John (Chung Hee) Kim

I am a PhD student in the Robotics Institute at Carnegie Mellon University. I am currently a graduate student researcher at the Kantor Lab advised by Dr. George Kantor.

I received my B.Eng. in Mechanical Engineering and M.Phil. in Electronic and Computer Engineering from the Hong Kong University of Science and Technology, where I worked as a research assistant in the Robot Manipulation Lab advised by Dr. Jungwon Seo.

My research interest lies in the intersection of robotic manipulation, perception, and reinforcement learning. I enjoy solving real world problems by working with robots to develop practical systems and solutions.

We present a framework for learning robotic contact manipulation of tree-like crops by leveraging graph representations.

We develop a computer vision method to extract skeleton representation of trees from images featuring large amounts of foliage and self-occlusion.

We develop a method for creating 3D models of sorghum panicles and a non-destructive approach to estimate seed count and weight. This is acheived using seeds as semantic landmarks in both 2D and 3D and a novel density-based clustering approach. Additionally, we present a new metric to evaluate reconstruction quality in the abscense of ground-truth.

This enables human-level dexterity in the tasks of ungrasping, e.g., the placement of Go stones that AlphaGo is unable to perform.

There's more to our Dexterous Ungrasping -- robustness.

Effective bin picking using an asymmetric gripper with different finger lengths.

We present an effective manipulation technique for picking thin objects from a flat surface.

We address how to insert a thin object (such as the phone battery) into a shallow-depth hole through dexterous manipulation.

A system and method for accurately placing or inserting objects using robotic manipulation is disclosed.

Stolen from source code. Last updated Jan 29, 2024.