Огляд засобів та можливостей високопродуктивного моделювання в UNREAL ENGINE

Abstract

Modern robotics is transitioning to virtual environments like Unreal Engine 5 to reduce costs and enhance testing safety for autonomous systems. This study examines the integration of UE5 with ROS and technologies like Lumen and Nanite to create high-fidelity simulations for robot training. The use of Procedural Content Generation and domain randomization helps bridge the "reality gap" by providing diverse and realistic synthetic datasets. Ultimately, combining UE5’s visual realism with physics engines like MuJoCo offers a powerful framework for advancing navigation and SLAM algorithms