Paper Sweeps

Humanoid Papers from August 9 to August 15, 2026

3Locomotion2Loco-Manipulation1Whole-Body Control2Training Infrastructure

3 papers

Locomotion

Terrain, gait, robustness, and agile movement.

paper preview Spatiotemporal Agility: Time-Constrained Reinforcement Learning for Vision-Guided Dynamic Quadrupedal Interception visual

01arXiv

Spatiotemporal Agility: Time-Constrained Reinforcement Learning for Vision-Guided Dynamic Quadrupedal Interception

Teaches a Go2 to chase and intercept moving objects by predicting where and when they will land.

Yidong Zhu et al.

project clip

02arXiv

Learning Fault-Tolerant Locomotion with Adaptive Gait Timing

Keeps a large quadruped walking after actuator failures by changing gait timing from body-sensor history.

Giovanbattista Gravina et al.

paper preview Hip Energized Monopedal Hopping visual

03arXiv

Hip Energized Monopedal Hopping

Makes a one-legged robot hop faster by reusing hip torque to add energy while staying balanced.

Shane Rozen-Levy et al.

2 papers

Loco-Manipulation

Whole-body task contact: feet, torso, hands, and objects.

paper preview LUCID: Latent-Skill Unified Control via Imagined Dynamics for Long-Horizon Humanoid Loco-Manipulation visual

04arXiv

LUCID: Latent-Skill Unified Control via Imagined Dynamics for Long-Horizon Humanoid Loco-Manipulation

Combines reusable humanoid skills with imagined future dynamics so a humanoid can finish long object-interaction tasks.

Cheng Guo et al.

project clip

05arXiv

Learning Loco-Manipulation From SMPC Demonstrations With Sparse Offline-to-Online RL

Turns MPC demonstrations into sparse-reward policies for moving and manipulating on Spot and G1.

Martin Schuck et al.

1 paper

Whole-Body Control

MPC, impedance, safety layers, and whole-body coordination.

project clip

06arXiv

Whole-Body Planning for Humanoids Navigating Confined Spaces via Self-Collision Avoidance References

Plans G1 whole-body paths through tight spaces by shaping poses that avoid self-collisions and support contacts.

Carlos Gonzalez et al.

2 papers

Training Infrastructure

Benchmarks, datasets, rewards, simulators, and tooling.

paper preview Curriculum Generation under Structured Parametric Environments for Robust Navigation Policies visual

07arXiv

Curriculum Generation under Structured Parametric Environments for Robust Navigation Policies

Makes robot navigation training harder at the right pace so walking and driving policies become more robust.

Prishita Ray

paper preview An AI Scientist that Doesn't Drift: Taste, Structure, and Falsifiable Findings in a Quadruped Navigation Research Loop visual

08arXiv

An AI Scientist that Doesn't Drift: Taste, Structure, and Falsifiable Findings in a Quadruped Navigation Research Loop

Runs a structured research loop that designs and tests quadruped navigation experiments without drifting from falsifiable claims.

Yiwen Zhang et al.