Academic Projects
Some of my projects done in college:
dSpace - A real-time ray-tracer
This was my team project for the final year of my undergrad course. We built a real-time ray tracer to run on a Beowulf cluster of upto 8 nodes. Our work was based on the paper Ray tracing deformable scenes using dynamic bounding volume hierarchies by Ingo Wald et. al. Although their methodology is primarily well suited for deformable scenes with animations we only looked at implementing it for static scenes and achieving interactive frame rates. Due to time constraints certain optimizations, like using SSE to speed up ray testing, could not be implemented but eventually we did manage to achieve frame rates of 25+ for around 500 polys at a resolution of 512x512 which included texturing with bilinear filtering and some post-process effects.
An early screen capture showing the RT server control panel on the left along with the output window on the right.
GPU based infinite procedural terrain
GPU terrain
This was a small project for my Interactive Graphics course where I render an endless terrain entirely on the GPU. A regular grid mesh is passed to the GPU and a vertex program offsets each vertex based on a heightmap that is generated procedurally each frame using Perlin noise. I also added some water and fog to make it look better.
The transition from one heightmap to the next isn't perfectly smooth and you can notice jitters if you observe. There's a simple fix for this- use bilinear filtering for height value lookup for each vertex based on the lateral displacement. Don't ask my why I didn't add it in that case. :)
The transition from one heightmap to the next isn't perfectly smooth and you can notice jitters if you observe. There's a simple fix for this- use bilinear filtering for height value lookup for each vertex based on the lateral displacement. Don't ask my why I didn't add it in that case. :)
| Demo and source |
PredPrey (ongoing)
My current research as a Graduate RA at Purdue under Prof. Christoph Hoffmann is about developing a new gameplay concept for MMORPGs based on natural evolution of species. We're trying to build a system that will offer an interesting gameplay with good replay value. The initial stage of the work involves building mathematical models to represent predator-prey interactions in an ecosystem with certain constraints. There are numerous models that already exist, like for example the Lotka-Volterra model, but we're trying to adapt one to be used in the game that would help the AI understand the current state of a game and adjust settings to keep the game interesting.
PredPrey simulation
