Adaptive elastic echo state network for channel prediction in IEEE802.11ah standard-based OFDM system

Abstract

As a promising wireless communication technology, the IEEE802.11ah standard is designed to connect various sensors in the Internet of Things (IoT) in future. It is important to investigate adaptive transmission in the IEEE802.11ah standard. However, exact channel state information (CSI) is required. Channel prediction is an available approach. Therefore, an adaptive elastic echo state network (AEESN) for channel prediction in the IEEE802.11ah standard-based orthogonal frequency division multiplexing (OFDM) system is introduced in this paper. The AEESN includes two key components, a basic echo state network and an adaptive elastic network. The latter is imported to overcome collinearity problems due to vast neurons in the former and to avoid ill-conditioned solutions when estimating output weights in the former. Moreover, the latter can produce sparse output weights, which reduces memory storage requirements. To evaluate system performances, 1MHz and 2MHz bandwidth cases with specified parameters are tested. One-step prediction, multi-step prediction and robustness are evaluated for various signal to noise ratios (SNRs). The results indicate that the AEESN not only offers satisfactory prediction performance, but also effectively avoids ill-conditioned solutions and produces sparse output weights. Therefore, it can assure adaptive IoT communication.