<ul data-eligibleForWebStory="true"><li>This paper introduces the use of Gaussian Mixture Copula Models for statistical modeling of driving scenarios for safety validation of automated driving systems.</li><li>Understanding the joint probability distribution of scenario parameters is crucial for scenario-based safety assessment and risk quantification.</li><li>Gaussian Mixture Copula Models combine the multimodal expressivity of Gaussian Mixture Models with the flexibility of copulas, allowing separate modeling of marginal distributions and dependencies.</li><li>The study compares Gaussian Mixture Copula Models with Gaussian Mixture Models and Gaussian Copula Models using real-world driving data based on United Nations Regulation No. 157 scenarios.</li><li>Evaluation across 18 million scenario instances shows that Gaussian Mixture Copula Models offer a better fit in terms of both likelihood and Sinkhorn distance compared to the other approaches.</li><li>The findings indicate that Gaussian Mixture Copula Models can serve as a strong foundation for future scenario-based validation frameworks.</li></ul>

Estimating the Joint Probability of Scenario Parameters with Gaussian Mixture Copula Models

Discover more