Luo, Dingsheng
Associate Professor
Research Interests: Embodied intelligence, machine learning, intelligent information processing, robotics
Office Phone: 86-10-6275 2502
Email: dsluo@pku.edu.cn
Luo, Dingsheng is an associate professor in the Department of Machine Intelligence, School of EECS, Peking University, and has served as the Deputy Director of the department since 2015. He received his Bachelor degree from North-west Normal University in 1997, Master degree from Lanzhou University in 2000, and Ph.D. degree from Peking University in 2003 respectively. His research interests include embodied intelligence, machine learning, intelligent information processing, robotics, intelligent humanoids, and service/assistive robots.
Dr. LUO has published more than 60 research papers, including journal papers on PONE, IJARS, etc., and conference papers on Humanoids, IROS, IEEE ICDL-ER, ACL, ICASSP, IEEE ICMLA, ECML, IJCNN, ICONIP, etc. He also has more than 5 China national invention patents. He was awarded Best Paper Award of CNNC in 2002, Zhengda Award of Peking University in 2008, First Prize Award of IJCAI-13 Robot Competition Technical Skills Category in 2013 and Best Paper Finalist Award of IEEE ICIA in 2015 and 2016 respectively. Dr. Luo is an IEEE member, a member of IEEE-RAS Technical Committee on Humanoid Robotics, a member of the Climbing and Walking Robots Association, a senior member of Chinese Association for Artificial Intelligence, serves as the Deputy Secretary General of the Educational Committee of the Chinese Association for Artificial Intelligence. Dr. Luo also serves as editor and reviewer of several Domestic or International journals, and as session chair and PC member of several international conferences, e.g. AAAI, Humanoids.
Dr. Luo has/had undertaken more than 10 projects funded by Chinese Government, including NSFC, 973 programs, etc. His research achievements are summarized as follows:
1) Autonomously Acquiring Human-like Robot Skills: During past decades, helping the robot to acquire human-like capabilities was heavily focused, and unlike most previous methods, he focuses on autonomous acquisition of these skills, and proposed a hierarchical framework with the ideas of embodiment cognition and developmental learning under the trial-and-error paradigm.
2) Exploring Human Inspired Mechanisms: In AI and robotics community, inspiration from how human do offers a promising way in solving many tasks. He focuses on this point and proposed a HMCD (human motion capture data) based multi-stage approach for robot standing up, an ankle-hip-step combined strategy for robot push recovery, a human COP (Center of Pressure) trajectory based foot rolling for bipedal walking, a human-infant inspired approach for robot reaching, etc.
3) Environmental Perception and Activity Recognition: Due to that the real world is always extremely complicated and full of miscellaneous variations, he focuses on the topic of robot situation awareness, and proposed a multi-sensory fusion based environmental perception method, a new raw-sensor-data based DNN model for human activity recognition, and a randomly state-reset technique based RNN model for robot activity self-awareness.
Machine Learning Algorithms: His team also focus on learning algorithms, and proposed a discriminative apprenticeship learning (DAL) algorithm to dig out potentials from both preference and non-preference demonstration cases, a cyclic contrastive divergence (CCD) algorithm to solve the input-independent problem that traditional contrastive divergence (CD) algorithm encountered when High-order RBMs are involved, etc.
