Abstract
The deployment of future intelligent transportation systems is contingent upon seamless and reliable operation of connected and autonomous vehicles (CAVs). One key challenge in developing CAVs is the design of an autonomous controller that can make use of wireless connectivity and accurately execute control decisions, such as a quick acceleration when merging to a highway and frequent speed changes in a stop-and-go traffic. However, the use of conventional feedback controllers or traditional machine learning based controllers, solely trained by the CAV's local data, cannot guarantee a robust controller performance over a wide range of road conditions and traffic dynamics. In this paper, a new federated learning (FL) framework enabled by the CAVs' wireless connectivity is proposed for the autonomous controller design of CAVs. In this framework, the learning models used by the controllers are collaboratively trained among a group of CAVs. To capture the varying CAV participation in FL and the diverse local data quality among CAVs, a novel dynamic federated proximal (DFP) algorithm is proposed that accounts for the mobility of CAVs, the wireless channel dynamics, as well as the unbalanced and non-independent and identically distributed data across CAVs. A rigorous convergence study is performed for the proposed algorithm under realistic wireless environments. Then, the impact of varying CAV participation in FL process and diverse local data quality of CAVs on the convergence is explicitly analyzed. Simulation results that use real vehicular data show that the proposed DFP-based controller can accurately track the target speed over time and under different traffic scenarios, and it yields a distance error two times smaller than controllers designed using traditional machine learning solutions trained with the CAV's local data. The results also show that the proposed DFP algorithm is well-suited for the autonomous controller design in CAVs when compared to popular FL algorithms, such as federated averaging (FedAvg) and federated proximal (FedProx) algorithms.
