Pith. sign in

REVIEW 35 cited by

ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2502.01143 v3 pith:D5XAGIUP submitted 2025-02-03 cs.RO cs.AIcs.LGcs.SYeess.SY

classification cs.ROcs.AIcs.LGcs.SYeess.SY
keywords dynamicsreal-worldagileasapsimulationwhole-bodydeltahumanoid
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Humanoid robots hold the potential for unparalleled versatility in performing human-like, whole-body skills. However, achieving agile and coordinated whole-body motions remains a significant challenge due to the dynamics mismatch between simulation and the real world. Existing approaches, such as system identification (SysID) and domain randomization (DR) methods, often rely on labor-intensive parameter tuning or result in overly conservative policies that sacrifice agility. In this paper, we present ASAP (Aligning Simulation and Real-World Physics), a two-stage framework designed to tackle the dynamics mismatch and enable agile humanoid whole-body skills. In the first stage, we pre-train motion tracking policies in simulation using retargeted human motion data. In the second stage, we deploy the policies in the real world and collect real-world data to train a delta (residual) action model that compensates for the dynamics mismatch. Then, ASAP fine-tunes pre-trained policies with the delta action model integrated into the simulator to align effectively with real-world dynamics. We evaluate ASAP across three transfer scenarios: IsaacGym to IsaacSim, IsaacGym to Genesis, and IsaacGym to the real-world Unitree G1 humanoid robot. Our approach significantly improves agility and whole-body coordination across various dynamic motions, reducing tracking error compared to SysID, DR, and delta dynamics learning baselines. ASAP enables highly agile motions that were previously difficult to achieve, demonstrating the potential of delta action learning in bridging simulation and real-world dynamics. These results suggest a promising sim-to-real direction for developing more expressive and agile humanoids.

Discussion (0). Sign in to comment.

Forward citations

Cited by 35 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Light-Loco-Parkour: Versatile Perceptive Whole-Body Locomotion via Multi-Skill Distillation

    cs.RO 2026-08 conditional novelty 7.0 of 10

    A single neural-network policy, trained in simulation, makes a humanoid climb, vault, and traverse uneven terrain from onboard depth and a velocity command, with no skill labels or runtime motion graphs.

  2. Handroid: Bridging Dexterous Hand and Humanoid

    cs.RO 2026-07 conditional novelty 7.0 of 10

    A single 27-DoF body doubles as an anthropomorphic dexterous hand and a 0.33 m desktop humanoid, with a unified control stack for manipulation, locomotion, and embodiment switching.

  3. ARDY: Autoregressive Diffusion with Hybrid Representation for Interactive Human Motion Generation

    cs.GR 2026-07 accept novelty 7.0 of 10

    An autoregressive diffusion model with a hybrid explicit-root/latent-body representation generates real-time, controllable 3D human motion from text and spatial constraints.

  4. MOVE: Motion-Guided Few-Shot Video Object Segmentation

    cs.CV 2025-07 conditional novelty 7.0 of 10

    MOVE provides a new motion-guided few-shot video object segmentation benchmark, and the proposed DMA baseline outperforms six existing methods across all settings.

  5. $\omega$-0: A Latent Predictive World Action Model for Concurrent Humanoid Loco-Manipulation

    cs.RO 2026-08 conditional novelty 6.0 of 10

    A latent-predictive world-action model lets one humanoid policy coordinate locomotion and manipulation across 11 real-world household tasks, reporting 79-82% success.

  6. Weights or Skills? A Survey of Robot-Learning Techniques: from Action-Predicting Weights to Robots that Write their Own Skills

    cs.RO 2026-08 conditional novelty 6.0 of 10

    A taxonomy of robot learning on a weights-versus-skills axis, with a five-rung self-improvement ladder whose top cell (feedback plus memory plus search) holds only a few recent systems.

  7. GenTrack: Physical Alignment for Robot-Native Motion Generation and Zero-Shot Humanoid Tracking

    cs.RO 2026-08 conditional novelty 6.0 of 10

    Online co-training of a text-to-motion generator and a humanoid tracker on simulated G1 improves generator executability and zero-shot tracker coverage beyond static replay or one-way filtering.

  8. Scaling Behavior Foundation Model for Humanoid Robots

    cs.RO 2026-07 conditional novelty 6.0 of 10

    A scaling recipe for humanoid behavior foundation models—global-frame motion tracking, on-policy data quantity plus reference diversity, and a transformer with hyperspherical latents—cuts global tracking error by roug...

