🚀 Introduction

What is TeraSim-NDE-NADE?

TeraSim-NDE-NADE is the traffic environment for TeraSim, which is an advanced generative simulation framework designed for testing autonomous vehicles (AVs) in naturalistic and adversarial environments. Unlike traditional scenario-based approaches, our method generates interactive, real-world-like traffic environments to accelerate AV safety testing by 10³ - 10⁵ times.

Key Features

• Naturalistic Driving Environment (NDE):
  • Trained and calibrated using large-scale real-world driving data
  • Fully interactive and dynamic
• Adversity Injection:
  • Vehicle Adversity (e.g., aggressive lane changes, hard braking)
  • Vulnerable Road User (VRU) Adversity (e.g., jaywalking, running red lights)
  • Static Adversity (e.g., construction zones, faulty traffic signals)
  • Includes trigger conditions, effects, and probability modeling
• Naturalistic and Adversarial Driving Environment (NADE):
  • AI-powered dynamic control of adversity injection
  • Optimized based on real-world accident data
  • Ensures large-scale, accelerated safety evaluation

🏗 Code Architecture

System Architecture

🔧 Installation & Configuration

1. Install Dependencies

Ensure you have TeraSim installed:

git clone https://github.com/mcity/TeraSim
cd TeraSim
pip install -e .
cd ..

2. Install Poetry (if not already installed)

curl -sSL https://install.python-poetry.org | python3 -

3. Install TeraSim-NDE-NADE

git clone https://github.com/mcity/TeraSim-NDE-NADE
cd TeraSim-NDE-ITE
poetry install

📌 Usage Example

Please refer to the /examples directory for detailed usage examples and tutorials.

The examples demonstrate:

• Setting up a naturalistic driving environment
• Configuring adversarial scenarios
• Running simulations with different parameters
• Analyzing simulation results

🤝 Contributing

We welcome contributions to improve TeraSim-NDE-NADE! To contribute:

1. Fork the repository
2. Create a feature branch (feature/new-adversity-model)
3. Submit a Pull Request (PR)
4. Join the discussion and help advance AV safety testing!

🔄 System Workflow

Below is a detailed workflow diagram showing the key processes in TeraSim-NDE-NADE:

  graph TD
    subgraph NADE Main Process
      start[Initialize: on_start] --> step[Main Loop: on_step]
      step --> stop[Terminate: on_stop]
    end
    subgraph on_step Core Process
      step --> getObs[Get Observation Data]
      getObs --> makeDecisions[Make Environment Decisions]
      makeDecisions --> executeMove[Execute Movement]
      executeMove --> NADEDecision[NADE Decision Making]
      NADEDecision --> applyCommands[Execute Control Commands]
    end
    subgraph NADE Decision Core Process
      NADEDecision --> predictTrajectory[Predict Future Trajectories]
      predictTrajectory --> getManeuverChallenge[Get Maneuver Challenges]
      getManeuverChallenge --> getAvoidability[Calculate Avoidability]
      getAvoidability --> addAvoidCommands[Add Avoidance Commands]
      addAvoidCommands --> modifyNDD[Modify NDD Distribution]
      modifyNDD --> getCriticality[Get Criticality Values]
      getCriticality --> NADEImportanceSampling[NADE Importance Sampling]
      NADEImportanceSampling --> applyCollisionAvoidance[Apply Collision Avoidance]
    end
    subgraph Maneuver Challenge Assessment
      getManeuverChallenge --> getNormalTrajectories[Get Normal Trajectories]
      getNormalTrajectories --> getNegligenceTrajectories[Get Negligence Trajectories]
      getNegligenceTrajectories --> checkCollisions[Check Collisions]
      checkCollisions --> updateContext[Update Context]
    end
    subgraph Importance Sampling Process
      NADEImportanceSampling --> getISProb[Calculate IS Probability]
      getISProb --> calculateWeight[Calculate Weights]
      calculateWeight --> selectCommand[Select Control Command]
    end
    subgraph Collision Avoidance Process
      applyCollisionAvoidance --> getNegligencePairs[Get Negligence Pairs]
      getNegligencePairs --> checkAvoidability[Check Avoidability]
      checkAvoidability --> selectAvoidanceCommand[Select Avoidance Command]
      selectAvoidanceCommand --> updateWeight[Update Weights]
    end
    subgraph Data Recording and Updates
      recordStep[Record Step Data]
      recordNegligence[Record Negligence Information]
      updateDistance[Update Distance]
      calculateTotalDistance[Calculate Total Distance]
      alignRecordEvent[Align Record Events]
    end
    %% Define key data flows
    class NADEDecision,predictTrajectory,getManeuverChallenge,NADEImportanceSampling,applyCollisionAvoidance mainFlow;
    class getISProb,calculateWeight,selectCommand samplingFlow;
    class getNegligencePairs,checkAvoidability,selectAvoidanceCommand avoidanceFlow;
    class recordStep,recordNegligence,updateDistance dataFlow;
    %% Style definitions
    classDef mainFlow fill:#f9f,stroke:#333,stroke-width:2px;
    classDef samplingFlow fill:#bbf,stroke:#333,stroke-width:2px;
    classDef avoidanceFlow fill:#bfb,stroke:#333,stroke-width:2px;
    classDef dataFlow fill:#fbb,stroke:#333,stroke-width:2px;

Loading

Join us in making autonomous vehicles safer with realistic, generative simulation! 🚗💡