Luzhe Sun(孙鲁喆)
Ph.D. Student AT Toyota Technological Institute at Chicago
Master of Computer Science AT University of Chicago
Bachelor of Engineering AT Xiamen University(Graduates With Honor)
Email: luzhesun@ttic.edu / luzhesun@uchicago.edu
Interest: Robotics Perception Algorithm, diffusion model,Graph Theory
Currently I am a PhD Student at TTIC
Robot Intelligence
through Perception Lab
supervised by Professor Matthew Walter
Biography
I'm Luzhe Sun,a first-year Ph.D. student at TTIC. My research focuses on Robotics Planning and Graph Theory. Prior to joining TTIC, I completed my Masters Program in Computer Science at the University of Chicago.
I studied at the School of Information Science at Xiamen University as an undergraduate, where I conducted research on graph theory with Professor Zhihong Zhang. I worked with Xing Ai to complete a paper on quantum neural networks for graph classification under the supervision of Edwin Hancock. This has greatly increased my interest in graph theory and machine learning.
In the summer of 2020, I spent seven months with Professor Guang Lin at Purdue University learning about multi-fidelity data prediction and building the Deep Multi-Fidelity Gaussian Process Prediction framework.
In my free time I really enjoy photography and painting Chinese paintings, so if you're interested I can teach you how to get started. I also play the erhu, but not as distinguished as painting.
Publications
-
Xing Ai, Luzhe Sun, Zhihong Zhang, Junchi Yan, Edwin Hancock Decompositional Quantum Graph Neural Network
-
Takuma Yoneda, Luzhe Sun, Ge Yang, Bradly Stadie, Matthew R. Walter To the Noise and Back: Diffusion for Shared Autonomy [RSS 2023] [Website Link]
Awards and Honors
-
Xiamen University Outstanding Student Worker Scholarship 2018.
Xiamen University. China. -
Xiamen University "Xilie Huang" Annual Scholarship(2/200) 2018.
Xiamen University. China. -
National Third Prize of China Students Service Outsourcing competition. 2018.
Ministry of Education of the People's Republic of China, Ministry of Commerce of the People's Republic of China, China. -
National Second Prize of Contemporary Undergraduate Mathematical Contest in Modeling. 2019.
China Society for Industrial and Applied Mathematics, China. -
China National Scholarship(Highest scholarship given by Chinese government Top 0.1%) 2019.
Chinese Ministry of Education. China. -
China National Scholarship(Highest scholarship given by Chinese government Top 0.1%) 2020.
Chinese Ministry of Education. China. -
Honored Graduates Student, 2021.
Xiamen University. China
Research
To the Noise and Back: Diffusion for Shared Autonomy [RSS 2023]
Robot Intelligence through Perception Lab, TTIC
Using diffusion model to correct user input, thus achieving shared-autonomy. This model aims to help
user finish hard operations swiftly and stably.
User intent is preserved by controlling the number of diffusion steps in the diffusion model.
The experiment is successful on hard control problems like the LunarLander and Block Push environment.
Trajectory Planning via Diffusion Model 2022
Robot Intelligence through Perception Lab, TTIC
Using the diffusion model to predict the trajectory of the manipulator. At the same time explored
the properties of the diffusion model on trajectory stitching and conditional sampling to achieve trajectory inpainting.
The experiment was simulated in Maze2D and Ravens environment.
Deep Multi-Fidelity Gaussian Process (Purdue University)
In daily life, we often encounter situations where it is difficult to
obtain high-fidelity data during the training phase of some data
predictions, such as changes in the earth’s atmosphere and heat
conduction data. But for these experiments, we can use artificially
simulated physical models to obtain a large amount of data. But these
artificially simulated data often have a gap with the real situation,
that is, low-fidelity data. How to train a model with a large number
of low-fidelity data obtained by simulation combined with very few high-precision
high-fidelity data so that the model has a high correct rate of prediction for
high-fidelity data has become a problem worthy of research.
(website link)
Quantum Graph Classification Neural Network (Xiamen University, Data Mining Lab)
Quantum machine learning is a fast emerging field that aims to tackle
machine learning using quantum algorithms and quantum computing.
We propose a novel hybrid quantum-classical algorithm for graph-structured data, which we refer to as the
Decompositional Quantum Graph Neural Network (DQGNN).
DQGNN implements the GNN theoretical framework using the tensor product and unity
matrices representation, which greatly reduces the number of model parameters required.
(website link)
Projects
Embodies ChatGPT | Invited demonstration at MSI(Chicago) Robot Block Party 2023
we implemented the Code as Policies algorithm on the UR5 robot. The robot is able to respond to the user’s
voice command and execute the corresponding task. The project was presented during the 2023 National
Robotics Week at the Museum of Science and Industry in Chicago, and at the University of Chicago for
the 2023 South Side Science Festival.
With the economic development and the improvement of living standards, people's demand for spiritual culture has further increased. Tourism has become people's basic way of life and one of the best choices for people to use leisure time. In recent years, the booming transnational free travel has gradually entered the scope of public acceptance. The accompanying problem is that there are inconveniences in tourism, accommodation consumption and other fields due to the relationship of language. The "Self-Travel Assistant" APP is designed based on this situation and is an auxiliary application for tourists from China and Japan to travel independently. Using modern AI technology to assist tourists, provide and recommend information on local attractions, restaurants, accommodation, transportation, etc. (website link)