|
I am currently pursuing an M.Sc. in Data Science at City University of Hong Kong. I earned my B.Eng. in Biomedical Engineering from Guangdong Medical University, graduating with excellent academic results. I also received the university’s scholarship for graduating undergraduates. I was a visiting student in the Department of Electronic and Electrical Engineering at Southern University of Science and Technology, where I researched vision-based tactile perception and helped develop and integrate sensor SDKs and tactile data collection systems. I later joined STouching Robotics as an intern during its early stages. I worked on vision-based tactile algorithms and embodied tactile systems, and managed the sensor SDK team. Research Interest: My research focuses on building embodied systems that generalize in open-world settings. I aim to unify tactile perception and physical state representations to help robots understand the physical world. My interests include environmental perception, task planning, dexterous manipulation, and action correction. Research Question: Embodied tactile sensing and dexterous robotic manipulation. I am particularly interested in the full pipeline: sensors, physical representations, policy learning, and closed-loop control. The goal is to enable robots to perform contact-rich dexterous manipulation tasks in unstructured environments. These tasks include insertion and extraction, wiping, and assembly. Email: canyonchency@gmail.com |
LeRobot · ACT · π0 / π0.5 · Piper
An experimental framework for robot manipulation with vision and touch. Extends LeRobot with Piper data collection and explores tactile policy inputs and residual action correction, with FTP-1 and DreamTacVLA prototypes.
Isaac Sim · Isaac Lab · UIPC
A tactile simulation platform for robot contact tasks. Uses Isaac Sim, Isaac Lab, and UIPC to model compliant membrane deformation and generate tactile images and force fields, with extensible robot, sensor, and object setups.
Tactile Images · Physical Measurements · Tactile Recognition
A tactile recognition model combining tactile image sequences with six calibrated physical signals. Uses feature fusion and temporal modeling, and provides tools for data collection, training, offline inference, and real-time recognition.
Dense Optical Flow · ONNX · RKNN · RK3588 · RV1126B
A compact dense optical-flow network for visuotactile sensing. Estimates image displacement through magnitude estimation, direction reconstruction, and residual refinement, with tools for export, quantization, and multi-stream inference on RK3588 and RV1126B.
Medical Imaging · 3D Imaging Reconstruction · Model Evaluation
A local bilingual desktop workbench for learning and research in medical imaging. Combines image viewing, CT reconstruction experiments, model inference, and evaluation, with offline core workflows, guided lessons, and an optional AI assistant.