Summer Term 2019

Project group DragRace

September 30, 2019, 2:00 pm, F1.110

Dynamic Rendering And Generation of Real-time Adaptive Complex Environments

In this presentation we will give an overview over the work of the DRAGRACE project group. We have created a 3D-environment generator capable of dynamically generating randomized landscapes during runtime in virtually arbitrary size featuring among other things cities, different vegetation and various terrain types. The generator has a server component enabling multiple users to explore these landscapes at the same time. Our work has been implemented using the PADrend platforms to provide easily extendable rendering functions.

(Final presentation)

Ansgar Mährlein

August 14, 2019, 2:00 pm, F1.110

Performance Optimization for FlightSimulation through CullingAlgorithms

In the final presentation of my master thesis, i will present occlusion culling algorithms for flight simulation. Scenes encountered in flight simulators are usually quite flat with a viewpoint high above the scene. However, when simulating low enough flight in alpine regions, the undulating terrain can provide a high level occlusion. Additionally, the three-dimensional cockpit model of some simulators can also occlude large parts of the scenery. I will present two algorithms that take advantage of the occlusion in the two aforementioned cases, and show that performance gains of up to 400% are attainable in flight simulation use cases in mountainous regions.

(Final presentation Master's thesis)

Sören Möllers

May 08, 2019, 14:15, F1.110

Effiziente prozedurale Generierung von Flüssen und Seen mithilfe lokaler Simulation von Erosion

Prozedural Generierung virtueller Landschaften ist für viele Computerspiele und Simulationen essentieller Bestandteil. Oft werden Erosionsprozesse von Wasser simuliert, die Flüsse und Seen produzieren und die Landschaft natürlicher wirken lassen. Diese Methoden bauen auf regulären Gittern als Repräsentation der Landschaft auf und führen unnötige Berechnungen aus auch in Bereichen, die nicht von Erosion betroffen sind. Um diese aufwändigen Simulationen zu beschleunigen, wird eine Methode vorgeschlagen, die die Simulationsgitter in einer hierarchischen Datenstruktur, genauer einer Art von Quadtree, speichert und so die Simulation auf Bereiche starker Erosion beschränkt. Dabei passt sich der Algorithmus adaptiv der Verteilung von Wasser an und erhöht in entsprechenden Regionen die Auflösung der Landschaft. Die Evaluation zeigt, dass eine solche adaptive Datenstruktur die Perfomance in einigen Situationen tatsächlich verbessern kann ohne die Simulationsqualität merklich zu mindern.

Simon Pukrop

May 08, 2019, 14:45, F1.110

Scheduling Algorithms for Multi-Operation Jobs with Setups on a Single Machine

In this final presentation of my master thesis, we will have a look at a realistic, but not much researched, scheduling setting. In the problem class we consider composed jobs represented by a set of operations, which themselves are ordered into different families. When scheduling operations of different families back to back one induces additional costs in the form of a setup time. We describe the quality of a schedule by its total completion time. I will present and explain our new (and first known) approach that yields polynomial time constant factor approximations for this NP-hard problem.