Qihao (Keith) Lin 林祺皓

Hi! I am an M.S. student at Carnegie Mellon University working on assistive robotics. My research interests lie in robot perception and manipulation, machine learning, and human-robot interaction. I am a member of the Robotic Caregiving and Human Interaction Lab, advised by Prof. Zackory Erickson.

I received my B.S. in Aerospace Engineering from Sun Yat-sen University. During my undergraduate research, I worked on robotic perception and manipulation with the Robotics and Intelligent Sensing Lab, advised by Chongkun Xia. My projects included deformable and transparent object sensing and manipulation.

In addition to academic research, I worked as a machine learning engineer intern at Shenzhen Haoya IoT Co., Ltd. on algorithm development for Chinese Sign Language translation systems. My work included on-device models and sign-language data processing.

Outside research, I enjoy music and sports. I was a singer and guitarist in the band Decide It Tomorrow, and I also enjoy basketball, jogging, and working out.

Email: qihaol@andrew.cmu.edu
CV: Qihao Lin CV

Selected Projects

In-bed assistance

Uncover to Re-cover: Model-Based Blanket Manipulation for In-Bed Assistance

Model-based blanket manipulation for in-bed assistance, connecting entropy-aware uncovering with a graph dynamics model fine-tuned on cloth smoothing. The system plans the blanket state left by uncovering to support subsequent re-covering, with simulation evaluation and transfer to a 10-participant human study.

Visual-tactile transparent fragment contour estimation pipeline

Visual-tactile perception

Transparent fragment reassembly

A visual-tactile framework for transparent fragment contour estimation and autonomous reassembly, supported by TransFrag27K, GelSight tactile sensing, material classification, and contour matching.

Paper Code

Chinese Sign Language translation system interface

Accessibility AI systems

Standardized-lexicon-scale Chinese Sign Language translation

Industry work on a Chinese Sign Language translation system connecting a multimodal computational CSL dictionary, CSL-to-text recognition, text-to-CSL rule-assembled signer rendering, and edge-deployable local interaction. The public-facing scope is around 8,000 CSL vocabulary items.

System Article Article