Digital Twin Edge Services With Proximity-Aware Longitudinal Lane Changing Model for Connected Vehicles

Edge-assisted digital twin (DT) is an efficient way to monitor, analyze, and guide connected vehicles by creating a virtual representation of the vehicle at the edge. This allows for real-time monitoring of vehicle performance, location, and condition, as well as providing guidance and edge services to optimize the vehicle's performance and ensure safe operation. Current traffic digital twinning methods primarily rely on data-driven approaches for predicting and replicating vehicle behavior, which do not adequately capture the complex dynamics and interactions between vehicles. In this paper, we propose a proximity-aware longitudinal lane changing (PALLC) model that simply and efficiently models longitudinal lane changing behavior to develop a DT system. More specifically, we first observe vehicle behavior, and find that the lane-changing behavior of the target vehicle is mainly influenced by neighboring vehicles, and ignoring the influence of distant vehicles can improve the accuracy of the PALLC model. Next, we theoretically analyze the convergence properties of the PALLC model for forecasting and assessing collision risks. Uncertainty factors originating from DT modeling, including model parameter biases, measurement deviations, and environmental noise, are fully considered to explore their effects on the DT system. In the experiments of different types of vehicles and different indicators, the results show that the PALLC model all have higher accuracy compared with the traditional method. Moreover, the PALLC model is more stable under the influence of parameter deviation and measurement deviation. These findings highlight its potential to provide DT edge services for connected vehicles.

I. Introduction

With the increasing popularity of connected vehicles (CVs), driving safety and efficiency have been significantly enhanced, but also drives the increasing need for low-latency and high-access-rate communications [1]. Edge artificial intelligence (AI) is vital for automated transportation systems as it enhances data processing and decision making, reduces system dependencies, and improves traffic safety, efficiency, and user experience. However, it is insufficient for intelligent transportation because it faces challenges in verifying processing and computation results, which may lead to a risk of traffic accidents [2]. Additionally, it is difficult to monitor vehicle conditions in real time, making the integration and allocation of transport resources challenging. Digital twins (DTs) based on the integration of AI, the Internet of Things (IoT), and Big Data are becoming promising technology to digitise physical entities, systems, or data in the physical world. DTs are representations of physical entities that can interact and synchronize with the physical world in real time [3]. Edge-assisted DT, as a localized approach, is used to enhance real-time monitoring, decision-making, and strategy-optimizing capabilities for connected vehicles. This allows the digital twin to provide reliable edge services for the physical system, enabling real-time monitoring of vehicle performance, location, and condition, as well as providing risk assessment and driving guidance to ensure safe operation and optimize the vehicle's performance.