Measurement Model Optimization For Channel Prediction Improvement In Wireless Networks

In this paper, a novel approach for improving channel prediction is proposed. Its main novelty consists in optimizing the post processing of acquired measurements intended to assess channel quality. To this end, a Multi-Armed Bandit based algorithm is designed to choose in an adaptive and online manner the optimal filtering parameters, taking into account the target usage in the network as well as the employed prediction algorithm. Performance evaluation is made following 3GPP assumptions and compared to conventional fixed filtering operation. Substantial gains in terms of prediction accuracy and UE throughput are obtained, especially in cases where the user context is fast varying.