Reinforcement learning training stack for the LimX TRON1 bipedal robot, built on Isaac Lab and using PPO to train locomotion policies. This repository extends the Isaac Lab template to support sim-to-real training for TRON1 robot variants.

• Isaac Sim + Isaac Lab, with isaaclab / isaaclab_tasks / isaaclab_rl importable
• Python 3.10
• GPU (>= 12 GB VRAM recommended for multi-env training)

# 1. Clone the repository
git clone https://github.com/limxdynamics/tron1-rl-isaaclab.git
cd tron1-rl-isaaclab

# 2. Editable install of the extension and vendored rsl_rl
pip install -e exts/bipedal_locomotion
pip install -e rsl_rl

Note: Robot model assets (USD files) for SF_TRON1A and WF_TRON1A are bundled inside exts/bipedal_locomotion/bipedal_locomotion/assets/usd/. No additional model download is required.

Task IDs are registered in exts/bipedal_locomotion/.

# Solefoot (SF)
python scripts/rsl_rl/train.py --task Isaac-Limx-SF-TRON1A-Blind-Flat-v0 --num_envs 4096 --headless

# Wheelfoot (WF)
python scripts/rsl_rl/train.py --task Isaac-Limx-WF-TRON1A-Blind-Flat-v0 --num_envs 4096 --headless

Common options:

• --checkpoint_path -- resume from a specific .pt checkpoint
• --video -- enable video recording
• --max_iterations N -- override the maximum iteration count

Log path: logs/rsl_rl/<experiment_name>/_<run_name>/

Apache 2.0.