Mazda’s MILS Environment Supports Autonomous Driving - MATLAB & Simulink

Mazda Builds Design Concept MILS Environment for Autonomous Driving and Explores Merging Decision Algorithm

Environment Can Verify Technical Ideas in Real-World Scenarios

“By utilizing Simulink, we clarified component roles, and it was easy to reconfigure each component. The sensor models—such as radar and cameras—along with the vehicle models provided by Automated Driving Toolbox facilitated the smooth building of the environment.”

Yu Yamauchi and Daichi Ishizaki, Mazda

Key Outcomes

  • The Design Concept MILS environment built with Simulink allowed automatic extraction of roads at merging scenes using ZENRIN® map data
  • Realistic scenarios were utilized for cosimulation between PTV Vissim and Simulink in the developed environment
  • Deep Learning Toolbox was used to train a new AI model for merging decisions and implementation in Simulink
Current Time 0:00
Duration 30:13
Loaded: 0.55%
Stream Type LIVE
Remaining Time 30:13
 
1x
  • Chapters
  • descriptions off, selected
  • captions off, selected
  • en (Main), selected
    View slides (in Japanese)

    Mazda’s goal is to increase value for customers by expanding the operational design domain (ODD) for autonomous vehicles. However, as the ODD expands, the software used requires an increasing amount of code. To speed up development under such circumstances, Mazda needed to build a simulation environment to verify new technological concepts and algorithms at an early stage of development. Mazda’s “Design Concept MILS” environment can verify technical ideas in real-world scenarios. The Mazda team confirmed its usefulness by simulating a new merging decision algorithm based on deep learning techniques.

    The simulation environment enables the automatic extraction of merging scenes—for example, from on- and off-ramps for highways—using external map data. It also enables cosimulation between Simulink® and a traffic simulator, so that users can run simulations in the application created by App Designer. Using Automated Driving Toolbox™ and Simulink, the Design Concept MILS environment allows for easy reconfiguration of sensor model combinations, the vehicle model, and each algorithm by replacing Simulink blocks. Mazda also utilized Deep Learning Toolbox™ to seamlessly develop the merging decision algorithm in this environment, as well as confirm its effectiveness.

    Learn more about Mazda Motor Corporation

    Products Used

    Get a free trial
    Contact sales