Qijin She
Education
National University of Defense Technology 2019.9 - 2022.6
Changsha, Hunan, China
- Master of Science (M.Sc.) in Computer Science and Technology
- GPA: 3.73/4.0 (8/10 courses are the highest grades)
- Supervised by Prof. Kai Xu
- Master Thesis: Real-time Reaching-and-Grasping with Dynamic Interaction Geometry Representation
Shandong University 2015.9 - 2019.6
Jinan, Shandong, China
- Bachelor of Engineering (B.Eng.) in Software Engineering
- GPA: 89.4/100; Ranking 8/292
- Data Science track
Research Experience
Project 1: Reaching-and-grasping Planning for Dexterous Hand October 2020 - January 2022
Advised by Prof. Ruizhen Hu, Prof. Hui Huang, Prof. Kai Xu Visual Computing Research Center, Shenzhen University
Project Leader
- introduced a novel geometric state representation to characterize hand-object interactions for grasping.
- designed and implemented a novel RL algorithm utilizing vectorized reward functions and imperfect demonstrations.
- built a physics-based grasp simulation system based on Pybullet.
Project 2: Online 3D Bin Packing Problem August 2019 - October 2020
Advised by Prof. Chenyang Zhu, Prof. Ying Yang, Prof. Kai Xu iGRAPE Lab, National University of Defense Technology
Project Co-Leader
- designed and implemented a RL-based method for the online 3D packing task.
- extended the packing method to the looking-ahead setting with the modified Monte Carlo Tree Search method.
- developed a simple user-study app to collect user test data.
Publications
Learning High-DOF Reaching-and-Grasping via Dynamic Representation of Gripper-Object Interaction
Qijin She*, Ruizhen Hu*, Juzhan Xu, Min Liu, Kai Xu, Hui Huang. ACM Transactions on Graphics (SIGGRAPH’22)
Proposed a novel geometric representation that dynamically characterizes gripper-object spatial relation, as well as a learning method
to train a policy model using this representation to solve the High-DOF reaching-and-grasping task.
Online 3D Bin Packing with Constrained Deep Reinforcement Learning
Hang Zhao*, Qijin She*, Chenyang Zhu, Ying Yang, Kai Xu. AAAI 2021
Proposed a new constrained reinforcement learning method with modified Monte Carlo Tree Search (MCTS) to solve the online 3D bin packing problem where only the first k items are known and in a fixed order.
Honors & Scholarships
NUDT Excellent Master’s Thesis (about top 5% of master’s graduates) 2023
The Best Poster Award at CCF CAD&CG 2022 (top 10 popular poster) 2022
Outstanding Postgraduate Scholarship (for excellent academic performance) 2020
Linglong Scholarship (top 2.5% in the School of Software) 2016,2018
First Prize Scholarship (top 10% in the School of Software) 2016,2018
Teaching
Teaching Assistant, Data Structure and Algorithm NUDT, Fall 2020
Teaching Assistant, Seminar on Research Methods and Academic Writing in Computational Science NUDT, Spring 2020
Services
Journal Reviewer: CVMJ 2022
Skills
Programming: Python, Latex, C++, C, ROS1, Shell, Java
Tools: Blender, Pybullet, Meshlab, Scrapy, Photoshop
Courses
Math Courses: Calculus, Discrete Mathematics, Linear Algebra, Numerical Computing, Probability and Statistics, Statistical Decision Theory and Bayesian Analysis, Stochastic Process
Professional Courses: Advanced Computer Architecture, Artificial Intelligence, Augmented Reality and Human-Computer Interaction, Computer Graphics, Computational Geometry, Computer Network, Data Structure and Algorithms, Database System, Data Mining, Game Theory (MOOC), Machine Learning, Operating System, Pattern Recognition, Parallel Computing, Reinforcement Learning (MOOC), Software Engineering, Software Testing, Web Data Management