Enhancing smart road safety with federated learning for Near Crash Detection to advance the development of the Internet of Vehicles

• DOI: 10.1016/j.engappai.2024.108350
• Publication Date: 2024-04-05T19:25:23Z
• Authors: Youcef Djenouri; Ahmed Nabil Belba...; Tomasz Michalak; Asma Belhadi; Gautam Srivastava
• Keywords: Identification

We introduce an innovative methodology for the identification of vehicular collisions within Internet Vehicles (IoV) applications. This approach combines a knowledge base system with deep learning model selection in ensemble setting. It is designed to provide general near-crash detection capability without relying on domain-specific knowledge, enabling development generic models. Our proposed employs novel approach, wherein multiple models are individually trained each image. Subsequently, visual features computed and stored image, along associated loss values from training phase. information utilized select most suitable processing new image data during testing To facilitate efficient selection, we employ kNN (k Nearest Neighbors) strategy. enhance both security IoV environments, implement intelligent federated (FL) Users organized into clusters, two distinct aggregation methods, departing conventional approaches. In initial stage, aggregate all users create global representing collective knowledge. subsequent cluster generate customized local provided models, allowing them their specific crash needs. test our that call Knowledge Guided Deep Learning Near Crash Detection (KGDL-NCD), well-known NCD benchmarks. The results demonstrate KGDL-NCD surpasses baseline solutions, achieving AUC (Area Under Curve) metric 0.95.