About Me
I am a Tenured Professor in Department of Computer Science and Technology, Tsinghua University, Beijing, China. I'm also in the State Key Lab. of Intelligent Systems and Technology, and Beijing National Research Center for Information Science and Technology.
My research work involves several fields of Embodied Intelligence, including Robotics, Dynamic Systems and Machine Learning, with particular emphasis on robotic perception, learning and control which address the three widely-known challenging issues:
Control: With more robotic systems equipped with multi-joint, flexible-link and soft materials, it becomes increasingly demanding to achieve complex dynamic control.
Perception: With the un-structured environment, limited sensing capability and constrained physical interaction, it is intrinsically difficult for a single robot or a group of robots to perceive their environments precisely.
Learning: With the dynamic and consecutive multiple heterogeneous tasks, it is extremely challenging for the robots to perform lifelong learning.
Research Summary
We have been working on tackling the above problems and our long term goal is of establishing a cognitive pathway between the manipulation control and robotic perception using innovative learning methodologies. We believe multi-modality of perception, high degree-of-freedom of action, and persistent accumulation of learning are essential to simulate intelligent behavior. Our research work during the past five years can be summarized as:
We established the incremental robust dictionary learning framework for the robotic lifelong learning tasks and addressed the exemplar extraction problems for the non-Euclidean space. This work lays a theoretical foundation for the robotic life-long learning in the dynamic environment.
We established the multi-modal fusion method to tackle the weakly-paired heterogeneous fusion problem. This work provides effective theoretical approaches for the robotic perception of complicated environment.
We established the reinforcement learning method to realize active perception for mapping between the multi-modal perception information and the high-dimensional action space. This work provides solutions to the robotic adversarial environment perception.
We developed several platforms including unstructured video summarization, robotic material analysis and intelligent vehicle to validate the developed methods.
Teaching
I am currently responsible for teaching three courses in Tsinghua University: “Dynamic Systems Analysis and Control” for undergraduate students, “Robotics” and “Uncertain Artificial Intelligence” for graduate students, all of which are closely related to my research fields. The various research topics in the different fields motivate me to investigate the intrinsic relationship among them from the perspective of a teacher. Therefore, I develop several illustrative projects which bridge the pedagogical gap between Systems Control and Machine Learning. A representative example is the dynamic video texture for dynamic video texture for dynamic system stability analysis, which provides the opportunity for students with no access to mechanical or electronic equipment to learn system stability. A detailed description can be found in [TE_2014].
I was afforded the Distinguished Teaching Award for Young Scholars in Tsinghua University at 2016.
The courses Dynamic Systems Analysis and Control has been ranked as Excellent Course in Tsinghua University from 2015.
Publication [TE_2014]
Huaping Liu, Wei Xiao, Hongyan Zhao, Fuchun Sun, Learning and understanding system stability using illustrative dynamic texture examples, IEEE Transactions on Education, vol.57, no.1, 2014, pp.4-11.