Scenario Parameter Generation Method and Scenario Representativeness Metric for Scenario-Based Assessment of Automated Vehicles

TL;DR

This work tackles the challenge of generating realistic, diverse AV test scenarios without assuming specific parametric forms. It introduces a data-driven pipeline that parameterizes high-dimensional scenarios, reduces dimensionality via Singular Value Decomposition, estimates the reduced-parameter pdf with Kernel Density Estimation, and samples new scenarios from the learned distribution. Central to the contribution is the Scenario Representativeness (SR) metric, based on the Wasserstein distance, which quantifies how well generated scenarios cover the real-world variety while avoiding overfitting. Through case studies on leading-vehicle deceleration and cut-in scenarios, the method demonstrates improved representativeness over fixed-parameter or Gaussian-assumption baselines and provides a practical path to integrate with importance-sampling–based AV assessments.

Abstract

The development of assessment methods for the performance of Automated Vehicles (AVs) is essential to enable the deployment of automated driving technologies, due to the complex operational domain of AVs. One candidate is scenario-based assessment, in which test cases are derived from real-world road traffic scenarios obtained from driving data. Because of the high variety of the possible scenarios, using only observed scenarios for the assessment is not sufficient. Therefore, methods for generating additional scenarios are necessary. Our contribution is twofold. First, we propose a method to determine the parameters that describe the scenarios to a sufficient degree without relying on strong assumptions on the parameters that characterize the scenarios. By estimating the probability density function (pdf) of these parameters, realistic parameter values can be generated. Second, we present the Scenario Representativeness (SR) metric based on the Wasserstein distance, which quantifies to what extent the scenarios with the generated parameter values are representative of real-world scenarios while covering the actual variety found in the real-world scenarios. A comparison of our proposed method with methods relying on assumptions of the scenario parametrization and pdf estimation shows that the proposed method can automatically determine the optimal scenario parametrization and pdf estimation. Furthermore, we demonstrate that our SR metric can be used to choose the (number of) parameters that best describe a scenario. The presented method is promising, because the parameterization and pdf estimation can directly be applied to already available importance sampling strategies for accelerating the evaluation of AVs.

Figures (10)

• Figure 1: Schematic representation of the scenario category "lvd". The left vehicle is the ego vehicle.
• Figure 2: Schematic representation of the scenario category "cut-in". The left vehicle is the ego vehicle.
• Figure 3: Speed of the leading vehicle during 100 randomly-selected observed lvd scenarios. For plotting purposes, the starting time of each scenario is set to 0.
• Figure 4: The first $n_{\mathrm{t}}=50$ coordinates of the first four columns of $U$ after scaling with $\alpha$ for the lvd scenarios. Note that $\oslash$ denotes element-wise division.
• Figure 5: Five scenarios that require the highest average deceleration of the follower. The black lines denote the observed scenarios and the gray lines denote their approximations based on the $d=4$ parameters. The corresponding initial time gaps are listed in \ref{['tab:corner cases']}.

Theorems & Definitions (3)

• Definition 1: Scenario
• Definition 2: Scenario category
• Remark 1