Enhancing Traffic Management for Connected Vehicles using Deep Learning with Integrated Trust Analysis and CyberGIS Systems

Abstract

With the exponential increase in vehicles and traffic congestion, next-generation traffic light monitoring & control systems (NTLMCS) are crucial for improving traffic management. These systems provide real-time traffic data to drivers and traffic police via a holographic digital ad wall at traffic signals. Vehicle-to-everything communication is established using the well-known dedicated short-range communication protocol. Vehicles near the hologram communicate essential parameters such as traffic timing, vehicle density, trustability factor, and fake packet alerts using a strategic connected dominating set approach. Traffic timing, derived from vehicle density, is communicated from NTLMCS to connected vehicles, aiding in accurate traffic management. The trust analysis stability system within vehicles evaluates the integrity of packets exchanged between connected vehicles. CyberGIS, which integrates cyber infrastructure, geographic information systems, and spatial analysis, were combined with autoencoders for identifying anomalies in global positioning system (GPS) data by comparing expected and actual signal patterns, thereby detecting potential manipulation. Next, the research involved the analysis of differentiating the normal and GPS-manipulated data by applying autoencoder models with various optimization techniques. We found that the autoencoder model trained with the Adam optimizer shows a promising prediction accuracy of 97% for differentiating the normal and GPS-manipulated data. Overall, the proposed framework not only improves traffic signal flow during peak hours but also enhances the overall security and efficiency of urban traffic management systems.