Papers
10
Total Citations
101
H-Index
6
About
Chang-Kyu Song is a roboticist whose research lies at the intersection of non-prehensile manipulation, differentiable physics, and automation. His core contributions focus on enabling robots to interact with unknown objects through sliding, pushing, and rearrangement—tasks critical for real-world applications like recycling and warehouse packing. Song’s most influential work, “Learning to Slide Unknown Objects with Differentiable Physics Simulations” (28 citations), pioneers the use of differentiable physics models to infer mechanical properties such as mass and friction distribution, allowing robots to stably push objects to goal configurations without prior knowledge. He further advanced this area with a probabilistic model for planar sliding (20 citations), which learns robust manipulation strategies under uncertainty. In “Object Rearrangement with Nested Nonprehensile Manipulation Actions” (20 citations), Song tackled the complex problem of planning manipulation sequences in confined spaces. His applied work includes “Toward Fully Automated Metal Recycling using Computer Vision and Non-Prehensile Manipulation” (8 citations), addressing the challenge of sorting scrap metal with attached impurities. Song’s research has garnered over 100 citations, demonstrating its impact on both theoretical robotics and practical automation. His early work on neuro-fuzzy navigation for mobile robots also showcases his versatility.
Research Focus
Key Achievements
Top Papers
- 1Learning to Slide Unknown Objects with Differentiable Physics Simulations28 citations · 2020
- 2Object Rearrangement with Nested Nonprehensile Manipulation Actions20 citations · 2019
- 3
- 4
- 5
- 6Identifying Mechanical Models through Differentiable Simulations6 citations · 2020
- 7
- 8Towards Robust Product Packing with a Minimalistic End-Effector2 citations · 2019
- 9
- 10