Skip to content
RobothecaBeta
Log in

HIL-SERL

  • Software
  • Humanoids and robot learning
  • Framework

A set of libraries, environment wrappers and examples for training robot manipulation policies with reinforcement learning, using demonstrations and human corrections.

Made by Robotic AI & Learning Lab Berkeley.

Source on GitHub: github.com/rail-berkeley/hil-serl (opens in a new tab). Facts read at commit c32939b (opens in a new tab).

Licenses

Code
Apache-2.0 (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit c32939b (opens in a new tab), on 16 Oct 2026.

Commercial use
Yes

What it does

  • Trains RL policies for robotic manipulation from a mix of demonstrations and human corrections.
  • Actor and learner nodes run asynchronously and exchange data over the network, with the learner periodically syncing the policy to the actor.
  • Includes agents such as SAC and BC, Gym environment wrappers, a replay buffer and data store, and vision models.
  • Provides scripts for policy training, demonstration data collection and reward classifier training.
  • Robot infrastructure includes a Flask server that sends commands to the robot via ROS and a Gym environment for the Franka robot.
  • Includes step-by-step walkthroughs for tasks such as RAM insertion, USB pickup and insertion, object handover and egg flip.

Written by Robotheca from the project's README (opens in a new tab) at commit c32939b.

Getting started

Difficulty
Advanced
Hardware it needs, as the project states it
Real-robot use is for a Franka arm and needs the serl_robot_infra setup with ROS; no other hardware requirements are stated.

Maintenance

Status
Maintained: a commit in the last 12 months
Newest commit
27 Oct 2025 on main
Latest release
None published on GitHub

Checked on GitHub on 16 Oct 2026.

Links

Robots that can run it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.