A Study of a GPU Based Real-time 3D Live Controlled Molecular Kinetics Simulation on Multiple Computing Platforms
GUTMANN, GREGGutmann, Greg
In the emerging field of molecular robotics researchers are faced with the problem of creating controlled movement within organic systems. Due to the difficulty in this problem we have been developing a molecular kinetics simulation to aid in the study of the microtubule swarm motion. However creating such a system poses many research challenges, since the scale of the simulation is far greater than past molecular simulations, and new simulation approaches must be taken. This problem is especially difficult due to the scale of the biological systems. Some past works have looked at making general models which have been able to reproduce swarm motion but are not able to elucidate the cause of these swarms. Other works have looked at explicitly simulating microtubule motion, but due to performance limitations have not been able to verify the creation of swarms based on their model. In order to combine the strengths of these two approaches, large scales as well as simulating the interacting elements of the biological system explicitly, we need to face challenges such as the immense amount of data which is produced, computational demands, as well as memory demands. In addition platform flexibility and usability must be addressed; otherwise, the impact of the work may be limited by hardware demands for the users or ease of use. We have solved these problems by creating a real-time live controlled 3D molecular kinetics simulation. Choosing a real-time 3D simulation helps to reduce the amount of data produced, by reducing storage demands and post processing or rendering. However, choosing real-time facilities further increases the performance demands by giving a very small time budget per update. To solve this, we have chosen to use general purpose graphical processing units which offer a high level of computational performance. We have also developed specialized algorithms designed to efficiently manage the memory demands alongside of the computation. Lastly, we have applied these methodologies in various ways to ensure platform flexibility and usability for future studies with the simulation. By using the proposed methods formulated throughout my graduate studies we have been able to reach the needed performance for simulating large-scale molecular swarms explicitly in real-time. In addition to adding to the emerging field of real-time visual simulations, this work has also created a tool for scientists to test and observe molecular kinetics based interactions, specifically microtubule gliding assay swarms. This will open a door for future research and study of molecular swarms. The scope of my graduate studies has been the creation of the entire simulation software and model.