Embedded Federated Learning for VANET Environments

In the scope of smart cities, sensors scattered throughout city generate information that supplies intelligence mechanisms to learn city’s mobility patterns. These patterns are used in machine learning (ML) applications, such as traffic estimation, allow for improvement quality experience city. Owing Internet-of-Things (IoT) evolution, monitoring points always growing, and transmission mass data generated from edge devices cloud, required by centralized ML solutions, brings great challenges terms communication, thus negatively impacting response time and, consequently, compromising reaction improving flow vehicles. addition, when moving between exposed, privacy. Federated (FL) has emerged an option these challenges: (1) It lower latency communication overhead performing most processing on devices; (2) it improves privacy, do not travel over network; (3) facilitates handling heterogeneous sources expands scalability. To assess how FL can effectively contribute scenarios, we present framework, which built a testbed integrated components infrastructure, where NVIDIA Jetson were connected cloud server. We deployed our lightweight container-based framework this testbed, evaluated performance devices, effectiveness aggregation algorithms, impact server, consumption resources. carry out evaluation, opted scenario estimated vehicle inside outside city, using real collected Aveiro Tech City Living Lab sensing infrastructure Aveiro, Portugal.