Galactica, a digital planetarium that explores the solar system and the milky way

Abstract

This paper describes a new Digital Planetarium system that allows interactive visualization of astrophysical data and phenomena in an immersive virtual reality (VR) setting. Taking advantage of the Cave Hollowspace at Lousal infrastructure, we have created a large-scale immersive VR experience, by adopting its Openscenegraph (OSG) based VR middleware, as a basis for our development. Since our goal was to create an underlying system that could scale to arbitrary large astrophysical datasets, we have splitted our architecture in offline and runtime subsystems, where the former is responsible for parsing the available data sources into a SQL database, which will then be used by the runtime system to generate the entire VR scene graph environment, for the interactive user experience. Real-time computer graphics requirements lead us to adopt some visualization optimization techniques, namely, GPU calculation of textured billboards representing stars, view-frustum culling with octree organization of scene objects and object occlusion culling, to keep the user experience within the interactivity limits. We have built a storyboard (the “Galatica” storyboard), which describes and narrates a visual and aural user experience, while navigating through the Solar System and the Milky Way, and which was used to measure and evaluate the performance of our visualization acceleration algorithms. The system was tested with an available dataset of the complete Milky Way (including the solar system), featuring 100.639 textured billboards representing stars and additional 104.328 polygons, representing constellations and planets of the solar system. We have computed the frame rate, GPU traverse time, Cull traverse time and Draw traverse time for three visualization conditions: (A) using standard OSG view frustum culling technique; (B) using view frustum culling with and our octree organizing the scene’s objects; (C) using view frustum culling with our octree organizing the scene’s objects and our occlusion culling algorithm. We have generally concluded that our octree organization and octree plus object culling techniques out-performs the standard OSG view frustum culling, when around half or less than half of the dataset is in view of the virtual camera.