Legged Locomotion RL — Sim-to-Real

    1

    Train and deploy reinforcement learning locomotion policies for quadrupeds and humanoids with sim-to-real validation.

    $49

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Legged Locomotion RL — Sim-to-Real

    Example session with this skill installed

    Generate a PPO reward structure for a quadruped robot dog using Isaac Lab. The goal is a trot gait at 0.5 m/s while minimizing energy consumption and foot slip on flat terrain.

    • Read your context and instructions
    • Compiled the legged locomotion rl

    I have generated the reward structure and the corresponding environment observation space for the quadruped trot task. The rewards use exponential kernels for tracking and squared penalties for hardware longevity. ## Quadruped Trot Reward Configuration | Reward Term | Weight | Formula / Description | Goal | | :--- | :--- | :--- | :--- | | **Lin Vel Tracking** | 1.0 | `exp(-|v_xy - cmd_v_xy|^2 / 0.25)` | Match commanded 0.5 m/s velocity | | **Ang Vel Tracking** | 0.5 | `exp(-|w_z - cmd_w_z|^2 / 0.25)` | Maintain heading stability | | **Torque Penalty** | -0.0001 | `sum(torques^2)` | Reduce motor heat and battery drain | | **Foot Slip** | -0.1 | `sum(foot_vel_xy^2 * contact_force)` | Prevent sliding during stance phase | | **Action Rate** | -0.01 | `sum((action_t - action_t-1)^2)` | Ensure smooth joint movements | | **Air Time** | 0.2 | `sum(time_in_air > threshold)` | Encourage distinct trotting steps | ```python # Environment Observation Baseline # Total Dimension: 48 (Proprioceptive) obs_spec = { "base_ang_vel": 3, # Angular velocity in base frame "projected_gravity": 3, # Gravity vector in base frame "commands": 3, # [vx, vy, yaw_rate] "joint_pos": 12, # Relative to default pose "joint_vel": 12, # Current joint velocities "last_actions": 12 # Previous step actions for smoothness } # Domain Randomization Ranges randomization = { "friction": [0.5, 1.25], "added_mass": [-1.0, 1.0], # kg "pd_gains": [0.85, 1.15], # Scaling factor "latency": [0.0, 0.02] # up to 20ms } ``` **Next steps** 1. Integrate these kernels into your `LeggedRobotCfg` reward class. 2. Run `python scripts/train.py` and monitor the `episode_sum_lin_vel` in Tensorboard. 3. Validate the export with `deploy_config_check.py` before moving to the harness.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Reward recipes, randomization, actuator models, deployment checker. Train and deploy reinforcement-learning locomotion policies for legged robots (quadruped robot dogs and humanoids) with sim-to-real transfer - Isaac Lab / MuJoCo (MJX) environments, reward design, domain randomization, actuator modelling, PPO training, policy export and a validated deployment config (control rate, PD gains, torque and joint limits). Use when the user wants a robot dog or humanoid to walk, run, climb stairs or recover from falls with a learned controller, or when a policy that works in simulation fails on the real robot. Includes an EXPERT.md reference (exact parameters, thresholds, failure diagnosis). Tested scripts: scripts/deploy_config_check.py. Acceptable use: civil and commercial applications only. Any military, paramilitary or weapon-related use is prohibited, as stated in the skill itself.

    Requires a harness for initial hardware tests. Does not support military applications. Policies are limited to their training distribution and require a safety standing-down mode.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 28 days ago

    Permissions

    Terminal / Shell
    Read Files
    Listed28 days ago

    What's inside

    Frequently Asked Questions

    The skill provides everything needed to build a training pipeline: validated reward kernels, environment configuration scripts, domain randomization templates, and export utilities for TorchScript or ONNX.

    It is optimized for Isaac Lab (formerly Isaac Gym) and MuJoCo/MJX, supporting both GPU-parallelized training for speed and JAX-based simulations for high-accuracy modeling.

    Yes, the agent includes automated URDF/MJCF validation to ensure mass, inertia, and joint limits are correctly defined before training begins to prevent simulation errors.

    This skill uses specialized reward kernels and domain randomization ranges—such as latency modeling and friction scaling—specifically designed to prevent 'sim-to-real' artifacts like motor overheating or gait instability.

    While the skill automates reward and environment setup, you will need a workstation with an NVIDIA GPU for Isaac Lab or a JAX-compatible environment for MJX to execute the training.

    New in Agents & Orchestration