查看更多>>摘要：The following quote was obtained by the news editors from the background informa tion supplied bythe inventors: “A high frequency with abundant spectrum resourc es is an effective means to improve theperformance of a wireless system. Howeve r, since the high frequency has a relatively high carrier frequencyand a large path loss, a sending direction needs to be aligned with a user according to the beamformingtechnique, thereby concentrating energy to transmit information so a s to overcome performance degradationcaused by an excessively large path loss. Generally, a method for obtaining a direction of a transmitor receive beam is t o perform beam sweeping in each direction and select a beam direction with relatively good performance as the beam direction for transmitting the information. W hen the beam sweepingis performed, each beam corresponds to one reference signa l resource. In a case of a very thin beam,many beam directions need to be swept , and correspondingly, an overhead of the reference signal resourceis very larg e. As shown in FIG. 1, if a regular beam, such as a single beam based on discret e Fouriertransform (DFT), is used, more beams in different directions need to b e swept when beam sweeping isperformed, thereby requiring a larger overhead of pilot resources. As shown in FIG. 2, if an irregular beamresource, such as a be am including multiple superimposed beam directions, is used, beams in multiple directions can be simulated at a time, thereby reducing the number of times of be am sweeping. However, abeam orientation, beam width and beam gain of the irregu lar beam are all uncertain. At present, althoughartificial intelligence (AI) ca n reduce the number of reference signal resources for beam sweeping to acertain extent, too many samples need to be trained due to a relatively thin beam at a parameter trainingstage, resulting in prohibitive cost.”
NETL
NSTL