We permanently offer topics for bachelor and master thesis projects in all areas of our research field and in related areas. Currently, the research of our group covers the following topics: interactive simulation of rigid bodies and deformable solids, fluid simulation, cutting and fracturing, medical simulation, massively parallel computation on GPUs, game physics, collision detection and real-time visualization. Each thesis topic is usually specified in cooperation with one of our research assistants and/or Prof. Bender considering the student's individual interests and his/her previous knowledge as well as the current research topics of the Computer Animation group. In order to guarantee a successful completion of the thesis, we usually expect our student to have
- Good knowledge and practical experience in C/C++ and object-oriented programming
- Basic knowledge of numerics, algorithms and data-structures
Below you find a (non-complete) list of currently open theses. If you are interested in another research topic, please contact us.
A common approach to simulate deformable objects is to solve the underlying differential equations using the Finite Element Method (FEM). However, when the object is spatially discretized into a mesh, e.g. a tetrahedral or a hexahedral mesh, the boundary of the mesh must be aligned with the boundary of the geometric object. The goal of this master thesis is to develop and implement a method based on the eXtended Finite Element Method (XFEM) that allows us to embed an object boundary into a regular grid without the requirement to align the object's boundary with the grid cells. In this way objects of complex shape or with thin features can be simulated robustly. Working on this project includes implementing a deformable object simulation in C++ using our framework and comparing the newly developed approach with existing methods.
Prof. Dr. Jan Bender
Fluid simulation methods based on the Smoothed Particle Hydrodynamics (SPH) approach have been investigated for several years in computer graphics. Application areas are e.g. the generation of special effects in movies and computer games. For realistic results a large number of fluid particles must be simulated which is computationally expensive. The main bottle neck is the neighborhood search which determines the closest neighbors for each particle. The neighbors of a particle are required to solve the equation of motion of the fluid locally.
The goal of this bachelor thesis is to implement an efficient neighborhood search on the GPU using CUDA which is able to handle millions of particles. This neighborhood search should be integrated in our SPH simulation framework.
Prof. Dr. Jan Bender
The smoothed particle hydrodynamics (SPH) method is widely used to simulate liquids like water. However, SPH is rarely used to simulate gaseous phenomena like steam or smoke. One major challenge is to model interactions between the gas particles and the surrounding air. The standard approach is to simulate the air flow using SPH particles. But in practice it is computationally too expensive to simulate large volumes of invisible air particles. Another challenge is the simulation of highly turbulent motion, which is strongly damped because of numerical errors when standard SPH methods are used.
The goal of this master thesis is to develop and implement a smoke simulation method based on SPH, which simulates only the visible smoke particles (or a narrow band of air particles around them). Further, a micropolar fluid model shall be used to simulate detailed turbulences. Working on this project includes implementing a SPH fluid simulation in our framework using C++ and comparing the newly developed approach with existing methods.
When discretizing an object for simulation two different basic approaches can be identified: the Lagrangian discretization in which the degrees of freedom are located on and move with the object itself and the Eulerian approach which uses a static grid for the simulated degrees of freedom. However, both approaches have pros and cons. When simulating deformable solids the Lagrangian way, problems can occur where objects deform into an invalid inverted state. An advantage of the Lagrangian formulation is that it naturally conserves mass which is a problem with the Eulerian approach. In order to combine the advantages of both approaches while avoiding their problems a hybrid method can be devised: the material point method (MPM). MPM uses particles to ensure that no mass is lost while a special background grid is used for the accurate computation of gradient terms. Since it is a meshless method which uses particles that are not connected in a mesh, typical problems like the aforementioned entanglement cannot occur. A constitutive model can then be defined to model different kinds of materials. E.g., MPM was used with a constitutive model for snow for the snow simulations in the Disney film Frozen.
The goal of this master thesis is to implement an MPM simulation with a simple constitutive model for the simulation of deformable solids. The simulator should be written in C++ using our visualization framework and the results should be compared to existing approaches. This applies the hybrid approach, which has only rather recently been introduced to computer graphics, to the problem of simulating deformable solids which is largely dominated by purely Lagrangian methods.
Smoothed particle hydrodynamics (SPH) simulations deliver visually realistic animations of fluids like water. However, the effect of air-water mixtures like foam, spray and bubbles, which are important visual details, are not handled during fluid simulation. These effects can be added as a post processing step. Foam, spray and bubbles are simulated as particles that move along with the surrounding fluid particles.
The goal of this bachelor thesis is to implement a foam, spray and bubbles simulation using C++ and our framework and to evaluate the results.