physical-superintelligence-lab

physical-superintelligence-lab / SIMPLE

Public

Welcome to SIMPLE, a full-stack simulation environment for humanoid loco-manipulation, built on AMO/SONIC, with integrated support for mainstream VLAs such as Psi-0, Pi05, GR00T, and more.

44
0
100% credibility
Found Apr 16, 2026 at 44 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

SIMPLE is a simulation platform for benchmarking robot policies on manipulation and locomotion tasks using various robots and thousands of 3D assets.

How It Works

1
🔍 Discover SIMPLE

You find SIMPLE, a playground for testing robot brains in realistic simulations with arms, wheeled bots, and humanoids.

2
📦 Set up your simulator

Choose an easy install like quick package manager or container to get your virtual robot world running.

3
📥 Grab robot challenges

Download ready-made tasks and scenes with everyday objects for your robots to practice on.

4
🚀 Launch policy tests

Pick a robot policy and watch it tackle challenges like picking objects or walking while manipulating.

5
📹 Watch the action

See videos of robots succeeding or struggling, with live stats on how well they perform.

🏆 Compare and improve

Review success rates across policies to pick winners and refine your robot skills.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 44 to 44 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is SIMPLE?

SIMPLE delivers a full-stack Python simulation environment for humanoid locomotion-manipulation tasks, letting you evaluate vision-language-action models like Psi-0, GR00T, and Pi05 on Unitree G1 humanoids, Franka arms, or Aloha grippers. It packs 1000+ Objaverse assets, 50+ Habitat scenes, and 50+ benchmark tasks into Isaac Sim or MuJoCo backends, with CLI tools for datagen, replay, and leveled evals (visuals, lighting, poses). Like a simple GitHub repo for robot sims, it handles policy serving, teleop, and leaderboards out of the box.

Why is it gaining traction?

It stands out by integrating mainstream VLAs as drop-in baselines with success rates across OOD levels, skipping weeks of glue code. Nix, UV, or Docker setups mirror simple GitHub action workflows – bootstrap once, eval Psi-0 vs GR00T instantly. Low barrier hooks robotics folks tired of custom sim forks.

Who should use this?

Humanoid robotics researchers benchmarking VLAs on loco-manip; sim-to-real teams needing Objaverse/HSSD randomization; students building simple GitHub projects for embodied AI courses, like G1 pick-and-place evals.

Verdict

Worth forking for VLA humanoid benchmarks – CLI evals and leaderboards shine despite 44 stars and 1.0% credibility score signaling early days. Docs WIP, but Nix bootstrap keeps it simple; mature enough for research prototypes.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.