Skip to content
View RICHKUNG's full-sized avatar

Highlights

  • Pro

Block or report RICHKUNG

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
richkung/README.md

Hi, I'm Rich (Hsiang-Yu Kung)

I'm an MS Robotics student at the University of Michigan, and I want to build autonomous vehicles. At Michigan I've built a vision-guided manipulation stack on a real 6-DOF arm.

Before Michigan I studied Electrical Engineering at National Tsing Hua University in Taiwan. There I did research on open-vocabulary 3D perception, implemented planning, control and reinforcement-learning methods for robot navigation, and built small robots from the firmware up.

LinkedIn · Email

Featured work

Kinematics, RGB-D calibration and a height-based block detector on a real Lite 6 arm. 3rd of 12 teams; 24-block tower. ROB 550, University of Michigan.

Lite 6 arm carrying a side tower toward a tall rainbow tower, beside the block detector rejecting a hand and a phone

Implemented A*/RRT* planners and compared vehicle-tracking controllers on three simulated tracks.

RRT* search tree and final path through an obstacle map

Integrated YOLO, depth-based localization and visual servoing into a rover mission stack in Unity.

Unity rover and its camera view with a bear detection and depth

Training-free 3D part segmentation from RGB-D video, with hierarchical masks, temporal tracking and analysis of granularity failures. Undergraduate research, Vision Science Lab, NTHU.

FamilyPart single-frame example: RGB input and L4 drawer parts within an L2 parent

More projects

Toolbox

  • Manipulation: DH / PoX kinematics · closed-form IK · MuJoCo
  • Planning & control: ROS 2 · Nav2 · A* / RRT* · PID / Pure Pursuit / Stanley / LQR · PPO
  • Perception: RGB-D calibration · AprilTag · OpenCV · YOLO · SAM2 · CLIP / SigLIP · point clouds
  • Embedded: STM32 · ESP32 · Mbed OS · MQTT
  • General: Python · C++ · Linux · Git

Pinned Loading

  1. motion-planning-and-control motion-planning-and-control Public

    A*, RRT*, curvature-limited speed profiles, and PID / Pure Pursuit / Stanley / LQR path tracking compared on F1 tracks in simulation.

    Python

  2. ros2-autonomous-pick-and-place ros2-autonomous-pick-and-place Public

    ROS 2 mission stack for a rover with a 5-DOF arm: YOLO + depth backprojection, visual servoing and grasp verification, with Nav2-based and Nav2-free controllers.

    Python

  3. FamilyPart FamilyPart Public

    Training-free, open-vocabulary 3D part instance segmentation from posed RGB-D video (Semantic-SAM + SAM2 + SigLIP). Undergraduate research, Vision Science Lab, NTHU.

    Python 1

  4. rne-reinforcement-learning rne-reinforcement-learning Public

    PPO implementation and reward-design case studies for simulated robotic navigation.

    Python

  5. STM32-Line-Follower-MQTT STM32-Line-Follower-MQTT Public

    Mbed OS line-following car (B-L4S5I-IOT01A) that decides everything on board, with an LVGL touchscreen dashboard (STM32F769I-DISCO) over MQTT.

    C++

  6. lite6-vision-pick-and-place lite6-vision-pick-and-place Public

    Vision-guided pick-and-place on a real 6-DOF Lite 6 arm: kinematics, RGB-D calibration, height-based block detection. 3rd of 12 teams.