Intelligent Vehicle Operation Risk Assessment and Early Warning Method for Unsignalised Intersections by Integrating Spatiotemporal Status and LiDAR Point Cloud Top‐View Features
IET Intelligent Transport Systems · 2026-01 · 20 (1)
DOI: 10.1049/itr2.70365 ↗Kurz gefasst
Fachartikel in der Zeitschrift IET Intelligent Transport Systems, online erschienen am 10.11.2026. Verfasst von Ruibin Zhang. Die Arbeit nennt 44 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Unsignalised intersections are high‐risk for traffic accidents due to complex road topologies and variable traffic environments, while single risk assessment indicators in existing driver assistance systems insufficiently characterise vehicle operational risks. To solve this problem, this paper proposes an intelligent vehicle operational risk assessment and early warning method for unsignalised intersections, integrating spatiotemporal status and LiDAR point cloud top‐view features. 1DCNN extracts vehicle operational spatiotemporal features, and an improved MobileNetV2‐SAM network extracts LiDAR point cloud top‐view scene features; the two types of features are fused via a weighted concatenation strategy and input into the Attention‐BiLSTM model to predict target vehicle motion states. An improved risk field model considering target vehicle heading angle quantifies the ego vehicle's operational risk, and risk levels are determined by comparing the risk field with the closest distance between target and ego vehicles for corresponding early warnings. A local path planning model based on the intersection of predicted risk field equipotential lines and quintic spline curves provides safe driving paths. Experimental results show that the Attention‐BiLSTM model achieves superior target vehicle state prediction with the lowest MSE/MAE and highest R 2 for all state parameters; the full model has an inference latency of 15.6 ms, meeting real‐time requirements of on‐board systems. The improved risk field model effectively characterises risk distribution of turning and lane‐changing vehicles, and the path planning method accurately identifies safe paths. The proposed closed‐loop framework integrating prediction, assessment, warning and path planning provides a practical technical solution for intelligent vehicle applications at unsignalised intersections.
Zitieren
APA
Zhang, R. (2026). Intelligent Vehicle Operation Risk Assessment and Early Warning Method for Unsignalised Intersections by Integrating Spatiotemporal Status and LiDAR Point Cloud Top‐View Features. IET Intelligent Transport Systems, 20(1). https://doi.org/10.1049/itr2.70365
Vancouver
Zhang R. Intelligent Vehicle Operation Risk Assessment and Early Warning Method for Unsignalised Intersections by Integrating Spatiotemporal Status and LiDAR Point Cloud Top‐View Features. IET Intelligent Transport Systems. 2026;20(1). doi:10.1049/itr2.70365
BibTeX
@article{Zhang2026,
author = {Zhang, Ruibin},
title = {Intelligent Vehicle Operation Risk Assessment and Early Warning Method for Unsignalised Intersections by Integrating Spatiotemporal Status and LiDAR Point Cloud Top‐View Features},
journal = {IET Intelligent Transport Systems},
year = {2026},
volume = {20},
number = {1},
doi = {10.1049/itr2.70365}
}
EUVIMED