一种Massive MIMO单小区权值优化方法

• Format: Patent
• Language: Chinese
• Published: 11.03.2022
• Subjects: ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; TRANSMISSION; WIRELESS COMMUNICATIONS NETWORKS

本发明公开了一种Massive MIMO单小区权值优化方法，包括步骤：S1获取数据；S2确认待优化小区：根据运营商的KPI数据及业务需求确认待优化的小区；S3强化学习输出调整权值：结合小区的KPI数据的指标和用户的采样点位置分布，给出候选权值动作组合，选择波束权值方案对候选权值进行强化学习并调整，输出调整权值；S4现网执行：将步骤S3中选择的波束权值方案在现网中执行，并收集反馈后的数据；S5判断是否继续优化：根据收集的反馈数据，判断是否完成优化，若完成优化，则结束该小区的优化过程；若未完成，则返回步骤S3，并循环步骤S3~S5，直至完成优化，并将小区的Q-table保存。 The invention discloses a Massive MIMO (Multiple Input Multiple Output) single cell weight optimization method. The method comprises the following steps: S1, acquiring data; S2, confirming a to-be-optimized cell: confirming the to-be-optimized cell according to KPI data of an operator and a service demand; S3, performing reinforcement learning and outputting an adjustment weight: giving a candidate weight action combination in combination with the indexes of the KPI data of the cell and the sampling point position distribution of the user, selecting a beam weight scheme to carry out reinforcement learning and adjustment on the candidate weight, and outputting the adjustment weight; S4, executing the current network: executin