paper preview
01arXiv
Robust bipedal locomotion on flowable slopes via foot-driven terrain manipulation
Keeps bipedal robots walking on loose slopes by using the feet to shape and push the ground.
Paper Sweeps
Humanoid Papers from July 12 to July 18, 2026
7Locomotion2Motion Tracking2Loco-Manipulation2Whole-Body Control2Training Infrastructure1State Estimation / Perception
7 papers
Locomotion
Terrain, gait, robustness, and agile movement.
paper preview
02arXiv
Stop to Decide: Latency-Aware Proprioceptive Navigation Primitives for Mapping-Free Quadruped Inspection
Helps a quadruped slow down or stop before hard turns and stairs so navigation stays stable.
paper preview
03arXiv
Learning Roller-Skating Motions of Humanoid Robots Based on Adversarial Motion Priors
Teaches humanoids to roller-skate by copying skating motion patterns and learning how to balance on passive wheels.
project clip
04arXiv
PUMA: Perception-driven Unified Foothold Prior for Mobility Augmented Quadruped Parkour
Helps quadrupeds choose better footholds for parkour over gaps, stepping stones, and tilted walls using depth perception.
project clip
05arXiv
Agile perceptive multi-skill locomotion for quadrupedal robots in the wild
Teaches quadrupeds multiple outdoor movement skills so they can cross rough terrain from vision and body sensing.
paper preview
06arXiv
Learning Agile Navigation in Crowded Environments for Quadruped Robots
Teaches quadrupeds to move through crowds by choosing fast paths that avoid people and obstacles.
project clip
07X
GaitSpan: Growing Humanoid Locomotion from Walking to Running
Grows a humanoid walking controller into one policy that can walk, jog, and run on command.
2 papers
Motion Tracking
Reference motion, imitation, retargeting, and behavior priors.
project clip
08arXiv
EgoHTR: Egocentric 4D Demonstrations of Human Terrain Traversal
Turns egocentric human terrain videos into demonstrations that humanoids can use for walking over complex ground.
project clip
09arXiv
Scaling Behavior Foundation Model for Humanoid Robots
Builds one humanoid behavior model that can track many full-body motion references across different robot setups.
2 papers
Loco-Manipulation
Whole-body task contact: feet, torso, hands, and objects.
paper preview
10arXiv
PAKE: Learning Whole-Body Loco-Manipulation with Partial Kinematic Embeddings
Teaches a quadruped with an arm to move its body and manipulate objects while tracking only the important motion cues.
paper preview
11arXiv
Vision-Based Dribbling for Humanoid Soccer via Privileged Representation Learning
Teaches a humanoid to dribble a soccer ball using onboard vision while avoiding simple opponents.
2 papers
Whole-Body Control
MPC, impedance, safety layers, and whole-body coordination.
paper preview
12arXiv
WarpMPC: Large-Batch MPC on GPU via ADMM with Unrolled LDL^T Factorization
Runs many robot motion plans on a GPU so humanoid whole-body controllers can react faster.
project clip
13arXiv
Semantic Audio-driven Understanding for Dynamic Humanoid Whole Body Control
Uses sound and language cues to make a humanoid whole-body controller react to dynamic events around it.
2 papers
Training Infrastructure
Benchmarks, datasets, rewards, simulators, and tooling.
paper preview
14arXiv
SKooP: Symmetric Koopman Predictions for Faster and More Generalizable Legged Robot Locomotion with Reinforcement Learning
Speeds up legged robot training by using left-right body symmetry to predict more useful future motion.
paper preview
15arXiv
Deployable Human Preference Alignment in Robotics: Learning Representative Rewards from Diverse Human Preferences
Learns a small set of robot rewards from different people so deployed robots can match user preferences with fewer labels.
1 paper
State Estimation / Perception
Terrain, state, slip, sensors, and local context.
paper preview
16arXiv
StratMamba: Strategic and Reactive Stream Partitioning for Path-Efficient LiDAR-Based Obstacle Avoidance
Helps quadrupeds avoid obstacles by splitting LiDAR memory into quick reactions and longer route planning.