strands-isaaclab-open-drawer-franka

A Franka Panda grasps a cabinet drawer handle and pulls the drawer open, in NVIDIA Isaac Lab, recorded through strands-robots. The PPO policy was trained with strands' isaaclab train_policy provider (PR #4227); its rollouts, written with strands' DatasetRecorder, are this LeRobot v3 dataset.

4 recorded envs (2×2) from the strands camera — mp4. Policy: cagataydev/strands-isaaclab-open-drawer-franka-policy.

episodes / frames	12 / 3600 at 60 fps (12 × 300 frames = 5 s each)
camera	observation.images.front 320×240 RTX, fixed (eye 1.5,-1.5,1.4 → target 0.5,0,0.4)
observation.state	47-D = 9 joint pos (7 arm + 2 finger) + 31-D policy observation (policy_obs.*) + root pos (3) + root quat xyzw (4)
action	8-D raw policy action: 7 arm joint targets + 1 gripper command, 60 Hz
task string	"grasp the drawer handle and pull the drawer open"
episode return (mean)	60.0 (range 58.2 – 60.4)
checks	strands verify_dataset ok · LeRobotDataset load ok · video decode ok · NaN/inf = 0 · Hub round-trip ok

Results (the policy that generated this data)

Trained with 4096 parallel envs × 400 PPO iterations (157 M env steps) in 21 min 6 s on one NVIDIA L40S (Newton/MJWarp). Mean reward 1.63 → 97.9 (best) → 97.8 (last); drawer-open success 1.00 (≥ 0.99 from iteration 37); drawer opened ≈ 0.43 m at the end of training; throughput median 150 k env-steps/s (max 155 k, GPU shared with the AnymalD run).

PPO iteration	mean reward	drawer-open success	env-steps/s
0	1.63	0.002	88,003
25	36.14	0.940	95,673
50	50.87	1.000	95,401
100	71.89	1.000	95,659
200	90.95	1.000	153,266
300	96.91	1.000	152,153
399	97.82	1.000	125,944

How it was made with strands-robots

# Isaac Lab in its OWN venv (its pins clash with strands; strands never imports it)
uv venv --python 3.12 ~/il && uv pip install --python ~/il/bin/python --prerelease=allow \
  --index https://pypi.nvidia.com --index-strategy unsafe-best-match "isaaclab[rsl-rl,isaacsim]==3.0.0rc1"
export ISAACLAB_PYTHON=~/il/bin/python
export OMNI_KIT_ACCEPT_EULA=YES          # you accept the NVIDIA Omniverse / Isaac Sim EULA yourself
pip install "git+https://github.com/cagataycali/robots@feat/isaaclab-trainer"   # strands-robots with PR #4227

1. Train (strands train_policy, isaaclab provider)

As an agent tool call (the train_policy tool is a Strands @tool):

from strands import Agent
from strands_robots.tools.train_policy import train_policy

agent = Agent(tools=[train_policy])
agent("Train the Franka to open the cabinet drawer with the isaaclab provider: task Isaac-Open-Drawer-Franka, 400 iterations, seed 1.")
# -> train_policy(action="train", provider="isaaclab", steps=400, seed=1, output_dir="runs/c6_open_drawer",
#                 extra={"task": "Isaac-Open-Drawer-Franka", "timeout_s": 7200})

As plain Python (exactly what produced this run):

from strands_robots.tools.train_policy import train_policy

job = train_policy(action="train", provider="isaaclab", steps=400, seed=1, output_dir="runs/c6_open_drawer",
                   extra={"task": "Isaac-Open-Drawer-Franka", "timeout_s": 7200})   # task default: 4096 envs, Newton/MJWarp physics
train_policy(action="status", provider="isaaclab", job_id="<job_id from the result>")

Under the hood: python -m isaaclab train --rl_library rsl_rl --task Isaac-Open-Drawer-Franka --max_iterations 400 --seed 1 in $ISAACLAB_PYTHON. Job id of this run: isaaclab-20260929-074641-de56a64da9e9. Docs: docs/learn/training/isaaclab.md · PR: strands-labs/robots#4227.

2. Record (strands Policy + DatasetRecorder)

The final checkpoint was rolled out and recorded with examples/record_trained_policy.py (included): rebuild the task env in play mode with an RTX camera → load model_399.pt with rsl_rl and export TorchScript/ONNX → wrap the actor as a strands Policy (RslRlJitPolicy, max |Δa| vs rsl_rl = 3.6e-07) → step with policy.get_actions_sync(...) → write every frame through strands DatasetRecorder (LeRobot v3) → verify with strands verify_dataset.

