Digital Twin-Driven Asymmetric Driving Aggressiveness Assessment for Connected Autonomous Vehicles

Accurately evaluating the driving aggressiveness could greatly improve the driving efficiency for connected autonomous vehicles (CAV) in the mixed traffic. Current aggressiveness evaluation methods merely focus on the collision probability, inadequately assessing hazard levels due to symmetry between interactive vehicles. Meanwhile, existing methods operate on a microcosmic scale, guiding motion planning within short time. With the advent of digital twin technology, aggressiveness assessments of road segments and mesoscopic planning become feasible. Thus, a novel driving aggressiveness assessment model is proposed in this paper based on asymmetric interactions between different types of vehicles. The interaction behaviors are analyzed, and the mathematic model of the aggressiveness is deduced. Subsequently, an application case of the proposed model in lane-change decision-making is, verifying its capacity to generate asymmetric driving behavior. The aggressiveness model provides a new perspective on asymmetric road safety evaluation and heterogeneous driving behavior in connected traffic environments.

I. Introduction

Heterogeneous driving threat perception constitutes a critical factor in determining the acceptance of autonomous vehicles by human drivers or passengers in mixed traffic, encompassing human-driven vehicles, autonomous vehicles, and various vehicle types [1]. Conventional driving aggressiveness assessment methods typically consider collision probability, neglecting potential collision severity evaluation, resulting in symmetrical evaluation for interactive vehicles [2]. This inconsistency with actual driving situations leads to identical driving behavior generation for distinct vehicles if decision-making models rely on symmetric threat model [3], [4]. Therefore, enabling autonomous vehicles to identify asymmetric threats like human drivers is an essential issue in automated driving technology.