# LeCAR-Lab/ASAP

[RSS 2025] "ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills"

Repository: https://github.com/LeCAR-Lab/ASAP
Canonical: https://ross.abutalabs.com/products/asap
Homepage: https://agile.human2humanoid.com/
Language: Python
License: MIT
License Family: permissive
Topics: humanoid, reinforcement-learning, robotics
Last push: 2026-01-06T04:58:36+00:00

## Health v2 (maintenance only)
Score: 48/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 61, release rhythm 35, longevity 41
- inputs: {"age_days": 583, "days_push": 239, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2100, forks 197 (observed 2026-08-28T04:06:13.588825+00:00)

## What it is
ASAP is a two-stage framework for training agile humanoid whole-body skills by aligning simulation and real-world physics. It pre-trains motion tracking policies in simulation, then trains a delta (residual) action model from real-world data to compensate for dynamics mismatch and fine-tune policies for sim2real deployment.

## Use cases
- train humanoid robots to perform agile whole-body motions like jumps and kicks
- close the sim-to-real gap for humanoid locomotion policies
- retarget human motion capture data (AMASS/SMPL) to arbitrary humanoid robots
- deploy motion tracking policies on real humanoid hardware
- fine-tune simulation-trained policies with a residual dynamics model

## When to choose
- you are doing humanoid robotics research on agile whole-body skills
- you need a sim2real pipeline built on IsaacGym, IsaacSim, or Genesis
- you want to retarget SMPL/AMASS human motions to a humanoid robot

## When to avoid
- you need a production-ready robotics middleware rather than research code
- you work with non-humanoid robots
- you lack GPU hardware or NVIDIA Isaac simulator access

## Facets
- artifact type: library
- maturity: active
- function: reinforcement-learning, simulation, robotics, machine-learning
- domain: robotics, reinforcement-learning, simulation, machine-learning
- platform: python
- tags: humanoid, sim2real, motion-tracking, isaacgym, isaacsim, genesis, motion-retargeting, delta-action-model, research-code, linux, gpu

## Member repositories
- LeCAR-Lab/ASAP (main) score 48

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:13.588825+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:54:19.875136+00:00, confidence not recorded.
  - readme: https://github.com/LeCAR-Lab/ASAP (fetched 2026-08-28T04:06:13.588825+00:00, sha 6c520642b2c8)
  - homepage: https://agile.human2humanoid.com/ (fetched 2026-08-29T10:34:37.049418+00:00, sha 6a4d863595f0)
- Data as of 2026-08-30T08:39:29.467469+00:00.