OMNI_KIT_ACCEPT_EULA=YES PYTHONPATH=/path/to/strands-robots $ISAACLAB_PYTHON examples/record_trained_policy.py \
  --task Isaac-Open-Drawer-Franka --checkpoint model_399.pt --episodes 12 --frames 300 \
  --cam fixed --cam_name front --eye 1.5,-1.5,1.4 --target 0.5,0,0.4 \
  --task_str "grasp the drawer handle and pull the drawer open" --robot_type franka_panda --root out/ds --repo_id cagataydev/strands-isaaclab-open-drawer-franka
$ISAACLAB_PYTHON examples/record_trained_policy.py --verify out/ds --repo_id cagataydev/strands-isaaclab-open-drawer-franka

Use it

from lerobot.datasets.lerobot_dataset import LeRobotDataset
ds = LeRobotDataset("cagataydev/strands-isaaclab-open-drawer-franka")
print(ds.num_episodes, ds.num_frames)            # 12 3600
x = ds[0]; print(x["observation.state"].shape, x["action"].shape, x["observation.images.front"].shape)
# (47,) (8,) (3, 240, 320)

Train an imitation policy on it with strands:

from strands_robots.tools.train_policy import train_policy
train_policy(action="train", provider="lerobot", dataset_repo_id="cagataydev/strands-isaaclab-open-drawer-franka",
             output_dir="runs/act_drawer", steps=20000, batch_size=32, extra={"policy_type": "act"})

Provenance

• strands-robots: feat/isaaclab-trainer @ fa66fc68 — strands-labs/robots#4227 (isaaclab train_policy provider; DatasetRecorder; verify_dataset)
• Isaac Lab 3.0.0rc1 · Isaac Sim 6.1.0.0 · Newton / MJWarp (task default physics) · rsl-rl-lib 5.4.1 (PPO) · lerobot 0.6.1
• GPU: 1× NVIDIA L40S (46 GB), shared with the AnymalD rough-terrain training
• Seeds: training seed 1 (params/agent.yaml, params/env.yaml); recording seed 7
• Training job: isaaclab-20260929-074641-de56a64da9e9, 2026-09-29

Limitations

• Simulation only; no real Franka or cabinet was used; no sim-to-real claims.
• Release candidates: Isaac Lab 3.0.0rc1 / Isaac Sim 6.1.0.0. Trained on the task's default Newton/MJWarp physics; the checkpoint does not remember the preset (IL-X-011) — replay on the same physics.
• Metrics/success_rate is Isaac Lab's drawer-open criterion; it saturates early (≥ 0.99 from iteration 37) while the reward keeps rising (the drawer gets opened further/faster), so compare runs by reward too.
• Recording: 12 parallel envs, 300 frames (5 s @ 60 fps) each from one fixed camera; the episodes are fixed-length windows, not cut at success. observation.state = 9 joint pos + 31-D policy observation + root pos/quat (47-D), not a standard LeRobot robot state.
• create_policy("rl") in strands cannot load rsl_rl checkpoints yet (IL-X-006); use the exported TorchScript + the wrapper above.

License

Card choice: license: other — our generated data / weights under CC-BY-4.0, plus NVIDIA notices. Why:

• Trajectories, rendered camera video, playback clips and trained weights are user-generated content made with NVIDIA Isaac Sim / Isaac Lab; the NVIDIA Omniverse License Agreement (governs Isaac Sim 6.1, isaacsim/LICENSE.txt) §2.1 allows distributing "user generated content that you develop using Omniverse, such as video, audio, stills, models, 3D assets and screen captures". We release it under CC-BY-4.0.
• No NVIDIA Content is redistributed: the Franka Panda USD (IsaacLab/Robots/FrankaEmika/Legacy/panda_instanceable.usd) and the Sektion cabinet USD (Props/Sektion_Cabinet/sektion_cabinet_instanceable.usd) come from the Isaac Lab / Isaac Sim asset packs and are not in this repo (params/env.yaml only references their paths). "Franka" is a trademark of Franka Robotics; no endorsement by Franka Robotics or NVIDIA is implied.
• params/*.yaml are Isaac Lab configurations (BSD-3-Clause); examples/* are Apache-2.0 like strands-robots. Running Isaac Sim requires your own acceptance of the NVIDIA Isaac Sim / Omniverse EULA. Full text: LICENSE.md.