Distributed Federated-Multi-Armed Bandit Learning for Content Management in Connected UAVs

Abstract

This article proposes an unmanned aerial vehicle (UA V) aided content management system in communication-challenged disaster scenarios. Without cellular infrastructure, community of stranded users can be provided access to situation-critical content using a hybrid UAV. A set of relatively static anchor UA Vs with vertical as well as lateral links can provide content access to its local users. A set of ferrying UA Vs with only lateral links, but with wider mobility, can provision content while visiting different communities of stranded users. The objective is to design a content dissemination system that learns content caching policies on-the-f1y in order to maximize content availability to the users. This article proposes a distributed Federated Multi-Armed Bandit Learning model for UA V-caching decision-making that takes geo-temporal differences in content popularity and heterogeneity in content demands into consideration. The proposed paradigm combines the expected reward maximization attribute of Multi-Armed Bandit, and the distributed intelligence sharing of Federated Learning for caching decision at the UAVs. It is demonstrated that Federated aggregation of individual models improve system performance while making the learning fast and adaptive via sharing models learnt by individual UAVs. The article does functional verification and performance evaluation of proposed caching framework for heterogeneous popularity distributions, and different inter-community geographical characteristics.