Li-Heng Lin

I am a second year master student in the computer science department at Stanford University and a member of ILIAD advised by Dorsa Sadigh. I completed my undergraduate degree at National Taiwan University where I worked with Professors Chung-Wei Lin and Hsuan-Tien Lin.

Email  /  CV  /  Google Scholar  /  Github  /  LinkedIn

profile photo

Research

I'm boardly interesteed in machine learning and robotics, especially when human factors are considered in the loop.

Distilling and Retrieving Generalizable Knowledge for Robot Manipulation via Language Corrections
Lihan Zha, Yuchen Cui, Li-Heng Lin, Minae Kwon, Montserrat Gonzalez Arenas, Andy Zeng, Fei Xia, Dorsa Sadigh
International Conference on Robotics and Automation (ICRA) 2024
paper / website

Propose a LLM-based system that responds to language feedback, distills knowledge from corrections, and retrieves relevant past knowledge.

Gesture-Informed Robot Assistance via Foundation Models
Li-Heng Lin, Yuchen Cui, Yilun Hao, Fei Xia, Dorsa Sadigh
Conference on Robot Learning (CoRL) 2023
paper / website

Enable robots to understand human gestures and generate corresponding plans using LLMs.

Graph-Based Deadlock Analysis and Prevention for Robust Intelligent Intersection Management
Kai-En Lin, Kuan-Chun Wang, Yu-Heng Chen, Li-Heng Lin, Chung-Wei Lin, Iris Hui-Ru Jiang
ACM Transactions on Cyber-Physical Systems

Refine the graph-based intersection model and propose a robustness-aware greedy scheduling algorithm.

Reinforcement-Learning-Based Job-Shop Scheduling for Intelligent Intersection Management
Shao-Ching Huang, Kai-En Lin, Cheng-Yen Kuo, Li-Heng Lin, Muhammed O. Sayin, Chung-Wei Lin
Design, Automation & Test in Europe Conference & Exhibition (DATE) 2023

Apply reinforcement learning to solve the scheduling problem on graph-based intersection models.

Deadlock Resolution for Intelligent Intersection Management with Changeable Trajectories
Li-Heng Lin, Kuan-Chun Wang, Ying-Hua Lee, Kai-En Lin, Chung-Wei Lin, Iris Hui-Ru Jiang
IEEE Intelligent Vehicles Symposium (IV) 2022

Ensure deadlock free by proposing a protection mechanism based on limiting the number of vehicles. The whole system reduces vehicle wait time by 52% on average compared to traditional traffic light systems.

Industry

Software Engineering Intern, Google
Summer 2021

Developed an Android app "Braille Image Translator" that takes in an image of a braille device and outputs its corresponding text.


Website template is adapted from here.