学术报告

题目：Imparting Human Intelligence to Machines

报告人：徐扬生教授

时间：2004年11月26日9:00

地点：自动化系407会议室

徐扬生教授简介

徐扬生教授，国际著名机器人学领域专家。1989年在美国宾夕法尼亚大学计算机系获博士学位。现任香港中文大学自动化与计算机辅助工程学系主任。2004年受聘于清华大学自动化系客座教授。1989 年至1999年，任美国卡内基梅隆大学（CMU）计算机科学系教授。

徐教授在CMU创建和主持全球第一个零重力环境空间机器人实验室。1993年成功研制月球探测的两用机器人。1996年首次提出陀螺稳定的独轮机器人。1997年，徐教授研究的人类控制规律自动建模在世界上第一辆横跨美洲大陆的自动汽车上获得成功。1998年被美国IBC评为全美最有影响力的500 名科学家之一。徐教授已经主持开发设计了20余个机器人或智能系统，几乎所有系统均为全球首创。

徐教授学术兼任包括：IEEE Fellow，香港工程师院院士，欧亚科学院院士，中国高科技发展计划航天领域咨询委员会高级顾问等。2003年因对“机器人及人机界面”的贡献，获21世纪成就奖。

报告内容简介

For over a few decades, people have come from a long way to develop intelligent machines with various schemes of machine intelligence. To build an intelligent machine, the central problems is to abstract intelligence and implant the intelligence to machines so as to allow them to make a right decision (the cognition problem), to sense the world (the perception problem), and to response with actions (the manipulation problem). The question is where the intelligence comes from.

In this talk, Professor XU will share the outstanding research fruits fulfilled in both Carnegie Mellon University and The Chinese University of Hong Kong. He and his students developed methodologies for modeling and transferring a class of human intelligence, i.e., human control strategies in response to real-time inputs. He will discuss the architectures, practice, pitfalls, and perspective in relation to the following work: (1) how they can efficiently model human control strategy in continuous and discontinuous real-time inputs, (2) whether they need to validate the learned models, (3) how about evaluation of the quality and performance of the human skills associated in the model, (4) what human skills can be transferred to another person, or to a robot, (5) why they need to model the human skill in handling dynamically stable systems, and (6) what sensing inputs should be selected for meaningful models.