Robot perception and semantic understanding, which is essential in robotics, have been greatly improved based on the large amount of training data and advent of the artificial intelligence. However, artificial intelligence techniques which heavily rely on statistics and vast amount of training data becomes kind of a double-edge sword; in the environment that are slightly different from thee training data, the perception performance is significantly reduced. This limitation makes AI hard to be adopted to the autonomous driving or the adaptive mobile robot in various environment, since the various environments have their own characteristics based on the countries, driving road composition, weather condition and even the diverse observational viewpoints...
전체댓글 0