Festivus

Policies

I want to train a robot

Find policies, datasets, and simulation environments that work with your hardware

25 policies

ACT/AlohaTransferCube

Verified on real hardwareManipulation

transfers a cube from one arm to the other using bimanual coordination

Robots:ALOHA 2
Tested on 1 robot, 1 env, 500 episodesLeRobot

ACT/AlohaInsertion

Verified on real hardwareManipulation

inserts a peg into a socket using coordinated bimanual manipulation

Robots:ALOHA 2
Tested on 1 robot, 1 env, 500 episodesLeRobot

DiffusionPolicy/PushT

Verified on real hardwareManipulation

pushes a T-shaped block to a target pose on a tabletop

Robots:Franka Emika Panda
Tested on 1 robot, 1 env, 10,200 episodesPyTorch

DiffusionPolicy/RobomimicCan

Verified on real hardwareManipulation

picks a can from the table and places it in a target bin

Robots:Franka Emika Panda
Tested on 1 robot, 1 env, 300 episodesPyTorch

Octo-Base

Verified on real hardwareManipulation

executes diverse manipulation tasks conditioned on language or goal images across multiple robot embodiments

Robots:Franka Emika PandaALOHA 2Hello Robot Stretch RE1
Tested on 22 robots, 9 envs, 900,000 episodesJAX

Octo-Small

Verified on real hardwareManipulation

executes manipulation tasks with a lightweight generalist model suitable for fine-tuning

Robots:Franka Emika PandaSO-100Koch v1.1
Tested on 22 robots, 5 envs, 850,000 episodesJAX

pi0-Base

Verified on real hardwareManipulation

performs diverse dexterous manipulation tasks including folding laundry, busing tables, and assembling boxes

Robots:Franka Emika PandaALOHA 2
Tested on 7 robots, 5 envs, 10,000 episodesJAX

RT-2-X

Verified on real hardwareManipulation

follows natural language instructions to manipulate objects on a tabletop across different robot embodiments

Robots:Franka Emika Panda
Tested on 6 robots, 3 envs, 130,000 episodesTensorFlow

LeRobot/DiffusionPushT

Verified on real hardwareManipulation

pushes a T-shaped block to a target pose using learned diffusion-based action generation

Robots:Franka Emika Panda
Tested on 1 robot, 1 env, 206 episodesLeRobot

VINN/FrankaPickPlace

Verified on real hardwareManipulation

picks and places objects by retrieving nearest-neighbor actions from a visual demonstration library

Robots:Franka Emika Panda
Tested on 1 robot, 2 envs, 20 episodesPyTorch

WalkTheseWays/Go2

Verified on real hardwareLocomotion

locomotes with diverse gait styles (walk, trot, bound, pronk) controlled via a continuous gait parameter space

Robots:Unitree Go2
Tested on 2 robots, 5 envs, 500,500 episodesIsaacLab

ExtremeParkour/Go2

Verified on real hardwareLocomotion

performs parkour maneuvers including climbing, gap jumping, and leaping over obstacles

Robots:Unitree Go2
Tested on 2 robots, 8 envs, 2,000,200 episodesIsaacLab

MobileALOHA/Cooking

Verified on real hardwareManipulation

performs mobile kitchen tasks like sauteing and wiping while navigating on a wheeled base

Robots:ALOHA 2
Tested on 1 robot, 1 env, 50 episodesLeRobot

ACT/SO100PickPlace

Verified on real hardwareManipulation

picks small objects from a table and places them at a target location

Robots:SO-100
Tested on 2 robots, 1 env, 200 episodesLeRobot

ACT/AlohaClothFolding

Verified on real hardwareManipulation

folds rectangular towels and T-shirts using coordinated bimanual manipulation

Robots:ALOHA 2
Tested on 1 robot, 2 envs, 300 episodesLeRobot

RT-1

Verified on real hardwareManipulation

follows natural language instructions for tabletop pick-and-place and drawer manipulation

Robots:Franka Emika Panda
Tested on 13 robots, 3 envs, 130,000 episodesTensorFlow

LeRobot/TDMPC-SimXArm

Tested in simulationManipulation

lifts an object using model-predictive control with a learned world model in the simulated xArm environment

Robots:Franka Emika Panda
Tested on 1 robot, 1 env, 25,000 episodesLeRobot

H1-BipedalLocomotion

Tested in simulationLocomotion

walks forward on flat ground at commanded velocities up to 1.5 m/s

Robots:Unitree H1
Tested on 1 robot, 2 envs, 1,000,030 episodesIsaacLab

DiffusionPolicy/KochPickPlace

Tested in simulationManipulation

picks small cubes and places them into a target cup

Robots:Koch v1.1
Tested on 1 robot, 1 env, 150 episodesLeRobot

G1-WholeBodyControl

Tested in simulationManipulation

walks to objects and manipulates them using coordinated whole-body motion

Robots:Unitree G1
Tested on 1 robot, 2 envs, 500,000 episodesIsaacLab

Stretch/HomeNavigation

Tested in simulationNavigation

navigates through home environments to reach specified goal positions while avoiding furniture

Robots:Hello Robot Stretch RE1
Tested on 1 robot, 10 envs, 50,030 episodesHabitat

BC-RNN/RobomimicSquare

Tested in simulationManipulation

places a square nut onto the correct peg in simulation

Robots:Franka Emika Panda
Tested on 1 robot, 1 env, 300 episodesPyTorch

Tello/VisualLanding

Model card onlyAerial

detects an ArUco marker and lands precisely on it using visual servoing

Robots:DJI Tello
Tested on 1 robot, 2 envs, 10,040 episodesStable Baselines3

CrossEmbodiment/LocoGo2H1

Model card onlyLocomotion

walks forward using a unified policy across quadruped and humanoid morphologies

Robots:Unitree Go2Unitree H1
Tested on 2 robots, 1 env, 400,000 episodesIsaacLab

ACT/SO100StackCubes

Model card onlyManipulation

attempts to stack 2 colored cubes on a tabletop

Robots:SO-100
Tested on 1 robot, 1 env, 50 episodesLeRobot