cv

Advanced Robotics Internship

Siemens

Developed a task and motion planning framework using ROS in Python for automatic and precise handling of COVID-19 testing strips using a UR5 robot with RGB-D images in flexible workspace.

• robotic manipulation
• task and motion planning
• recognized by the ARM Institute
• use of Python, ROS, MoveIt

Researcher

PRACSYS Group at Rutgers University

Served as a researcher at PRACSYS Group focusing on vision-based robotic manipulation for objects in confined environments in simulation and on a real robot.

• robotic manipulation
• task and motion planning
• use of Python, C++, ROS, MoveIt, PyBullet, Mujoco

Researcher

ARCLab at University of California, San Diego

Served as a researcher and developed a deep-learning based kinodynamic motion planning algorithm and an active continual learning variation of the MPNet model. Accelerated the performance of motion planning algorithms. Improved data efficiency through active continual learning.

• learning-based motion planning
• acceleration and improving data efficiency
• use of Python, C++, PyTorch, OMPL

Junior Research Assistant

the Chinese University of Hong Kong

Worked on a theoretical roblem in graph theory on fixed-parameterized tractability of finding length-constrained paths. Surveyed literature and constructed a dynamic programing algorithm for a simpler setting.

• graph theory
• fixed-parameterized tractability
• theoretical computer science

M.Sc. in Computer Science

Rutgers University

• Pattern Recognition
• Computer Graphics
• Numerical Analysis
• Network and Combinatorial Optimization
• Robotics: SC Robotics, SC Robotics Design, Robotics and Society

M.Sc. in Computer Science

University of California, San Diego

• Probabilistic Reasoning & Learning (A+)
• Neural Network/Pattern Recognition (A+)
• ML on Geometrical Data (A+)
• Computer Vision I (A)
• Planning & Learning Robotics (A)
• Convex Optimization Algorithms (A)
• AdvTech/Computational Math II (A+)

B.Sc. in Computer Science with Honours, First Class

the Chinese University of Hong Kong

• Topics in Graph Algorithms (A)
• Foundations of Optimization (A)
• Matrix Analysis and Computations (A)
• Web-scale Information Analytics (A)

NSF NRT SOCRATES program

Socially Cognizant Robotics for a Technology Enhanced Society

An NRT traineeship that focuses on integrating robotics, machine learning, and artificial intelligence with social and behavioral sciences, including psychology, cognitive science, urban planning, and policy development.

Outstanding Manipulation Paper Award-Finalist at ICRA 2022

IEEE Robotics and Automation Society

For the paper co-authored with Shiyang Lu, Rui Wang, Chaitanya Mitash and Kostas E. Bekris, "Online Object Model Reconstruction and Reuse for Lifelong Improvement of Robot Manipulation"

ELITE Stream & Intelligent Science Stream

Faculty of Engineering, CUHK

The ELITE Stream is offered by the Faculty to students with excellent academic performance. Its aims are to nurture outstanding engineering students and to develop their potentials through additional challenging coursework and invaluable extra-curricular activities. Any student who meets the entrance requirements is eligible for the Stream. The award of the ELITE Stream to qualified students will be officially recorded on the academic transcript.

Dean’s List (2013-15, 2016-18)

Talent Development Scholarship

To give recognition to students with achievements or talents in non-academic areas and provide support for these students to further develop their talent and potential. Have demonstrated talent or potential in at least one of the following areas: (1) Sports and games; (2) Music and performing arts; (3) Culture, arts and design; or (4) Innovation, science and technology; Good academic attainment.

Department Admission Scholarship 2013-14

Department of Computer Science and Engineer, CUHK

Resolution Complete In-Place Object Retrieval given Known Object Models

International Conference on Robotics and Automation

Safe, occlusion-aware manipulation for online object reconstruction in confined spaces

The International Symposium of Robotics Research

Efficient and high-quality prehensile rearrangement in cluttered and confined spaces

International Conference on Robotics and Automation

Online object model reconstruction and reuse for lifelong improvement of robot manipulation

International Conference on Robotics and Automation

MPC-MPNet: Model-Predictive Motion Planning Networks for Fast, Near-Optimal Planning under Kinodynamic Constraints

IEEE Robotics and Automation Letters

Motion planning networks: Bridging the gap between learning-based and classical motion planners

IEEE Transactions on Robotics

Towards End-To-End Learning-Based Algorithms in Motion Planning (Master's Graduation Thesis)

Trash Collection Robot

Course project at Rutgers University. Programmed a LoCoBot WX250 robot for trash collection. Set up navigation stack and ROS2 communication. Controlled the robot movement through a PID-like controller. Engineered motion planning for grasping objects given perception from YOLOv5s.

• ROS navigation stack
• ROS2
• control
• motion planning

Conditional Variational Autoencoder for Sampling-based Motion Planning

Course project at Rutgers University. Applied Conditional Variational Autoencoder for sampling-based motion planning, enabling learning of conditional distribution by user inputs.

• Variational Autoencoder
• sampling-based motion planning
• PyTorch

Deep Learning for 3D Collision Checking

Course project at UCSD. Generated dataset from online 3D scanned point cloud data and applied Siamese Network with PointNet for learning-based collision checking for 3D point cloud. Achieved more than 90% accuracy on complex point cloud objects.

• PointNet
• Siamese Network
• PyTorch

Affordable Smart Wheelchair

Student club at University of California, San Diego. Led the setup through ROS for the first semester. Configured ROS Navigation Stack for the second semester and experimented with ORB SLAM2.

• ROS
• ROS Navigation Stack
• ORB SLAM2