A Review of Scenario Complexity Assessment for Automated Vehicles

Abstract

As automated vehicles (AVs) transition towards Levels 3 and 4 of automation, their Operational Design Domain (ODD) inevitably expands. Consequently, the assessment of scenario complexity has emerged as a critical prerequisite for delineating the performance boundaries of AVs, facilitating critical functions such as emergency take-overs and minimal risk maneuver execution. This paper aims to review the method of scenario complexity assessment for autonomous vehicles. Totally 260 articles were identified through a systematic literature search and 62 articles were included in this review. Then, we categorized existing quantification methodologies into two distinct paradigms: element-based and interaction-based approaches. Element-based methods quantify complexity by enumerating static road and traffic constituents—such as road users and infrastructure—yet frequently neglect dynamic interdependencies. In contrast, interaction-based approaches largely deduce complexity through the analysis of behavioral interactions among road users or the response characteristics of autonomous driving systems; however, given their strong dependence on specific algorithmic capabilities, these methods are often limited in terms of universality. The contrastive analysis was performed. Finally, based on the review, this paper outlined the future work in the field of scenario complexity assessment. This paper provides an integrated review for the cutting-edge knowledge and fresh insights for research into the scenario complexity assessment, helping for the future test and evaluation of the AVs.