# RoboTwin-Platform/RoboTwin

[ICML 2026] RoboTwin 2.0 Offical Code Repo

Repository: https://github.com/RoboTwin-Platform/RoboTwin
Canonical: https://ross.abutalabs.com/products/robotwin
Homepage: https://robotwin-platform.github.io
Language: Python
License: MIT
License Family: permissive
Topics: benchmark, embodied-ai, robotics, data-generator
Last push: 2026-08-20T14:06:09+00:00

## Health v2 (maintenance only)
Score: 68/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 40, longevity 51
- inputs: {"age_days": 714, "days_push": 13, "days_rel": 188, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2772, forks 466 (observed 2026-08-28T04:07:19.068344+00:00)

## What it is
RoboTwin 2.0 is a scalable benchmark and data-generation platform for bimanual (dual-arm) robotic manipulation, built on simulated digital twins with structured domain randomization. It includes an object library of 731 instances, an MLLM-driven expert data synthesis pipeline, and unified evaluation protocols with a public leaderboard.

## Use cases
- benchmark dual-arm robot manipulation policies
- generate large-scale synthetic training data for robot learning
- evaluate sim-to-real transfer of manipulation policies
- run domain-randomized simulation experiments for embodied AI
- compare robot policy performance on a public leaderboard

## When to choose
- you need a standardized benchmark for bimanual manipulation
- you want automated large-scale synthetic robot data with domain randomization
- you are doing embodied AI / robot learning research and need reproducible evaluation

## When to avoid
- you need a single-arm-only or non-manipulation robotics benchmark
- you need a production robot control stack rather than a research benchmark
- you cannot run GPU-accelerated simulation locally

## Facets
- artifact type: framework
- maturity: active
- function: simulation, benchmarking, data-generation, machine-learning, robotics
- domain: robotics, artificial-intelligence, simulation, machine-learning
- platform: python
- tags: embodied-ai, bimanual-manipulation, domain-randomization, sim-to-real, digital-twin, benchmark, mllm, dataset-generation, linux, gpu

## Member repositories
- RoboTwin-Platform/RoboTwin (main) score 68

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:19.068344+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 2026-08-30T08:16:37.331328+00:00, confidence not recorded.
  - readme: https://github.com/RoboTwin-Platform/RoboTwin (fetched 2026-08-28T04:07:19.068344+00:00, sha 36eb5dce338f)
  - homepage: https://robotwin-platform.github.io (fetched 2026-08-29T09:56:02.055111+00:00, sha ea1f43fa671a)
- Data as of 2026-08-30T08:39:29.467469+00:00.