Main Content

How Unreal Engine Simulation for Automated Driving Works

Automated Driving Toolbox™ provides a co-simulation framework that you can use to model driving algorithms in Simulink® and visualize their performance in a virtual simulation environment. This environment uses the Unreal Engine® by Epic Games®. This co-simulation framework allows Automated Driving Toolbox to send data between the Unreal Engine and Simulink.

Understanding how this simulation environment works can help you troubleshoot issues and customize your models.

Note

Simulating models in the 3D visualization environment requires Simulink 3D Animation™.

Communication with 3D Simulation Environment

When you use Automated Driving Toolbox to run your algorithms, Simulink co-simulates the algorithms in the visualization engine through a lock-step mechanism. The lock-step mechanism is a synchronization approach where the simulation progresses in fixed time steps, and the two simulation engines, Simulink and the 3D simulation engine, run sequentially. One simulation engine waits while the other simulation engine proceeds, so each simulation produces consistent results.

During each simulation step:

  1. Simulink configures the initial settings of the 3D environment, including actor transformations and properties and the custom scene viewpoint. Simulink also determines the configuration parameters of the 3D environment and the actors using the 3D simulation engine feedback.

  2. The co-simulation framework sends data from Simulink to the 3D simulation engine.

  3. The co-simulation framework sends a command signal to the 3D simulation engine indicating the end of data transfer.

  4. The 3D simulation engine executes the simulation. The 3D simulation engine configures the 3D environment and the actor positions based on the data from Simulink. The 3D simulation engine then provides updated information about the 3D environment and the actors.

  5. The co-simulation framework sends data from the 3D simulation engine to Simulink.

  6. The co-simulation framework sends an acknowledge signal to MATLAB® or Simulink, indicating the end of data transfer from the 3D simulation engine.

The diagram summarizes the communication between Simulink and the visualization engine.

Block Execution Order

During simulation, the Unreal Engine simulation blocks follow a specific execution order:

  1. The Simulation 3D Vehicle with Ground Following blocks initialize the vehicles and send their X, Y, and Yaw signal data to the Simulation 3D Scene Configuration block.

  2. The Simulation 3D Scene Configuration block receives the vehicle data and sends it to the sensor blocks.

  3. The sensor blocks receive the vehicle data and use it to accurately locate and visualize the vehicles.

The Priority property of the blocks controls this execution order. To access this property for any block, right-click the block, select Properties, and click the General tab. By default, Simulation 3D Vehicle with Ground Following blocks have a priority of -1, Simulation 3D Scene Configuration blocks have a priority of 0, and sensor blocks have a priority of 1.

The diagram shows this execution order.

Execution order for Unreal Engine simulation blocks

If your sensors are not detecting vehicles in the scene, it is possible that the Unreal Engine simulation blocks are executing out of order. Try updating the execution order and simulating again. For more details on execution order, see Control and Display Execution Order (Simulink).

Also be sure that all 3D simulation blocks are located in the same subsystem. Even if the blocks have the correct Priority settings, if they are located in different subsystems, they still might execute out of order.

Simulation Speed

Simulation speed depends on the stop time and the sample time of the simulation. The stop time is the duration of the simulation. The sample time is the time interval at which the output is updated during the simulation. To set the sample time in Simulink, use the Sample time parameter in the Simulation 3D Scene Configuration block. If you decrease the sample time, the software takes more samples per second, so simulation speed decreases.

Simulation speed also depends on model complexity, computational speed, and the speed at which the system can render a scene, which depends on the system configuration. You can use the profile function in MATLAB or the Simulink Profiler (Simulink) in Simulink to analyze the distribution of simulation execution time among the simulation components and see the computational time of each component in the simulation. For more information on profiling, see Profile Your Code to Improve Performance.

Related Topics