Lab 01-01: Setting Up Your MuJoCo Environment

Objectives

• Install MuJoCo and verify the installation
• Load and visualize a simple robot model
• Understand the MuJoCo simulation loop
• Run your first physics simulation

Materials

Type	Name	Version	Tier
software	Python	3.10+	required
software	MuJoCo	3.0+	required
software	NumPy	1.24+	required
data	humanoid.xml	included	required

Prerequisites

• Basic Python programming knowledge
• Familiarity with command line operations

Instructions

Step 1: Install MuJoCo

MuJoCo is now free and open-source. Install it using pip:

# In your terminal or Jupyter notebook
!pip install mujoco

Verification: Run the following to verify installation:

import mujoco
print(f"MuJoCo version: {mujoco.__version__}")

Expected: Version 3.0.0 or higher displayed without errors.

Step 2: Load Your First Model

MuJoCo uses XML files to define robot models. Let's load the built-in humanoid:

import mujoco
import numpy as np

# Load the humanoid model (built into MuJoCo)
model = mujoco.MjModel.from_xml_path(
    mujoco.util.get_resource_path("humanoid/humanoid.xml")
)
data = mujoco.MjData(model)

# Examine model properties
print(f"Number of bodies: {model.nbody}")
print(f"Number of joints: {model.njnt}")
print(f"Number of actuators: {model.nu}")
print(f"Degrees of freedom: {model.nv}")

Verification: You should see output similar to:

Number of bodies: 14
Number of joints: 18
Number of actuators: 21
Degrees of freedom: 27

Troubleshooting: If you get a file not found error, ensure MuJoCo is properly installed.

Step 3: Understand the Simulation Loop

MuJoCo simulations advance through discrete timesteps:

# Get simulation parameters
print(f"Timestep: {model.opt.timestep} seconds")
print(f"Simulation will run at {1/model.opt.timestep:.0f} Hz")

# Reset simulation to initial state
mujoco.mj_resetData(model, data)

# Run simulation for 1 second
simulation_time = 1.0
steps = int(simulation_time / model.opt.timestep)

for i in range(steps):
    # Step the simulation forward
    mujoco.mj_step(model, data)

```python
```python
    if i % 100 == 0:

        print(f"Time: {data.time:.3f}s, Height: {data.qpos[2]:.3f}m")

Verification: The humanoid should fall under gravity. Height should decrease from ~1.3m toward 0. Expected: You see the height decreasing as simulation time increases.

Step 4: Visualize the Simulation

Launch the interactive viewer to see your robot:

import mujoco.viewer

# Launch passive viewer (non-blocking)
with mujoco.viewer.launch_passive(model, data) as viewer:
    # Run simulation with visualization
    while viewer.is_running():
        mujoco.mj_step(model, data)
        viewer.sync()

Verification: A window should open showing the humanoid model. Expected: You can rotate the view with mouse, see the robot fall, and close the window.

Step 5: Apply Control Inputs

Let's prevent the humanoid from falling by applying joint torques:

# Reset simulation
mujoco.mj_resetData(model, data)

# Simple standing controller - apply torques to maintain position
def simple_controller(model, data):
    """Apply proportional control to maintain initial pose."""
    kp = 100  # Proportional gain

    # Target is initial joint positions (all zeros for this model)
    target = np.zeros(model.nu)

    # Current joint positions (skip free joint - first 7 elements)
    current = data.qpos[7:7+model.nu]

    # Apply proportional control
    data.ctrl[:] = kp * (target - current)

# Run with controller
with mujoco.viewer.launch_passive(model, data) as viewer:
```python
```python
    while viewer.is_running() and data.time < 5.0:

        simple_controller(model, data)
        mujoco.mj_step(model, data)
        viewer.sync()

Verification: The humanoid should attempt to maintain its pose rather than immediately collapsing. Expected: While not perfect, the robot should show some resistance to falling.

Expected Outcomes

By completing this lab, you should:

1. Have a working MuJoCo installation
2. Understand how to load and inspect robot models
3. Know the basic simulation loop structure
4. Be able to visualize simulations
5. Understand how control inputs affect robot behavior

Rubric

Criterion	Points	Excellent	Satisfactory	Needs Improvement
Installation	20	MuJoCo installed, version verified	Installed with minor issues	Unable to install
Model Loading	20	All properties correctly displayed	Model loads with warnings	Cannot load model
Simulation Loop	25	Loop runs correctly, understands timestep	Loop runs but unclear on timing	Loop fails
Visualization	20	Viewer works, can interact with scene	Viewer opens but limited interaction	Viewer fails
Controller	15	Controller implemented and explained	Controller works but not understood	Controller fails

Total Points: 100

Extensions

For additional challenge:

1. Modify the humanoid: Edit the XML to change mass or joint limits
2. Data logging: Record joint positions over time and plot them
3. Different models: Load other MuJoCo models (ant, cheetah, etc.)

Common Errors

Error	Cause	Fix
ModuleNotFoundError: mujoco	MuJoCo not installed	Run pip install mujoco
GLFWError	Display issues	Set MUJOCO_GL=egl environment variable
FileNotFoundError	Model path wrong	Use mujoco.util.get_resource_path()