Resume

Download

Education

Publications

Experience

Skills

Courses

References

Education

M.S. Robotics, Systems and Control

ETH Zurich, 2020-2023

Publications

Experience

Master Thesis Student -Robotics and Perception Group, University of Zurich

Dec. 2022 - Present

  • Achieved comparable racing performance with onboard vision-only inputs (gate corners).
  • Improved data collecting efficiency and achieved generalizable performances for the imitation learning framework.
  • Improved visual representation expressiveness using cross-attention mechanism and fused multi-modal visual embeddings used for vision-based drone racing tasks.

Software Developement Intern -ABB Corporate Research Center in Switzerland

May. 2022 - Nov. 2022

  • Developed Physics-Informed Neural Networks(PINN) to fit the execution time distributions and predict the Worst-Case Execution Time(WCET) of applications running on ABB commercial motor drivers.

Semester Project Student -Robotic Systems Lab (RSL), ETHz

Oct. 2021 - Feb. 2022

  • Developed a base pose selector based on Reinforcement Learning (PPO) and trained policies for a wheeled-legged robot to ground on challenging terrains.
  • Formulated evaluation metrics, experimented with reward shaping and improved the performances.
  • Achieved a success rate of 80% at a grounding error of 0.05m on challenging terrains, e.g. Holes.
  • Integrated the pose selector into the global planner (RRT*) and benchmark against brute force search.

NeurlIPS21 Competition -AI Center, ETHz

Aug. 2021 - Oct. 2021

  • Implemented a self-distillation network by combining the DINO backbone (Vision Transformer) and SimCLRv2 backbone, and improved the performances by 20% over the published baselines.

Course Project Student -Autonomous Systems Lab (ASL), ETHz

Feb. 2021 - Jun. 2021

  • Implemented a semi-supervised triplet-loss-based graph convolution network modified from DeepGCNs to learn a deep-learned descriptor for the objects constellation for global localization in SLAM.
  • Introduced semantic embedding and graph attention into the base network for learning and improved the performance both on the simulation data and real data.
  • Achieved over 95% top 5 ratio on simulation datasets and over 80% top 5 ratio on real datasets, and the precision for localization outperforms PointNet descriptor and Hand-crafted descriptors, such as Random Walk descriptor.

Deep Learning Research Intern -Mech-Mind Robotics Technologies Ltd.

Jul. 2020 - Sep. 2020

  • Collected RGBD image data sets by manipulating robotic arms to grasp densely aligned medicines and food wrapper boxes in real scenarios and implemented offline and online data augmentation.
  • Implemented Bi-Directional Cascade Network(BDCN) using Tensorflow/Keras, and tested different feature extraction convolution networks as the backbone and different network structures.
  • Improved the online detection speed by 10 times compared to the original network DexiNed, with the detection accuracy decreasing by only 5%.

Bachelor Thesis Student -Biomanufacturing Center, Tsinghua University

Feb. 2020 - Jun. 2020

  • Collected a private data set for the instance segmentation task of cervical cancer screening based on the clinical medical data offered by cooperative hospitals.
  • Implemented binary class and multi-class instance segmentation using Mask-RCNN model with ResNet50 and ResNet101 as the backbone to recognize and segment various cancer cells, including overlapping cytoplasm and nuclei, and visualized the results.
  • Achieved an mAP of 0.866 for cytoplasm segmentation on the public data set, 0.825 for nuclei segmentation, and an mAP of 0.742 cytoplasm segmentation on the clinical data set respectively at the level of IoU = 0.5.

Undergraduate Student Researcher -Advanced Nanomaterials & Microrobotics Lab, CUHK

Jul. 2019 - Sep. 2019

  • Designed multi-scale prototype circuit boards and diminished their sizes to 5mm x 8mm x 30mm
  • Tested the energy harvesting, signal transmission, and receiving efficiency of multi-scale prototype circuit boards, which were powered by DC power supply and copper-zinc primary battery respectively.
  • Designed a mechanical reciprocating propulsion system installed on a stomach model to simulate the creeping of patients’ stomachs.

Undergraduate Research Assistant -Institute of Automation, Chinese Academy of Sciences

Mar. 2019 - Jul. 2019

  • Realized the relocalization and automatic obstacle avoidance functions of disinfection robots used in hospital scenes, and conducted C++ secondary development based on the one-time pre-built map and real-time map interfaces offered by SLAMTEC’s APOLLO robot development platform.
  • Designed a user-friendly operator interface for Windows systems based on MFC, including functions-integrated buttons and real-time display of a 2D bitmap.

Research Training Student -Department of Mechanical Engineering, Tsinghua University

Oct. 2017 - Aug. 2018

  • Addressed time synchronization problem using GPS, and located UAVs using optical flow methods.
  • Completed the navigation of UAVs based on MAVROS and developed the autopilot for speed and gesture control through PX4 Firmware.
  • Completed the gazebo simulation of UAVs and tested the algorithms on the simulation platform.

Skills

Note: I think these sections are silly, but everyone seems to have one. Here is a *mostly* honest overview of my skills.

Linux
5 / 5
Chinese
5 / 5
LaTex
5 / 5
Python
5 / 5
Git
4 / 5
ROS
4 / 5
English
4 / 5
PyTorch
4 / 5
Bash
4 / 5
C++
4 / 5
Numpy
4 / 5
Scikit-Learn
4 / 5
Jupyter
3 / 5
Tensorflow + Keras
3 / 5
MATLAB
3 / 5
Matplotlib
3 / 5
Pandas
3 / 5
German
2 / 5
HTML + SASS/SCSS/CSS
2 / 5
R
2 / 5
Javascript
2 / 5
React
2 / 5

Selected Courses

References are available upon request