Ziwen Zhuang (庄子文)

or Leo Zhuang

I am a master's student at ShanghaiTech University. And I am currently working as a research assistant with Professor Hang Zhao at Shanghai Qi Zhi Institute (affiliated to IIIS, Tsinghua University).

Feel free to contact me if you are interested in working on self-supervised robotic exploration and learning for control.

This page is probability out of date. Migrating to ziwenzhuang.github.io

Researches and Publications

Robot Parkour Learning

Under review

Adversarially Robust Imitation Learning

Jianren Wang*, Ziwen Zhuang*, Hang Zhao
Published on CoRL 2021
Available online: Adversarially Robust Imitation Learning | OpenReview

SEMI: Self-supervised Exploration via Multisensory Incongruity

Jianren Wang*, Ziwen Zhuang*, Hang Zhao
Published on ICRA 2022
Arxiv preprint: https://arxiv.org/abs/2009.12494

Tactile only Active Sensing using Reinforcement Learning

Ziwen Zhuang, Jianren Wang, David Held
CMU RISS summer research internship 2019
Poster, Journal (page 209)

Projects

Fully Autonomous Cotton Harvesting System

Research Project with a cotton farming project, with Jiawei Hou, Yifan Cao, Chenguang Deng under MARS lab at ShanghaiTech University.
Used FlexBE state machine to integrate all the modules.
Integrating FlexBE, MoveIt!, cotton recognition and localization, cotton picking module both on software level and hardware level.
Conducted several field trials.

Semi-supervised Video Objects Segmentation

Bachalor Thesis
A code base to implement and compare Video Object Segmentation algorithms at VideoObjSeg
Reproduced STM and achieved 71.11% Jacard Index on DAVIS 2017 validation set.
Proposed improvement on Enhanced Space-Time Memory Network at 71.11% Jacard Index on DAVIS 2017 vali-dation set which trained only on image datasets.

Meta RL framework integrated in rlpyt

Compatible performance as PEARL opensource code.
More comprehensive code structure than PEARL with seperated experiment tool
More readable code descriptions and extendable modules.
Code: https://github.com/ZiwenZhuang/rlpyt

Simulation in Social Network Cooperation Repeated Games

Ziyuan Hu, Ziwen Zhuang, Yincen Xia
Undergraduate Research Project with Prof. Dengji Zhao
Simulated Social Network cooperation graph.
Developed Memory-Augmented Monte Carlo Tree Search in Reinforcement Learning framework.

Digitize City Roads Structure using Image Process Technology

Letong Wang, Yihan Wang, Ziwen Zhuang
Undergraduate Research Project with Prof. Dengji Zhao
Code: https://github.com/ZiwenZhuang/DiDiProject

Customized experiment tools

Ziwen Zhuang exptools
Being able to deploy multiple variants in one command and easy to analyze results based on your experiment variants.
Extracted from rlpyt and integrating some of the utilities of tensorboard
Due to existance of an exptools package at pip, this package has to be installed by cloning.