  9. Behavior Foundations for Quadruped Robots: ABot-C0 Technical Report

    cs.RO 2026-07 conditional novelty 6.0 of 10

    A multi-source 16,074-clip quadruped motion library plus a flow-matching generalist tracker shows empirical data scaling and zero-shot unseen tracking, integrated with all-terrain locomotion and real-robot deployment.

  10. Closing the Reality Gap: Zero-Shot Sim-to-Real Deployment for Dexterous Force-Based Grasping and Manipulation

    cs.RO 2026-07 conditional novelty 6.0 of 10

    Zero-shot sim-to-real RL policies on a five-finger hand achieve commandable grasp-force tracking and in-hand reorientation using dense tactile simulation, current-to-torque calibration, and actuator randomization.

  11. VOFA: Visual Object Goal Pushing with Force-Adaptive Control for Humanoids

    cs.RO 2026-05 unverdicted novelty 6.0 of 10

    VOFA combines a high-level visuomotor policy with a low-level force-adaptive controller to let humanoids push objects up to 17 kg to arbitrary goals using only noisy onboard vision, achieving over 80% real-world success.

  12. Kinodynamic Motion Retargeting for Humanoid Locomotion via Multi-Contact Whole-Body Trajectory Optimization

    cs.RO 2026-03 conditional novelty 6.0 of 10

    A physics-aware motion-retargeting pipeline that uses ground-reaction-force-derived heel-toe contacts produces dynamically feasible humanoid references and improves downstream imitation learning.

  13. Humanoid Everyday: A Comprehensive Robotic Dataset for Open-World Humanoid Manipulation

    cs.RO 2025-10 conditional novelty 6.0 of 10

    A 10,300-demonstration, 260-task multimodal humanoid manipulation dataset with baseline policy evaluations and a cloud evaluation platform.

  14. A Scalable Whole-body Motion Transfer via Implicit Kinodynamic Motion Retargeting

    cs.RO 2025-09 conditional novelty 6.0 of 10

    A neural retargeting pipeline maps human motion to humanoid robot motion at 5000+ frames per second using a shared latent space and physics-based fine-tuning, filtering noise and producing physically feasible trajectories.

  15. In-between Motion Generation Based Multi-Style Quadruped Robot Locomotion

    cs.RO 2025-07 conditional novelty 6.0 of 10

    A CVAE-based in-between motion generator creates multi-style quadruped gaits from sparse motion data, and the trained controller runs gallop, tripod, trotting, and pacing on a real robot.

  16. Robot Drummer: Learning Rhythmic Skills for Humanoid Drumming

    cs.RO 2025-07 conditional novelty 6.0 of 10

    A simulated Unitree G1 humanoid learns to drum dozens of popular songs from MIDI with high F1 scores using a Rhythmic Contact Chain and temporal decomposition.

  17. Learning Motion Skills with Adaptive Assistive Curriculum Force in Humanoid Robots

    cs.RO 2025-06 conditional novelty 6.0 of 10

    A2CF uses an adaptive assistive-force agent to guide humanoid robots through training, yielding faster convergence and robust policies that work without the external force.

  18. GMT: General Motion Tracking for Humanoid Whole-Body Control

    cs.RO 2025-06 conditional novelty 6.0 of 10

    GMT trains a single unified humanoid policy using adaptive sampling and mixture-of-experts, achieving lower tracking errors than a re-implemented ExBody2 across diverse whole-body motions.

  19. KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills

    cs.RO 2025-06 conditional novelty 6.0 of 10

    A robot control method that adaptively tightens motion-tracking reward tolerances achieves lower tracking errors on dynamic skills and transfers zero-shot to a real humanoid.

  20. From Experts to a Generalist: Toward General Whole-Body Control for Humanoid Robots

    cs.RO 2025-06 conditional novelty 6.0 of 10

    BumbleBee, an expert-to-generalist pipeline using autoencoder-based motion clustering and per-cluster delta action models, reports state-of-the-art whole-body control on a Unitree G1 humanoid, with success rates of 89...

  21. RL from Physical Feedback: Aligning Large Motion Models with Humanoid Control

    cs.RO 2025-06 reject novelty 6.0 of 10

    RLPF uses reinforcement learning with a physics-simulator tracking reward and an alignment verification module to fine-tune a large text-to-motion model for physically feasible humanoid motions.

