I am a postdoctoral researcher in Robotic Manipulation and Mobility Lab (Prof. Matei Ciocarlie) at Columbia University. My current research focuses on optimal control and reinforcement learning (RL). I earned my PhD in control theory under the guidance of Prof. Richard Longman (Columbia University) and Prof. Minh Phan (Dartmouth). My thesis is “From Model-Based to Data-Driven Discrete Iterative Learning Control”. I presented my work in the Spring 2020 GRASP Seminar Series at the University of Pennsylvania.
I am passionate about the integration of RL and control following the human central nervous system in spirit. I believe it is the direction leading us to intelligent robots. I am developing a hierarchical learning method to improve the sample efficiency that can allow learning in the real world. To become an expert in learning algorithms, I am interested in any opportunities that can
- enrich my knowledge in numerical methods, particularly for optimization;
- sharpen my coding skills, i.e., the efficienct implementation of learning algorithms;
- broaden my experiences in real-world applications.
I also enjoy discussing topics related to “learning” beyond technology. “Learning” pushes forward human civilization. How shall we teach ourselves? How shall we teach each other? I am seeking these answers.
“All life is problem-solving”. I follow the cycle of learning to approach my answers:
- define the type of problems to be solved;
- propose tentative theories;
- revise and eliminate theories through critical thinking, discussion and trials;
- identify new problems that arise in this process.
Please contact me at bing.song@columbia.edu!
My CV can be found here.