paper preview
01arXiv
Natural Sit-to-Stand Motion Synthesis For Humanoids via Guided Assistance Curricula and Staged Rewards
Teaches a G1 humanoid to stand up from chairs of different heights by gradually reducing assistance.
Paper Sweeps
Humanoid Papers from August 23 to August 29, 2026
3Locomotion1Motion Tracking2Loco-Manipulation2Whole-Body Control2Training Infrastructure2State Estimation / Perception
3 papers
Locomotion
Terrain, gait, robustness, and agile movement.
paper preview
02arXiv
Physics Filtering Favors the Generalization of Robot Learning
Filters a learned controller through simple physics so legged robots keep working on new terrain and payloads.
paper preview
03arXiv
CARO: Contact-Agnostic Residual Observation for Zero-Shot Robust Quadruped Locomotion
Adds a torque-residual body cue so a quadruped keeps walking through payload, terrain, and impact changes.
1 paper
Motion Tracking
Reference motion, imitation, retargeting, and behavior priors.
paper preview
04arXiv
Demonstration-Guided Humanoid Stand-Up on an Emulated Deformable Surface
Adapts stand-up demonstrations so a G1 humanoid can get up on soft, compliant ground.
2 papers
Loco-Manipulation
Whole-body task contact: feet, torso, hands, and objects.
project clip
05arXiv
DreamMimic: Learning Visuomotor Whole-Body Loco-Manipulation via World Model
Trains a humanoid to walk and manipulate objects from vision by distilling stronger simulated teachers.
project visual
06arXiv
TONAV: Task-Oriented Navigation and Action-Velocity Chunk Learning for Articulated Object Quadrupedal Mobile Manipulation
Moves a quadruped with an arm into the right pose, then controls contact-heavy object manipulation.
2 papers
Whole-Body Control
MPC, impedance, safety layers, and whole-body coordination.
paper preview
07arXiv
Guided Riemannian Optimization (GuRO): Bridging Model Predictive Control and Decision Transformers
Guides a quadruped’s walking decisions with online planning so it can cross rough terrain more reliably.
paper preview
08arXiv
LAC: Linear and Angular Compliance for Humanoid Whole-body Control
Gives a G1 humanoid tunable body compliance so it can stay controlled when the upper body makes contact.
2 papers
Training Infrastructure
Benchmarks, datasets, rewards, simulators, and tooling.
project clip
09arXiv
GOLEM: Modular Humanoid Autonomy Towards Electric Vehicle Battery Disassembly
Builds a modular autonomy stack so a humanoid can navigate, stay stable, and help disassemble EV batteries.
project clip
10arXiv
WarpSAC: Towards the Pinnacle of Scalable Off-policy RL by Rethinking Exploration and Exploitation
Speeds up G1 walking training by changing when the learner explores and when it exploits what already works.
2 papers
State Estimation / Perception
Terrain, state, slip, sensors, and local context.
paper preview
11arXiv
DELTA: Deformable Elevation-Based Local Terrain Attention Encoder for Sparse-Terrain Quadrupedal Locomotion
Helps a quadruped read sparse terrain maps so it can keep walking over rough ground without extra policy cost.
paper preview
12arXiv
RAEM: Robust Autonomous Exploration for Multi-Floor Environments with a Quadruped Robot
Lets a quadruped explore multi-floor buildings, climb stairs, and recover when local planning gets stuck.