University of Nottingham, Ningbo, China
09.2018-07.2022 Bachelor of Engineering in Electrical and Electronic Engineering |
Overall GPA: 3.89/4.0 | Dean's scholarship (2021.12)
Johns Hopkins University
09.2022-06.2024 Master of Science in Engineering in Robotics | GPA: 3.80/4.0 |
Course Assistant of Algorithms for Sensor-based Robotics Courses:
Machine Learning | Algorithms for Sensor-based Robotics | Augmented
Reality | Robot Motion Planning | Learning-based Control for Robotics |
Robot System Programming
ONGOING ONLINE COURSES
Acwing: Fundamentals of Algorithms
DR_CAN: Modern Control Theory
PAST PROJECTS
Computer Vision & Machine Learning
Alzheimer's disease diagnosis based on MRI scans and clinical
data
Pre-processed the MRI scan data to fit the network input
Designed networks based on ResNet-18 by adding attention block
and clinical data input
Trained, tested and compared the model performance between
different architectures and hyperparameters and reached 83.8%
accuracy and 97.7% recall on test set.
Augment reality application development
Developed a zombie-chase-player AR game from scratch using Unity
3D and Vuforia and deployed it on an iPhone
Developing an AR application to map human anatomy onto the
person in image in real time for anatomy education
Deep Learning for Medical Image Segmentation
Redesigned the U-Net to output desired image size.
Tested the impact of padding and dropout in convolutional layers
on prediction accuracy.
Experimented with the elastic deformation method for data
augmentation, the overlap-tile strategy for seamless
segmentation, and the Adam optimizer for improved convergence
rate.
3D visualization of polyhedron by pygame
Implemented rotation matrix to achieve the rotation of
polyhedron by monitoring the motion of the cursor.
Implemented rasterization and depth buffer to correctly render
the polyhedron with color.
Calculated the angle between the surface and the parallel light
to dynamically change the color of the surface to simulate the
shading effect.
Robotics
Implementation of extended Kalman filter and particle filter for
mobile robot pose estimation
Implemented extended Kalman filter in C++ to estimate the
location of a mobile robot based on GPS and IMU on a simulated
rugged terrain to obtain location errors smaller than 0.5m
Implemented particle filter in C++ to estimate the position and
orientation of a mobile robot in a given map based on Lidar
Motion Planning for a 6-joint serial link manipulator
(UR5)
Implemented hand-eye calibration algorithm in MATLAB for
manipulator-camera systems
Implemented Probability RoadMap Planning in C++ to enable the
collision free operation of the manipulator
Designed the mechanical structures for package storage and
sensors integration.
Implemented a deep learning model in PyTorch to drive real-time
RGB-D camera frame segmentation to detect the collision-free
space for obstacle avoidance purpose on NVIDIA Jetson; further
enhance system robustness by integrating radar and ultrasonic
sensor data.
Developed a signal conditioning system on ROS to extract,
process, and fuse raw data from an array of sensors, including
GPS module, radars, IMU, wheel encoders, RGB-D cameras, and
ultrasonic sensors.
Fused the GPS and IMU data using Kalman filter to obtain a
continuous position estimation.
Developed a local path planning algorithm for obstacle
avoidance, with a focus on multi-sensor data fusion and utilize
A* algorithm for global path finding.
Developing the delivery service functionality and human-robot
interaction for delivery service using.
Line following robotic vehicle
Designed the PCB of H-Bridge motor control circuit in KiCad to
integrate an array of electronic components, including MOSFET
driver, MOSFETs among other peripheral equipment, followed by
soldering, wire connection, and hardware test.
Enabled recognition of traffic light and road signs based on an
array of classical computer vision algorithms implemented in C++
using Raspberry Pi and OpenCV API, with a focus on navigating
through a preset route while performing various dynamics
maneuvers following model road signs.
Enabled remote monitoring and control of the robotic vehicle via
an Arduino-based controller using a pair of wireless modules
nrf24l04.
Built and parametrically optimized a PID controller to improve
the line-tracking performance.
Utilized DWM1000 UWB compliant wireless transceiver module to
enable indoor localization with a precision of 10 cm, range of
100m, and data transmission rate of 6.8 Mb/s and utilize A*
algorithm for global path planning.
Converted the PyTorch model for collision-free space
segmentation to TensorRT inference engine to get about 3 times
faster real-time performance on Jetson board.
Developed a suite of control algorithms in Python to enable
real-time steering, speed control and path-finding based on
sensor data; perform microcontroller unit programming in C;
develop the system based on ROS mainly in Python.
uild a GUI in Swift general-purpose programming language for
monitoring and controlling the robot through Bluetooth on an iOS
device.
