Efficient One-Way Time Synchronization for VANET with MLE-Based Multi-Stage Update

As vehicular networks become increasingly pervasive, enhancing connectivity and reliability has emerged as a critical objective. Among the enabling technologies for advanced wireless communication, particularly those targeting low latency and high reliability, time synchronization is critical, especially in vehicular networks. However, due to the inherent mobility of vehicular environments, consistently exchanging synchronization packets with a fixed base station or access point is challenging. This issue is further exacerbated in signal shadowed areas such as urban canyons, tunnels, or large-scale indoor halls where other technologies, such as global navigation satellite system (GNSS), are unavailable. One-way synchronization techniques offer a feasible approach under such transient connectivity conditions. One-way schemes still suffer from long convergence times to reach the required synchronization accuracy in these circumstances. In this paper, we propose a WLAN-based multi-stage clock synchronization scheme (WMC) tailored for vehicular networks. The proposed method comprises an initial hard update stage to rapidly achieve synchronization, followed by a high-precision stable stage based on Maximum Likelihood Estimation (MLE). By implementing the scheme directly at the network driver, we address key limitations of hard update mechanisms. Our approach significantly reduces the initial period to collect high-quality samples and offset estimation time to reach sub-50 s accuracy, and subsequently transitions to a refined MLE-based synchronization stage, achieving stable accuracy at approximately 30 s. The windowed moving average stabilized (reaching 90% of the baseline) in approximately 35 s, which corresponds to just 5.1% of the baseline time accuracy. Finally, the impact of synchronization performance on the localization model was validated using the Simulation of Urban Mobility (SUMO). The results demonstrate that more accurate conditions for position estimation can be supported, with an improvement about 38.5% in the mean error.