paper preview
01arXiv
Isaac Sim-to-Real: Reinforcement Learning based Locomotion for Quadrupeds
Teaches a Unitree Go1 to walk from Isaac Lab simulation by shaping rewards, modeling actuators, and testing push recovery.
Paper Sweeps
Humanoid Papers from July 19 to July 25, 2026
3Locomotion4Motion Tracking3Loco-Manipulation1Whole-Body Control2Training Infrastructure
3 papers
Locomotion
Terrain, gait, robustness, and agile movement.
paper preview
02arXiv
DASH Robot: Minimalistic Design and Optimal Aerial-Terrestrial Locomotion via Contact-Implicit Control
Builds a small robot that can hop and fly, with a controller that chooses when to use ground contact or thrust.
paper preview
03arXiv
Towards Torque-Driven Reinforcement Learning for Quadruped Locomotion
Trains a Unitree B1 quadruped to walk using direct torque commands instead of position targets or velocity sensing.
4 papers
Motion Tracking
Reference motion, imitation, retargeting, and behavior priors.
paper preview
04arXiv
From Sign Language Generation to Humanoid Execution: Vision-Language Guided Retargeting with Collision Mitigation
Turns generated sign-language body motion into humanoid joint motion by cleaning collisions and retargeting the poses.
paper preview
05arXiv
What Matters in Humanoid General Motion Tracking? An Empirical Study
Tests which motion-tracking design choices matter most for making a Unitree G1 follow whole-body reference motions.
paper preview
06arXiv
Extreme-RGMT: Continual Learning of Highly Dynamic Skills for Robust Generalist Humanoid Control
Helps humanoids learn dynamic skills like flips while keeping reliable everyday motion tracking.
project clip
07arXiv
Scaling Behavior Foundation Model for Humanoid Robots
Builds a humanoid behavior model that scales motion tracking across many references and robot deployment paths.
3 papers
Loco-Manipulation
Whole-body task contact: feet, torso, hands, and objects.
paper preview
08arXiv
Closing the Lab-to-Store Gap: A Data-Efficient Post-Training and Experience-Driven Learning VLA Framework for Retail Humanoids
Improves a G1 retail humanoid by adding targeted post-training data from real task failures and experience.
paper preview
09arXiv
Towards Miniature Humanoid Tele-Loco-Manipulation Using Virtual Reality and Reinforcement Learning
Lets a small ROBOTIS OP3 humanoid walk while a VR operator controls its arms to move objects.
paper preview
10arXiv
Closing the Loop in Humanoid VLA: Persistent 3D Object Tokens for Verifiable Loco-Manipulation
Gives a G1 humanoid persistent 3D object memory so it can act on objects and verify task progress.
1 paper
Whole-Body Control
MPC, impedance, safety layers, and whole-body coordination.
project clip
11arXiv
FARO: Feasibility-Aware Robot Motion Optimization
Filters humanoid contact plans before expensive optimization, then tracks the selected motion on real hardware.
2 papers
Training Infrastructure
Benchmarks, datasets, rewards, simulators, and tooling.
paper preview
12arXiv
Predictive Training with Latent Imagination for Visual Quadruped Navigation
Trains quadrupeds to navigate moving obstacles by imagining short future states during training without slowing deployment.
project clip
13Project
UFO: A General Unsupervised Reinforcement Learning Framework for Humanoid Control
Trains humanoid control skills without task rewards, then uses latent commands for robust movement and teleoperation.