Research Highlights

Home > Research > Research Highlights


Research Highlights

Research Highlights

Prize awarded in the design contest of the International Semiconductor Symposium

2008.02.10 우정호,김동현 1

KAIST researchers have been awarded in the international semiconductor symposium for developing processors for 3D graphics in mobile devices and high performance real time visual process in robots.

KAIST electrical engineering and computer science professor, Yoo Hoi-Jun (47), on the 10th, announced that Ph.D candidate Woo Chung-Ho received the outstanding design prize in the 2007 Asian Solid-State Circuits Conference (ASSCC) for low power 3D graphics accelerator, which is programmable for high performance 3D graphics in mobile devices.

In addition, from the same laboratory, Kim Dong-Hyun received the outstanding design prize in the 2008 DAC/ISSCC student design contest for developing a real time visual processor in robots to rapidly identify moving objects.

Woo and Kim, for 4 days from the 4th have been invited to the International Solid State Circuits Conference (ISSCC 2008), held at the Marriot Hotel in San-Francisco, USA, to demonstrate the developed processors in front of semiconductor professionals.

What Woo and Kim has developed are low power programmable 3D graphics accelerators for mobile devices, and are expected to be used in cellular phones and PMP to realize high performance 3D graphic games, etc.

Realizing high quality 3D graphics in mobile devices like in PCs require programmable 3D graphic accelerators, but have been difficult due to high power consumption and limited silicon area.

However, the researchers succeeded in developing a new architecture, capable of realizing 3D graphics in mobile devices as in PCs, which reduces the silicon area and power consumption by 35% and 28%, respectively, while improving the graphics performance by a multiple of 2.1.

Same laboratory Kim Dong-Hyun, developed a high performance object identifying chip for real time intelligent robots, with twice the processing speed with highly reduced power consumption.

Object identifying in intelligent robots is essential for standalone navigation, distinguishing enemy robots, or automotive safety devices, but has had problems using high performance processors due to high power consumption.

However, Kim has integrated the rapid visual data process that comes in the camera, with the function to compare the object in the saved data, in one chip and reduced the power consumption by one tenth, while improving twice the processing speed.

Prof. Yoo said, “Not only did we finish in developing the chip, but have proven its performance in intelligent robots,” and “This system processes visual data rapidly and quickly compares it with the saved data in identifying the object.”