Dynamic D2D-Assisted Federated Learning Over O-RAN: Performance Analysis, MAC Scheduler, and Asymmetric User Selection

Existing studies on federated learning (FL) aremostly focused on system orchestration for static snapshots ofthe network and making static control decisions (e.g., spectrumallocation). However, real-world wireless networks are susceptibleto temporal variations of wireless channel capacity and users’datasets. In this paper, we study the impacts of the dynamics:1) wireless channels and 2) users’ datasets on the FL execution.The former is captured by introducing a set of discrete timeevents while the latter is characterized by a novel ordinarydifferential equation and the metric of dynamic model drift,formulated via a partial differential inequality, drawing concreteanalytical connections between the dynamics of users’ datasetsand FL accuracy. We then propose dynamic cooperative FLwith dedicated MAC schedulers (DCLM), exploiting the uniquefeatures of open radio access network (O-RAN) to execute FL.DCLM entails: 1) a hierarchical device-to-device (D2D)-assistedmodel training; 2) dynamic control decisions through dedicatedO-RAN MAC schedulers; and 3) asymmetric user selection. Weprovide extensive theoretical analysis to study the convergenceof DCLM and then aim to optimize its degrees of freedom (e.g.,user selection and spectrum allocation) through a non-convexoptimization problem. We develop a systematic and genericapproach to obtain the solution for this problem. We finally showthe efficiency of DCLM via numerical simulations and provide aseries of future directions.