報告時間：2023年12月8日（星期五）16:00
報告地點：學術會議中心二樓報告廳
報 告 人：盧添福 Tien-Fu LU
工作單位：School of Electrical and Mechanical Engineering, University of Adelaide, Australia
舉辦單位：儀器科學與光電工程學院
報告簡介：
Reinforcement Learning (RL), one of the model free techniques, has been widely employed in recent years to solve complex engineering and non-engineering problems including self-driving cars, industry automation in production lines, financial trading, and healthcare with proven satisfactory solutions. Nevertheless, when problems/systems are more complex, the action and state-space can increase exponentially, and most RL techniques would fail to efficiently compute optimal policies to these problems leading to excessive computational overheads. It remains a challenge for RL techniques to balance between exploitation and exploration in larger spaces to efficiently (in terms of the speed of convergence/sample efficiency) optimise the RL model policy for best cumulated rewards. Quantum computation offers certain computational advantages with potential improvements to traditional RL models. Quantum Variational Circuit (QVC) provides computational advantage by using parameterized circuit running in quantum environment and output the parameter to classical optimizer. This talk covers the recent work of the presenter’s team which compares the performances of one qubit quantum enhanced RL models through structure variations including gates and trainable parameters using OpenAI Gym CarPole motion control platform. Through varying the structure, better CarPole motion control performances are obtained and discussed.
報告人簡介：
Tien-Fu Lu is currently the associate head of school – international and external engagement at the School of Electrical and Mechanical Engineering, University of Adelaide.
His research interests are mainly in the fields of Mechatronics and Robotics covering piezoelectric actuators/energy harvester, nano-positioning and measurement technologies, and compliant mechanisms. He has been a Board member of ASPEN (Asia Society tor Precision Engineering and Nanotechnology since 2013, http://www.aspen-soc.org/#conferences and contributing as members of various conference committees and chair/co-chair of sessions and editorial members of journals (editorial board member and co-editors).
As a chief investigator, he has been awarded grants in the past 5 years jointly with colleagues for more than 22 million dollars in total. He has published over 160 articles including book chapters, journal, and conference articles in the field of robotics and mechatronics. More details can be found at:https://researchers.adelaide.edu.au/profile/tien-fu.lu.
