We often encounter some limitations in sensing capacity while collecting data for estimation purposes. The limitation may be, for instance, in the size of the sensory data to be measured, or in the number of samples we can collect for observations. In this talk, we draw specific examples from mechanical systems to introduce some of pre-processing techniques that can be used to cope with such limitations. We first look at how the compressive sampling (CS) can be applied to the estimation of mechanical strains using portable ultrasound devices in circumstances where the raw RF data need to be downsized for computational and transmissive purposes. The second case of our interest for limited sensing capacity relates to situations where the sequence of data collection needs to be carefully determined due to high cost of experiment. Such a constraint often arises from engineering applications of statistical inference such as the Gaussian process (GP) regression with sequential sampling. We look at how the optimal sensor placement (also known as the experiment design) techniques can be used for the efficient operation of the GP regression, and illustrate their utility with two practical examples; the wind field estimation for precision airdrop with a UAV (Unmanned Aerial Vehicle) and the 3D shape estimation via tactile exploration.

※ 학력
- 1998년 서울대학교 기계설계학과(학사)
- 2001년 서울대학교 기계항공공학부(석사)
- 2007년 UC Berkeley, Mechanical Engineering (PhD)

※ 주요경력
- 2015년-현재 Associate Professor (부교수), Department of Mechanical and Mechatronics Engineering, University of Waterloo, Canada
- 2009년~2015년 Assistant Professor (조교수), Department of Mechanical and Mechatronics Engineering, University of Waterloo, Canada
- 2007년~2009년 Mechanical and Systems Engineer, Applied Materials Inc., Santa Clara, CA USA