In mobile communication systems, channel aging caused by channel changes continuously poses a challenging issue. However, it would incur significant overhead to estimate the channel state frequently by shortening the pilot period. This paper proposes a channel prediction method for wideband multiple-input multiple-output (MIMO) systems based on Convolutional Long Short-Term Memory Network (ConvLSTM) to tackle channel aging with minimal overhead. This method takes advantage of the properties of ConvLSTM for processing spatiotemporal data, which enhances the time-domain prediction aided by the strong frequency-domain correlation of the wideband channel. The sliding window model is combined to achieve continuous prediction. Newly acquired data from each prediction iteration is employed in ongoing adaptation to channel fluctuations. In order to mitigate the error propagation issues in time-series calculations, the pilot is promptly inserted to update the channel when predictive performance falls below a predefined threshold. Experimental results demonstrate that the approach presented in this paper significantly reduces the pilot overhead while maintaining excellent communication performance.
Channel Tracking with Minimal Overhead for Wideband MIMO Mobile Communication Systems
2024-06-24
1445651 byte
Conference paper
Electronic Resource
English
Testing MIMO Systems with Coupled Reverberations Chambers: A Wideband Channel Model
British Library Conference Proceedings | 2006
|A Nonstationary Wideband MIMO Channel Model for High-Mobility Intelligent Transportation Systems
Online Contents | 2015
|