Zheyuan Xie

I am a second year master student majored in Robotics at GRASP Lab, University of Pennsylvania. Previously, I have completed my bachelors degree in automation in Shanghai Jiao Tong University. I am devoted to creating intelligent robotic systems, especially in designing perception and control software. Aspired to become a “full-stack” engineer, I am actively developing my skill set across all disciplines.

The section is under construction!

Physically-based grass rendering		Fall 2019
	This project is an implementation of the paper, Responsive Real-Time Grass Rendering for General 3D Scenes. A grass simulator and renderer is implemented with Vulkan.

CUDA Path Tracer		Fall 2019
	This is a CUDA-based path tracer capable of rendering globally-illuminated images very quickly.

PennOS		Spring 2019
	A user-level operating system implemented in C based on UNIX preemptive signaling and user context (ucontext).

Robotics Orchestra		Spring 2019
	IPD 516 Advanced Mechatronics Reactive Space final project.

Probablistic Robotics		Spring 2019
	ESE 650 Learning in Robotics course project. - Estimate orientation with a 6-DOF IMU based on UKF. - Implement a full SLAM system based on EKF. Applied to Victoria Park dataset.

Automatic Face Swapping Between Videos		Fall 2018
	This interesting project is able to creating fake videos by "seamlessly" swapping faces. A face in the source video that has similar pose and emotion with the face in the target video frame is selected by comparing facial landmarks. This source face is warpped using Barycentric transform, color-corrected, and finally Poisson blended into the target video.

Optical Flow Object Tracking		Fall 2018
	Kanade–Lucas–Tomasi (KLT) feature tracker implemented in Python.

Image Stitching and Panorama by Feature Matching		Fall 2018
	Create beautiful panorama photos!

Kinematics and Planning of Serial Manipulator		Fall 2018
	MEAM 520 Introduction to Robotics course projects. Formulated forward, inverse, and velocity kinematics of a 5 DOF Lynx manipulator arm. Implemented collision-free path planning with Probablistic Roadmap Method (PRM) and artificial poential field.

Tweet-Based Health Prediction		Fall 2018
	This is the final project of CIS 520 Machine Learning. Our team ranked 6/66 on the class leaderboard. We tried to predict health statistics from 2000 Tweet LDA topics collected from different counties across US. A number of classification algorithms are implemented including k-NN, MLP, random forest, Gaussian process regression.

Autonomous Robot for the DJI Robomaster AI Challenge		Summer 2018
	DJI RoboMaster Summer Camp project.

Control and Navigation of a Quadrotor with Reinforcement Learning		2018
	- End-to-end control policy represented by an MLP. - Control policy learned through trails in simulator using PPO, implemented in Python and TensorFlow. - Learn optimal path to goal region in a grid world with Q-Learning, verified with experiment.

Autonomous Navigation of a Quadrotor in Unknown Environment		2017
	- Realtime trajectory generation from RGB-D sensor with OpenCV. - Implemented the model predictive controller (MPC) in Python. - Performed simulation in ROS/Gazebo.

Human-Robot-Interface Platform		Summer 2017
	Quori is a low-cost HRI platform developed in Modlad at UPenn.

Cooperative Transport with Two Quadrotors		2017
	Basic PID control and parameter tuning. Cooperative transportation modeling ROS programming. Qt Interface programming. OptiTrack motion tracking.

Automated Jigsaw Puzzle Solving with Cartesian Manipulator		Fall 2016
	- Automated jigsaw puzzle solving on tounchscreen tablet. Maximum complexity supported:7x7! - Digital image processing (region segmentation, SIFT feature matching, template matching) with C++ and OpenCV. - Graphical user interface written in C++ and Qt. - Send GCode motion commands to serial port to control a cyberDIP device.

The section is under construction!