Abstract

The subject of this bachelor’s thesis is the development of a framework for simulating robots and rovers in the Unreal Engine 4 (UE4). The simulation will connect to the robotic framework Robot Operating System (ROS) and provides photo-realistic images for computer vision applications. The well-established hardware abstraction layer interface called ros_control was used to support actuator control in the simulation. To extract sensory information from the photo-realistic virtual environment the rosbridge package was used. The newly proposed framework is a proof of concept for a game-engine-based simu- lator with the Robot Operating System (ROS) to perform computer vision experiments without requiring specialized hardware. It is also possible to use this framework to gather a dataset to train neural networks on or to evaluate existing machine learning techniques. This framework will be evaluated on the example of the European Rover Challenge (ERC). The student team from the Scientific Workgroup for Rocketry and Spaceflight (WARR) already created a simulation, which will be used as reference. The results of this bachelor’s thesis show that the framework works as intended for controlling simple actuators and joints; however, the current physics simulation of Unreal Engine 4 (UE4) does not provide enough stability to simulate the proposed complex scenario without major artifacts. Moreover, the camera plugin used in this thesis influences the physics simulation negatively when parameters are changed to achieve real-time high-definition support. The framework does, however, provide a real alternative to already in-use state-of-the- art solutions, as it enables an easy-to-manipulate robotic simulator with a powerful graphics engine for photo-realistic simulations.