[GitHub Code for this Chapter: chapter4 branch](https://github.com/Lab-of-AI-and-Robotics/IsaacLab-Tutorial/tree/chapter4)

Objective

To transform our basic, template-generated Cartpole environment into a sophisticated locomotion environment for the Unitree Go2. In this chapter, we will learn a critical skill for any advanced Isaac Lab developer: how to read, understand, and adapt the official, high-performance environment configurations provided within Isaac Lab's source code to build our own custom tasks.

4.1. The Professional Workflow: Learning from the Source

So far, we have built a project from a template. Now, to create a high-performance locomotion agent, we will not start from scratch. Instead, we will learn from the experts: the Isaac Lab developers themselves.

We will use the official Go2 configuration file, rough_env_cfg.py, as our "golden reference." Our goal is to systematically modify our test_env_cfg.py to mirror the structure and logic of this professional-grade configuration. This process of refactoring is fundamental to building complex environments efficiently.

4.2. Step 1: Foundational Inheritance

The first step is to change the foundation of our environment. The default template inherits from a generic ManagerBasedRLEnvCfg. For locomotion, we need a more specialized base class.

1. Open your project's test_env_cfg.py file.
2. Find the main class definition: class TestEnvCfg(ManagerBasedRLEnvCfg):
3. Change it to inherit from LocomotionVelocityRoughEnvCfg, which is the same base class used by the official Go2 config. You will also need to add the corresponding import statement.

# .../test/test/tasks/manager_based/test/test_env_cfg.py

# Add this import at the top
from isaaclab_tasks.manager_based.locomotion.velocity.velocity_env_cfg import LocomotionVelocityRoughEnvCfg

# Modify the class definition
@configclass
class TestEnvCfg(LocomotionVelocityRoughEnvCfg):
    # ... (rest of the class)

This base class provides pre-configured managers and settings specifically designed for velocity-controlled locomotion tasks, saving us a tremendous amount of setup.

4.3. Step 2: Full Environment Overhaul

Now, we will perform the main refactoring. We will replace all the Cartpole-specific configurations inside TestEnvCfg with locomotion-specific ones, using rough_env_cfg.py as our guide.

For this step, delete the existing SceneCfg, ActionsCfg, ObservationsCfg, EventsCfg, RewardsCfg, and TerminationsCfg classes from your test_env_cfg.py. We will replace them within the __post_init__ method of our main TestEnvCfg class, just like the official example.

Modify your TestEnvCfg class to look like this: