A Neural Network for Estimating CQI in 5G Communication Systems

Channel quality indicator (CQI) is a key parameter in communication system design that encodes the state of the channel. With this information, a base station (BS) can adjust the quality of service that would best suit the channel at that time and place, thereby facilitating communications. As it is counterproductive to request CQI from all users, it is preferable to estimate it in some cases. This paper studies the current CQI request-estimation paradigm and proposes a neural network based solution to attain the best of both worlds. We show that the neural network based architecture outperforms the legacy based system. With radio access network (RAN) architectures being virtualized, we can expect a lot of baseband processing to be offloaded to the Cloud in proximity of the base station site. The Edge Cloud is expected to have large computational capabilities, which can suitably host our neural network based solution.