CV

Education

M.S. in Robitics, University of Michigan, 2023 (expected)
- GPA: 4.0/4.3
- Exchange student at the University of Toronto, GPA: 3.96/4.0
- Coursework: Mobile Robotics and Perception, Robot Modeling and Control, Machine Learning
B.S. in Mechanical Engineering, National Taiwan University, 2020
- GPA: 4.0/4.0
- Coursework: Mobile Robotics, Computer Vision, Robotics System Laboratory

Skills

Programming: C/C++, Python, MATLAB, LabVIEW
Software Tools: ROS, Git, OpenCV, PyTorch, Simulink, Raspberry Pi, Arduino, Docker

Work experience

Incoming Fall intern @ Tesla autopilot team
Incoming Summer intern @ Nvidia mapping team
Software Engineer Intern @ Delta Research Center, Delta Electronics (Jul. 2020 - Aug. 2020)
- Designed a pipeline for robotic arms with a camera to track multiple faces in real time with OpenCV library
- Implemented Haar-cascade and HOG object detection algorithm and compared their runtime and accuracy
- Reduced tracking latency by 50% by implementing multithreading techniques and a modified PD controller

Research experience

Graduate student researcher @ Biped Robotics Laboratory, Univeristy of Michigan
- Worked on the localization pipeline for Cassie, the biped robot, using topometric map and lidar measurement
- Extracted and Analyzed road and sidewalk feature from a Velodyne Lidar using PCL library in a ROS system
- Replaced the bus stops generation process with agglomeratively hierachical clustering algorithm
Undergraduate student researcher @ Microrobotics Laboratory, University of Toronto
- Constructed an automatic pipeline to acquire 1500 measurements of magnetic fields within 30 mins for microrobots
- Implemented interior-point algorithms in MATLAB to reduce the average magnetic field model errors by 50 %

Projects

SLAM and exploration of a mobile robot, University of Michigan (Sept. 2021 - Oct. 2021)
- Programmed a mobile robot to perform SLAM and exploration tasks (First place in the contest out of 24 teams)
- Implemented Monte Carlo Localization with sensor readings from a RPLidar, wheel encoders, and a gyroscope
- Implemented Frontier Exploration with A-star algorithm for the robot (Fully explored the maze in the competition)
- Reduced position errors from 90 cm t0 3 cm when commanding the robot to drive a 1m square 8 times
Autonomous blocks stacking robotics arm, University of Michigan (Oct. 2021 - Dec. 2021)
- Programmed a robotics arm to autonomously identify and manipulate blocks (First place in the contest out of 24 teams)
- Conducted camera intrinsic and extrinsic calibration using OpenCV PnP solver (with 4mm maximum depth error)
- Developed an algorithm that utilizes a RGBD camera to detect blocks’ size, color, position, orientation on a board
- Implemented Forward / Inverse Kinematics and fine-tuned a PID controller (reduced offset from 20mm to 3mm)
Visual Odometry using KITTI dataset, University of Michigan (Oct. 2021 - Dec. 2021)
- Extracted feature points from stereo images using two approaches: SIFT and D2-Net (a CNN-based feature extractor)
- Estimated motion using essential matrix for monocular camera and using PnP solver for stereo camera
- Compared error and runtime for different feature extraction model and different camera model
Deep Learning for Computer Vision, University of Michigan (Oct. 2021 - Dec. 2021)
- Trained a MiniVGG model with added batch normalization layers to predict categorized scenes with 55% accuracy
- Implemented Pix2Pix conditional GAN model to color shoe’s images containing only the edges to full images of shoe
- Implemented a single-stage object detector, based on YOLO v1, which achieved 12% mAP on the PASCAL VOC dataset