面向空地高速互联的太赫兹动中通架构与关键技术

摘要： 梳理了主动相控阵、被动相控阵和机械伺服3类太赫兹动中通体制，并讨论了2类跟踪技术，包括以四馈源和差单脉冲闭环架构为代表的机械伺服跟踪方案，以及相控阵波束电子扫描方案。通过和差网络形成合通道、方位差通道和俯仰差通道，其中合通道用于通信解调和链路监测，两路差通道经差和比计算后给出有符号二维角误差，以驱动方位—俯仰云台；针对主动相控阵和被动相控阵，提出了学习增强蜂窝周期波束跟踪方法，利用长短期记忆网络（LSTM）预测运动模型残差，采用扩展卡尔曼滤波（EKF）融合导频观测，并通过蜂窝搜索完成低开销波位校正。在收发两端均配置32×32阵列的代表性机动场景中，该方法获得了0.092 5°的径向均方根误差和0.162 2°的95%分位误差。上述两类方法为太赫兹动中通提供了兼顾跟踪范围、训练开销和工程可实现性的技术路径。

关键词：太赫兹通信；动中通；相控阵；波束跟踪；和差单脉冲跟踪

Abstract: Three technical architectures for terahertz communication on the move are reviewed, namely active phased arrays, passive phased arrays, and mechanical servo systems. Two beam-tracking approaches are discussed, including a mechanical servo tracking scheme based on a four-feed sum-and-difference monopulse closed loop and a phased-array electronic beam-scanning scheme. The sum-and-difference network forms a sum channel, an azimuth-difference channel, and an elevation-difference channel. The sum channel is used for communication demodulation and link monitoring, while the two difference channels are processed through difference-over-sum calculation to obtain signed two-dimensional angular errors for driving the azimuth–elevation gimbal. For active and passive phased arrays, a learning-enhanced periodic honeycomb beam-tracking method is proposed. A long short-term memory (LSTM) network predicts motion-model residuals, an extended Kalman filter (EKF) fuses pilot observations, and a honeycomb search performs low-overhead beam-position correction. In a representative maneuvering scenario with 32 × 32 arrays at both the transmitter and receiver, the proposed method achieves a radial root-mean-square error of 0.092 5° and a 95th-percentile error of 0.162 2°. The two approaches provide a technical pathway for terahertz communication on the move that balances tracking range, training overhead, and engineering implementation complexity.

Keywords: terahertz communication; communication on the move; phased array; beam tracking; sum-and-difference monopulse tracking