Skip to main content

Module 2: The Digital Twin - Gazebo & Unity Simulation

Before a robot can operate in the real world, it needs a safe, controlled environment for development, testing, and training. This is the role of a simulator—to create a digital twin of the robot and its environment. This chapter explores two of the most powerful simulation platforms in robotics: Gazebo and Unity.

1. Learning Objectives​

  • Explain the critical role of simulation in modern robotics.
  • Differentiate between Gazebo and Unity as simulation platforms.
  • Understand the key features and architecture of both simulators.
  • Create a simple world in Gazebo and spawn a robot model.
  • Understand the basics of connecting ROS 2 with Gazebo and Unity.
  • Identify the pros and cons of each simulator for various robotics tasks.

2. The Importance of Simulation in Robotics​

Why not just test on real hardware? While physical testing is the ultimate goal, simulation offers indispensable advantages:

  • Safety: Testing new algorithms on a physical robot can be dangerous, especially with early-stage, unpredictable code. Simulators provide a sandbox where failures have no real-world consequences.
  • Cost-Effectiveness: Simulators are free. Physical robots and sensors can cost thousands or millions of dollars. Crashing a simulated robot costs nothing.
  • Speed: You can run simulations faster than real-time to quickly gather data or test long-duration tasks.
  • Parallelization: You can run thousands of simulations in parallel on a cloud server to train AI models (Reinforcement Learning) on a massive scale.
  • Perfect Perception: Simulators can provide "ground truth" data that is impossible to get from real sensors, such as the exact position of every object in the environment.

3. Introduction to Gazebo​

Gazebo is a free, open-source, and robust physics simulator for robotics. It is one of the most widely used simulators in the ROS community, designed specifically for robotics research, development, and testing.

Gazebo Architecture​

A Gazebo simulation is composed of several key components:

  • World: The entire simulated environment, including objects, lighting, and physics properties. Described in a .sdf file.
  • Models: Any object in the world, from a simple shape to a complex robot. Models are also described using SDF.
  • Sensors: Special types of models that generate data, like cameras, LiDAR, and IMUs.
  • Plugins: Custom code that can be attached to models or sensors to control their behavior or generate data.
graph TD
    subgraph Gazebo World (.sdf)
        A[Physics Engine]
        B[Lighting]
        C[Environment]
    end

    subgraph Robot Model (SDF)
        D[Links<br/>(Rigid Bodies)]
        E[Joints<br/>(Connect Links)]
        F[Sensors<br/>(Camera, LiDAR)]
        G[Plugins<br/>(To control joints or publish sensor data)]
    end

    A -- Affects --> D
    F -- Generates Data --> G
    C -- Interacts with --> D

Figure 1: High-level architecture of a Gazebo simulation.

The Simulation Description Format (SDF)​

Gazebo uses SDF to define everything from worlds to robots. It's an XML-based format that describes the physical properties (mass, inertia), visual properties (color, shape), and collision properties of objects.

Hands-On: A Simple Gazebo World​

  1. Create a World File: Create a file named simple.world.

    <?xml version="1.0" ?>
    <sdf version="1.7">
      <world name="default">
        <include>
          <uri>model://sun</uri>
        </include>
        <include>
          <uri>model://ground_plane</uri>
        </include>
        <model name="box">
          <pose>0 0 0.5 0 0 0</pose>
          <link name="link">
            <collision name="collision">
              <geometry><box><size>1 1 1</size></box></geometry>
            </collision>
            <visual name="visual">
              <geometry><box><size>1 1 1</size></box></geometry>
            </visual>
          </link>
        </model>
      </world>
    </sdf>

  2. Run Gazebo: Open a terminal and run Gazebo with your world file.

    gazebo simple.world

You should see a simple world with a ground plane, lighting, and a floating cube.

4. Introduction to Unity for Robotics​

Unity is a professional game engine that has been increasingly adopted by the robotics community for its high-fidelity graphics, powerful physics engines, and ease of use.

Key Unity Features for Robotics​

  • High-Fidelity Graphics: Unity is renowned for its photorealistic rendering capabilities, making it ideal for training computer vision models that need to work in the real world.
  • Unity Robotics Hub: A set of open-source packages that makes it easy to connect Unity simulations with ROS 2.
  • Physics Engines: Unity supports multiple physics engines, including NVIDIA's PhysX and the more specialized Havok physics engine.
  • Asset Store: A massive marketplace of 3D models, environments, and tools that can be quickly integrated into your simulation.

