Time-dependent MIMO system channel prediction method based on multitask learning

Abstract

The invention discloses a time- dependent MIMO system channel prediction method based on multitask learning, which mainly solves a problem of mismatching of channel state information and channels of the channels in a MIMO system caused by time-variant characteristics. An implementation scheme of the method is as follows: 1) obtaining a time-dependent MIMO channel state information historical observed value through channel measurement; 2) regarding the channel state information on different transmitting and receiving antenna pairs as different tasks, simultaneously inputting the tasks in a multitask learning algorithm to perform joint learning, and based on an internal relation among the multiple antennae, acquiring more sufficient characteristic space, and obtaining multitask learning algorithm parameters; 3) jointly inputting the measured channel state information data and the multitask learning algorithm parameters into the multitask learning algorithm to perform channel prediction,and obtaining predicted channel state information. The method provided by the invention is simple and practicable, has excellent effects and small calculated quantity and can be used in the time-dependent MIMO system.

Description

technical field [0001] The invention belongs to the technical field of communication, and in particular relates to a channel prediction method, which can be used in a time-correlated MIMO system. Background technique [0002] In wireless communication, reflection, diffraction and scattering commonly exist in various environments, and multipath propagation phenomena inevitably exist; the relative movement of the transmitting end and the receiving end inevitably produces Doppler expansion, which makes the wireless channel appear frequency Selectivity and time-varying properties. Fading due to frequency selectivity and time-varying properties is considered as two different distortions. The former depends on the multipath extension and is characterized by coherent bandwidth; the latter depends on the time variation of the channel and is characterized by coherent time. In order to reduce the bit error rate of the system, the estimated channel state information is usually used t...

AI Technical Summary

Problems solved by technology

Compared with the traditional prediction method, the prediction accuracy of the neural network prediction method will be better, but the network parameters of the traditional neural network algorithm are real numbers, and the MIMO channel state information is complex data, so the traditional neural network algorithm is not suitable for channel prediction

Moreover, the existing nonlinear prediction algorithms only consider the channel state information on each transmit-receive antenna pair separately, without taking into account the internal relationship between each transmit-receive antenna pair, resulting in insufficient training samples and poor prediction results. question

Images