New methods of render-supported sensor simulation in modern real-time VR-simulation systems

Abstract

In recent years, VR-Simulation Systems have taken up an essential part in engineerers' and reserchers' daily work. For the close-to-reality simulation of entire environments the term 'Virtual Testbeds' has been coined. In a Virtual Testbed, complete mission scenarios are simulated as a whole instead of focusing on a variety of details. Sensor simulation is an important aspect in many envidaged simulation scenarios dealing with robotics. In this paper, we focus on a novel sensor simulation approach of optical sensors supported by the render module. By using data and techniques originally developed for 3D rendering, simulation results can be improved whilst simultaneously increasing the performance significally, which is essential for real-time simulation tasks. Beside the render-relevant data, our sensor simulation approach does not require any additional data to archieve high accurate results. Our approach is integrated into a modern VR-Simulation-System. Thus, different Virtual Testbed scanarios ranging from the simulation of wood harvesting processes in simulated forests up to the virtual prototyping of space exploration robots, are presented to emphasize the synergy between the sensor and render component.