CS547 Human-Computer Interaction Seminar (Seminar on People, Computers, and Design)
Fridays 11:30am-12:30pm PT · Gates B3 · Open to the public|
David Glowacki
University of Bristol||Stanford
A GPU-Accelerated Software Framework for using 3D Depth Sensor Data to Drive Interactive Physics Simulations
October 3, 2014
Advances in computational power have led to a considerable expansion in the role of computational/simulation methodologies in the physical sciences. The parallel development of new human-computer interaction technologies heralds an exciting new horizon for simulation: namely, is it possible to exploit new optical sensing technologies to imagine novel types of augmented-reality interactions with physics-based simulations? In this talk, I will describe a recently developed interactive system that utilises commodity 3d depth sensors in order to interpret the human form as an energy perturbation, which is subsequently embedded in a real-time molecular dynamics simulation. The net result is an augmented-reality system where humans can use their 'fields' to chaperone the motion of atoms in a rigorous molecular simulation. Efficient 3d image manipulation and physics calculations rely heavily on GPU software acceleration with the OpenCL parallel programming language. This has allowed us to achieve our target of 60 frames per second (FPS), giving an extremely fluid interactive experience. GPU acceleration has also allowed us to scale the system up for use in immersive 360 spaces with an array of up to ten depth sensors, allowing several users to simultaneously interact with a molecular physics simulation. Initial tests run with this system indicate that users have been able to accelerate simple molecular rare event dynamics by 3-4 orders of magnitude compared to standard 'brute force' simulation methodologies, raising the prospect for further development into a platform for citizen-science type applications and next-generation immersive digital education. |
|