Reinforcement learning & simulation

Reinforcement learning (soarm_mjlab, in progress) trains in MuJoCo via mjlab, following the manager-based config pattern from unitree_rl_mjlab — no C++ deployment stack or terrain/locomotion machinery, since a 6-DOF arm's control loop already runs comfortably in Python at ~50 Hz. Deploys through soarm_sdk.RobotInterface so trained policies drive sim and real hardware identically. See RL training pipeline, step by step for the full algorithmic walkthrough.

Beyond this workspace's own tracks, the broader SO-ARM ecosystem has useful options worth knowing about if you want to go further in a given direction: MoveIt/OMPL for classical motion planning, ros2_so_arm100 for a ROS2 description/control bring-up, IsaacSim/IsaacLab for GPU-accelerated, photorealistic RL at larger scale.

Status: Phases 0–3 done, Phase 4 (real training run) in progress — the first campaign didn't clear the promotion bar yet. See Roadmap for the phase-by-phase detail.

• Environment & Task Config — MDP terms, rewards/observations, adding a new task
• Training Runs — local CPU smoke tests, vast.ai GPU campaigns, the v1–v11 results
• Roadmap — phase-by-phase status