Hands-On: A Simple Unity Scene​

Setting up a Unity scene involves using the graphical editor:

  1. Install Unity Hub and a recent version of the Unity Editor.
  2. Create a new 3D project.
  3. Add a Plane to act as the ground (GameObject -> 3D Object -> Plane).
  4. Add a Cube to act as an object (GameObject -> 3D Object -> Cube).
  5. Add a Rigidbody component to the Cube to make it affected by physics.
  6. Press the "Play" button to start the simulation and see the cube fall onto the plane.

5. Connecting Simulators with ROS 2​

The real power of simulation is realized when you connect it to your robot's brain—ROS 2.

a. Gazebo & ROS 2 (ros_gz)​

The ros_gz packages provide the bridge between Gazebo and ROS 2.

  • Gazebo Plugins publish sensor data (like camera images or laser scans) directly to ROS 2 topics.
  • ROS 2 nodes can control robots in Gazebo by publishing messages to topics that Gazebo plugins are subscribed to (e.g., sending velocity commands to a wheel joint).
graph TD
    subgraph Gazebo Simulator
        A[Camera Sensor Plugin] -- Publishes to --> B(Gazebo Transport);
    end
    subgraph ROS 2
        D[ROS 2 Node<br/>(e.g., /image_processor)]
    end
    subgraph Bridge
      C[`ros_gz_bridge`]
    end

    B -- Gazebo Message --> C;
    C -- ROS 2 Message --> E[ROS 2 Topic<br/>/image_raw];
    E -- Delivered to --> D;

Figure 2: The ros_gz_bridge connects Gazebo's transport layer to the ROS 2 graph.

b. Unity & ROS 2 (Robotics Hub)​

The Unity Robotics Hub provides a package called ROS-TCP-Connector, which allows Unity to communicate with ROS 2 over a TCP connection.

  • A C# script in Unity can publish sensor data as ROS 2 messages.
  • Another C# script can subscribe to ROS 2 topics and use the data to control objects in the Unity scene.

6. Choosing the Right Simulator​

FeatureGazeboUnity
Primary Use CasePhysics-heavy simulation, robotics researchHigh-fidelity rendering, CV/ML model training
Physics FidelityGood, multiple engines (ODE, DART)Very good (PhysX, Havok)
Graphics QualityBasic to moderatePhotorealistic
ROS 2 IntegrationExcellent, native (ros_gz)Very good (Robotics Hub)
CommunityStrong in the ROS/robotics communityMassive in the gaming/real-time 3D community
CostFree and Open SourceFree for personal use, paid for enterprises

Rule of thumb:

  • If your focus is on dynamics, control systems, and traditional robotics algorithms, Gazebo is often the best choice.
  • If your focus is on computer vision, training AI models in photorealistic worlds, or creating public-facing demos, Unity is a superior option.

7. Summary & Key Takeaways​

  • Simulation is an essential tool for safe, fast, and cost-effective robotics development.
  • Gazebo is a powerful, open-source physics simulator with deep roots in the ROS community.
  • Unity is a professional game engine with stunning graphics, making it ideal for vision-based AI training.
  • Both simulators can be tightly integrated with ROS 2 to create a complete software-in-the-loop (SIL) system.
  • The choice of simulator depends on your specific project goals.

8. Chapter Quiz​

  1. What is the primary benefit of using simulation for training a Reinforcement Learning agent? a) It looks better than the real world. b) It allows for massive parallelization and faster-than-real-time execution. c) It is more expensive. d) It is required by ROS 2.

  2. What format does Gazebo use to describe worlds and models? a) XML b) JSON c) URDF d) SDF

  3. What is the main advantage of using Unity for robotics simulation? a) It is fully open source. b) It has the most accurate physics engine by default. c) It offers high-fidelity, photorealistic graphics. d) It has native integration with ROS 1 only.

  4. The ros_gz packages are used to: a) Bridge communication between ROS 2 and Gazebo. b) Run Unity simulations from the command line. c) Define robot models. d) Replace the ROS 2 build system.

  5. You are developing a self-driving car algorithm that relies heavily on camera data and needs to be trained in many different realistic lighting and weather conditions. Which simulator would be the better choice? a) Gazebo, because it has better physics. b) Unity, because of its high-fidelity graphics. c) Either would be equally good. d) Neither, you must use real-world data.

(Answers: 1-b, 2-d, 3-c, 4-a, 5-b)

9. Further Study​