- School of Information and Communication Engineering, Sungkyunkwan University
martedì 19 dicembre 2006
inizio 17:30, caffe & C ore 17:00
An overview of the research being carried out at the Intelligent Systems Research Center of Sungkyunkwan University, Korea, is presented. The talk will focus especially on the recent result on robust, human-like, vision based manipulation and navigation that are being developed for application to home service robots. Specific topics to cover include real-time 3D object recognition and pose estimation, real-time 3D workspace modeling, motion planning with accessibility and retrieval plans, and vision based SLAM under the framework of a service robot information system. The robustness in recognition and pose estimation is achieved by behavioral cognition where an optimal set of multiple evidences is selected and collected in-situ and is matched against the given model based on a sequence of images obtained from robot motion. This multiple evidence and model matching process continues till the certainty of recognition and pose estimation, accumulated by particle filtering, reaches sufficiently high enough for credible decision. Experimental results are shown.