aerial vehicles and the use of computer games as simulated environments.