Adaptive Federated Learning Empowered Wireless Localization Framework Using Vehicle Sensors

Currently, wireless localization plays a vital role in supporting a wide range of tasks within smart cities. For example, vehicle tracking services are increasingly used in tunnels and bridges to detect objects and prevent collisions. Both indoor and outdoor localization are equally important for applications such as traffic management and vehicle positioning. However, existing localization techniques still face significant challenges, particularly with respect to accuracy, latency, and resource consumption. Addressing these limitations is therefore essential to ensure reliable and efficient operation in smart city environments. This study proposes an Adaptive Federated Learning–Enabled Wireless Localization Framework (ALFLS) designed specifically for mobility-based vehicle tasks. The novelty of ALFLS lies in its ability to apply pattern learning for both indoor and outdoor localization, leveraging federated learning to execute tasks while maintaining high quality of service. The framework also incorporates an optimized placement strategy for edge and cloud node resources, supported by a training algorithm that enhances real-time localization accuracy. To evaluate performance, experimental localization datasets were tested on a benchmark testbed, highlighting the practical benefits of ALFLS. The simulation results demonstrate that the framework improves localization accuracy by up to 98%, reduces latency by approximately 30%, and achieves significantly higher resource utilization compared to existing methods. These results confirm that ALFLS provides a robust, efficient, and scalable solution for addressing the persistent challenges of wireless localization in smart city environments.