  22. SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending

    cs.RO 2025-06 conditional novelty 6.0 of 10

    SkillBlender pretrains reusable goal-conditioned skills and blends them with softmax per-joint weights to solve simulated humanoid loco-manipulation tasks with one or two reward terms.

  23. MoRE: Mixture of Residual Experts for Humanoid Lifelike Gaits Learning on Complex Terrains

    cs.RO 2025-06 conditional novelty 6.0 of 10

    A two-stage reinforcement learning pipeline with a mixture of latent residual experts gives a Unitree G1 humanoid multiple commanded human-like gaits over complex terrains.

  24. Hold My Beer: Learning Gentle Humanoid Locomotion and End-Effector Stabilization Control

    cs.RO 2025-05 conditional novelty 6.0 of 10

    A slow-fast two-agent reinforcement learning architecture with separate upper- and lower-body policies reduces end-effector shaking during humanoid locomotion.

  25. ComplexMimic: Human-Scene Interaction Imitation in Complex 3D Environments

    cs.CV 2026-07 unverdicted novelty 5.5 of 10

    Dual-expert RL plus difficulty-aware multi-teacher distillation improves physics-based human–scene interaction imitation under complex 3D geometry versus prior single-policy baselines.

  26. Tired Actor: Fatigue-Informed Character Control

    cs.RO 2026-08 conditional novelty 5.0 of 10

    Injecting a muscle-fatigue model into a general physics-based character controller preserves motion imitation accuracy while producing tired, more human-like behaviors such as shorter steps, corner cutting, and fall c...

  27. A Single Diffusion-Policy Controller for Multi-Task Block Pushing with Zero-Shot Sim-to-Real Transfer

    cs.RO 2026-07 conditional novelty 5.0 of 10

    One diffusion policy trained via energy-guided RL solves multi-shape block pushing without demos and transfers zero-shot to real robots under varied conditions.

  28. ZeroWBC: Learning Natural Whole-Body Humanoid Interaction from Human Egocentric Data

    cs.RO 2026-03 conditional novelty 5.0 of 10

    An open-loop generation-then-tracking system maps one egocentric image plus language into Unitree G1 whole-body interactions using only human egocentric motion data.

  29. RobotDancing: Residual-Action Reinforcement Learning Enables Robust Long-Horizon Humanoid Motion Tracking

    cs.RO 2025-09 conditional novelty 5.0 of 10

    Residual-action reinforcement learning, with selective corrections on hip and knee pitch joints, enables zero-shot long-horizon dance tracking on real humanoid robots.

  30. HERMES: Human-to-Robot Embodied Learning from Multi-Source Motion Data for Mobile Dexterous Manipulation

    cs.RO 2025-08 conditional novelty 5.0 of 10

    HERMES converts a single human motion demonstration into a deployable mobile bimanual dexterous manipulation policy, using RL, depth-image distillation, and closed-loop PnP pose refinement.

  31. Quantifying and Visualizing Sim-to-Real Gaps: Physics-Guided Regularization for Reproducibility

    cs.RO 2025-07 reject novelty 5.0 of 10

    A gain-regularized, parameter-conditioned RNN balances a low-cost 110:1 gearbox robot with matching simulated and real settling times, while naive domain randomization oscillates.

  32. Hierarchical Vision-Language Planning for Multi-Step Humanoid Manipulation

    cs.RO 2025-06 conditional novelty 5.0 of 10

    A three-layer hierarchical system using a VLM planner and VLM skill monitor with imitation-learned skills and an RL tracking policy achieved 73% success on a real humanoid pick-and-place task.

  33. SLAC: Safe and Efficient Real-Robot Reinforcement Learning via Unsupervised Simulation Pre-Training

    cs.RO 2025-06 conditional novelty 5.0 of 10

    SLAC learns a latent action space in a low-fidelity simulator and uses it for real-world reinforcement learning, solving whole-body mobile manipulation tasks in under an hour without demonstrations.

  34. Theory Foundation of Physics-Enhanced Residual Learning

    cs.LG 2025-08 reject novelty 4.0 of 10

    A set of conditional bounds shows PERL's advantages follow from assumed smaller Lipschitz constant and loss ceiling, without proving those assumptions or connecting them correctly to neural network training.

  35. A Survey: Learning Embodied Intelligence from Physical Simulators and World Models

    cs.RO 2025-07 conditional novelty 4.0 of 10

    Embodied intelligence learning is reviewed through the complementary lenses of physical simulators and world models, with a proposed IR-L0 to IR-L4 robot capability taxonomy.

Pith tools