FlmainGO Training Guide

Workflow

1

1. Go to the lab root

Run from your shell:

Change directory

cd lab/flamingo

2

2. Activate your Isaac Lab environment

Activate the conda environment (change name as needed):

Activate conda env

conda activate env_isaaclab

3

3. Install the package in editable mode

Install the repository as an editable package:

Install editable package

pip install -e .

4

4. Unzip USD assets (manual step)

Git may not correctly upload .usd files, so you must manually unzip the USD assets into the repository assets folder.

Example target path:

• lab/flamingo/assets/data/Robots/Flamingo/flamingo_rev01_4_1/

Quick check: confirm you can see .usd files under the expected directory before training.

---

Launch scripts

All commands in this section should be run from the repository root: lab/flamingo

Train Flamingo

Run:

Train command

python scripts/co_rl/train.py \
  --task {task name} \
  --algo ppo \
  --num_envs 4096 \
  --headless \
  --num_policy_stacks {stack number on policy obs} \
  --num_critic_stacks {stack number on critic obs}

Example (track velocity):

Train example — track velocity

python scripts/co_rl/train.py \
  --task Isaac-Velocity-Flat-Flamingo-v1-ppo \
  --algo ppo \
  --num_envs 4096 \
  --headless \
  --num_policy_stacks 2 \
  --num_critic_stacks 2

Play Flamingo

Run:

Play command

python scripts/co_rl/play.py \
  --task {task name} \
  --algo ppo \
  --num_envs 64 \
  --num_policy_stacks {stack number on policy obs} \
  --num_critic_stacks {stack number on critic obs} \
  --load_run {folder name} \
  --plot False

Example (track velocity):

Play example — track velocity

python scripts/co_rl/play.py \
  --task Isaac-Velocity-Flat-Flamingo-Play-v1-ppo \
  --algo ppo \
  --num_envs 64 \
  --num_policy_stacks {stack number on policy obs} \
  --num_critic_stacks {stack number on critic obs} \
  --load_run 2025-03-16_17-09-35 \
  --plot False

---

Where results are saved (logs)

Isaac Lab RL scripts typically write runs under:

• logs////

For this repository’s scripts, runs are commonly stored under:

• logs/rsl_rl///

When playing, point --load_run to the <date-time> folder name.

---

Troubleshooting

1) No module named ... or scripts import errors

• Make sure you run the commands from the correct working directory:

  • install/train/play should be run from: lab/flamingo/

• Confirm your Isaac Lab Python environment is activated.

• Re-run: pip install -e ..

2) Robot/scene fails to load or asset path errors

• Verify that .usd files exist under:

  • lab/flamingo/assets/data/Robots/Flamingo/flamingo_rev01_4_1/

• If needed, manually unzip the USD assets into the expected directory structure.

3) Isaac Sim/Isaac Lab mismatch

• Ensure your Isaac Sim version is compatible with the Isaac Lab version you installed.

• If you see startup crashes or missing extensions, re-check the Isaac Lab official install docs and quickstart.

---

References

• Isaac-RL-Two-wheel-Legged-Bot: https://github.com/jaykorea/Isaac-RL-Two-wheel-Legged-Bot

• Isaac Lab binaries install: https://isaac-sim.github.io/IsaacLab/main/source/setup/installation/binaries_installation.html

• IsaacLab (GitHub): https://github.com/isaac-sim/IsaacLab

• Isaac Lab quickstart: https://isaac-sim.github.io/IsaacLab/main/source/setup/quickstart.html

• Isaac Lab RL scripts overview: https://isaac-sim.github.io/IsaacLab/main/source/overview/reinforcement-learning/rl_existing_scripts.html

• Isaac Lab Docker guide: https://isaac-sim.github.io/IsaacLab/main/source/deployment/docker.html

• Isaac Sim docs: https://docs.isaacsim.omniverse.nvidia.com/