# rail-berkeley/hil-serl

Repository: https://github.com/rail-berkeley/hil-serl
Canonical: https://ross.abutalabs.com/products/hil-serl
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2025-10-27T16:39:08+00:00

## Health v2 (maintenance only)
Score: 44/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 49, release rhythm 35, longevity 49
- inputs: {"age_days": 688, "days_push": 310, "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 1487, forks 205 (observed 2026-08-28T04:04:52.078485+00:00)

## What it is
HIL-SERL is a Python library suite for training reinforcement learning policies for precise robotic manipulation using human demonstrations and human-in-the-loop corrections. It includes environment wrappers, an actor-learner training infrastructure, and examples for real robot arms such as the Franka.

## Use cases
- train RL policies for robotic manipulation from demonstrations
- fine-tune robot policies with human corrections during training
- run actor-learner asynchronous RL on real robot hardware
- control a Franka arm with impedance-based gym environments
- achieve near-perfect success rates on dexterous manipulation tasks

## When to choose
- you need sample-efficient RL for real-world robot manipulation
- you have a Franka arm or compatible robot and want human-in-the-loop training
- you want a research-grade JAX-based RL stack for robotics

## When to avoid
- you need simulation-only RL without robot hardware
- you want a plug-and-play product rather than a research codebase
- your robot is not supported by the provided gym environments

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, reinforcement-learning, robotics, sdk
- domain: robotics, machine-learning, reinforcement-learning
- platform: python
- tags: human-in-the-loop, robotic-manipulation, jax, franka-arm, rl-training, linux, gpu

## Member repositories
- rail-berkeley/hil-serl (main) score 44

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:52.078485+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 2026-08-30T04:33:48.754438+00:00, confidence not recorded.
  - readme: https://github.com/rail-berkeley/hil-serl (fetched 2026-08-28T04:04:52.078485+00:00, sha c30f29d814eb)
- Data as of 2026-08-30T08:39:29.467469+00:00.