Kyrylo Kolesnichenko
I’m a final-year undergraduate at Vilnius University and a researcher at Kent State University’s Advanced Telerobotics Research Lab, advised by Jong-Hoon Kim.
My main interests are reinforcement learning and humanoid robotics. I develop and test control policies in simulation and on hardware, working on expressive motion, locomotion for teleoperation, and walking with different foot geometries. I also work on bipedal running and LLM-based human–robot interaction.
Specification
- robots
- Booster K1 ×8Pepper ×1Microduck ×1Unitree G1 ×1
- robotics / rl
- Isaac Lab/Sim · mjlab · MuJoCo · ROS 2 · PPO/RSL-RL · AMP · BeyondMimic · NVIDIA SONIC · GVHMR · GMR · sim-to-real
- ml / ai
- PyTorch · Hugging Face · vLLM · ONNX Runtime · OpenCV · scikit-learn · TensorFlow · LLMs · VLMs · RAG · FAISS
- languages
- Python · SQL · JavaScript/TypeScript · Bash · R · LaTeX
- systems
- Linux · Git · Docker · conda · CMake · CUDA · H100 GPU servers · FastAPI · PostgreSQL · React · Node.js · CAD · 3D printing
- record
- 12 projects · 3 papers · 41 videos and photos
Projects
- P-11SONIC-K1
NVIDIA's SONIC whole-body motion tracker, ported to the Booster K1 and trained from scratch with under 1% of the original compute. The first open-weights motion tracker for the K1.
From a monocular runway video to a catwalk gait on the Booster K1.
Our Ghost Trial 04 entry. A G1 walks the runway, turns, poses and walks back, by fine-tuning SONIC.
- P-04K1 Shoes
3D-printed shoes for the Booster K1. Heels for runway walks, plus Dutch clogs and a light vine pair for the Humanoid Robot Fashion Challenge. Open source, with MuJoCo and URDF models of the K1 wearing them.
- P-06K1 Loco-Teleop
A walking policy for the Booster K1 that drives only the legs, so the arms and head stay free for teleoperation.




