Skip to main content

Quick Start

This guide will help you run your first DISCOVERSE simulation example in just a few minutes and experience the core features of the framework.

Verify Installation​

First, make sure DISCOVERSE is installed correctly:

python scripts/check_installation.py

If you see all core checks passed, the installation is successful.

First Robot Simulation​

Launch a Basic Robot Environment​

DISCOVERSE supports multiple robot platforms. Let's start with the simplest one:

# Launch Airbot Play robotic arm
python discoverse/robots_env/airbot_play_base.py

This will start a basic Airbot Play robotic arm simulation environment. You should see a 3D simulation window displaying the robot arm model.

Run an Operation Task​

Now let's run an automated operation task:

# Run the coffee cup placement task
python discoverse/examples/tasks_airbot_play/place_coffeecup.py

You will see the robotic arm automatically execute the pick and place actions for the coffee cup. This demonstrates DISCOVERSE's automated data generation capabilities.

Interactive Control​

While the simulation is running, you can use the following keyboard shortcuts for interaction:

Basic Controls​

  • 'h' - Show help menu
  • 'r' - Reset simulation state
  • 'F5' - Reload MJCF scene
  • 'p' - Print robot status information

View Controls​

  • '[' / ']' - Switch camera views
  • 'Esc' - Toggle free camera mode
  • Left mouse drag - Rotate view
  • Right mouse drag - Pan view
  • Mouse wheel - Zoom view

Rendering Toggle​

  • 'Ctrl+g' - Toggle Gaussian rendering (requires gaussian-rendering module)
  • 'Ctrl+d' - Toggle depth visualization

More Robot Platforms​

Dual-Arm Mobile Robot (MMK2)​

# Launch MMK2 dual-arm robot
python discoverse/robots_env/mmk2_base.py

# Run kiwi picking task
python discoverse/examples/tasks_mmk2/kiwi_pick.py

Dexterous Hand Simulation​

# Launch LeapHand tactile hand
python discoverse/examples/robots/leap_hand_env.py

Inverse Kinematics Demo​

DISCOVERSE provides interactive inverse kinematics functionality:

# Airbot Play inverse kinematics
python discoverse/examples/mocap_ik/mocap_ik_airbot_play.py

# Specify specific scene
python discoverse/examples/mocap_ik/mocap_ik_airbot_play.py --mjcf mjcf/tasks_airbot_play/stack_block.xml

# MMK2 inverse kinematics
python discoverse/examples/mocap_ik/mocap_ik_mmk2.py --mjcf mjcf/tasks_mmk2/pan_pick.xml

More Application Examples​

SLAM Tasks​

If you have installed the 3DGS rendering module, you can experience high-fidelity active SLAM:

Prerequisites

Requires installation of the gaussian-rendering module and downloading corresponding .ply model files. Please refer to the Installation Guide.

python discoverse/examples/active_slam/dummy_robot.py

This example showcases the simulator's high-fidelity 3D environment.

Multi-Agent Collaboration​

python discoverse/examples/skyrover_on_rm2car/skyrover_and_rm2car.py

Watch a demonstration of drone and ground robot collaboration.

Data Collection Workflow​

One of DISCOVERSE's core advantages is automated data collection with 100x efficiency improvement over real-world collection:

1. Run Data Collection Tasks​

# Automatically generate robotic arm operation data
python discoverse/examples/tasks_airbot_play/block_bridge_place.py

# Generate dual-arm collaboration data  
python discoverse/examples/tasks_mmk2/coffeecup_plate.py

2. View Generated Data​

Data is typically saved in the data/ directory, including:

  • Robot state trajectories
  • Sensor data (RGB, depth, point clouds, etc.)
  • Action sequences
  • Task labels

Next Steps​

Now that you've successfully run your first simulation, you can explore further:

  1. Basic Concepts - Understand DISCOVERSE's architecture and core concepts
  2. Basic Simulation Tutorial - Learn how to create custom simulation scenes
  3. Sensor Tutorial - Configure and use various sensors
  4. Imitation Learning Tutorial - Train your first robot policy

Getting Help​

If you encounter issues: