ODD-Centric Contextual Sensitivity Analysis Applied To A Non-Linear Vehicle Dynamics Model

TL;DR

This paper introduces an ODD-centric sensitivity analysis method for vehicle dynamics models, enhancing the evaluation of model robustness within specific operational design domains for automated driving systems.

Contribution

It presents a novel ODD-focused sensitivity analysis framework applied to a complex nonlinear vehicle model, aiding in the development of ODD-compliant behavior strategies.

Findings

Sensitivity analysis highlights model vulnerabilities within ODDs

Strategies to minimize model sensitivity improve safety and reliability

Framework applicable to complex vehicle dynamics models

Abstract

Advanced driving functions, for assistance or full automation, require strong guarantees to be deployed. This means that such functions may not be available all the time, like now commercially available SAE Level 3 modes that are made available only on some roads and at law speeds. The specification of such restriction is described technically in the Operational Design Domain (ODD) which is a fundamental concept for the design of automated driving systems (ADS). In this work, we focus on the example of trajectory planning and control which are crucial functions for SAE level 4+ vehicles and often rely on model-based methods. Hence, the quality of the underlying models has to be evaluated with respect to the ODD. Mathematical analyses such as uncertainty and sensitivity analysis support the quantitative assessment of model quality in general. In this paper, we present a new approach to assess the quality of vehicle dynamics models using an ODD-centric sensitivity analysis. The sensitivity analysis framework is implemented for a 10-DoF nonlinear double-track vehicle dynamics model used inside a model-predictive trajectory controller. The model sensitivity is evaluated with respect to given ODD and maneuver parameters. Based on the results, ODD-compliant behavior generation strategies with the goal of minimizing model sensitivity are outlined.