Rosen Diankov (出杏光魯仙) | About me

Ever since the beginning of high school, I've been madly programming 3D graphics and simulation engines. The fact that there are no limits to software really drove me to develop some pretty remarkable games and game development tools. After my first year at UC Berkeley, this graphics passion refocused itself on AI, computer vision, and real time systems. Although graphics were still exciting, in the end the product just appeals to the eye. Good graphics research is usually judged by how real or pretty it looks and how fast it renders. The former highly depends on the audience and usually programming hacks can be used to fool the eye. The latter is highly dependent on the amount of time you spend on the problem; it is determined by how well the hierarchies, partitioning structures, and caching algorithms are written. AI and computer vision problems are much more challenging, they cannot be hacked away like graphics problems.

Some fields in robotics, it turns out, have very similar problems to graphics and geometric reasoning. In particular path planning, AI, and computer vision. The only difference is that these are real problems that can help society really move forward. Everybody talks about how robots will be the future and how they will be doing laundry, driving cars, cleaning the house, and making dinner. However, the fact is that this hasn't happened yet. In fact, the problem is so hard that biped walking on rough terrain, a task most humans perform without thinking, is still a very hot area of research (and very interesting). Robots manipulating objects with in an unknown, uncertain, and unconstrained environment still haven't left the top research labs in the world. That's why the robotics problem is worthy enough to spent a person's entire life developing the technology and searching for the answers.

My advisors at CMU are James Kuffner and Takeo Kanade.