面向航天测控的高效自适应滤波算法及其加速实现

An efficient adaptive filtering algorithm and its accelerated implementation for aerospace TT&C

  • 摘要: 在航天测控中,日益复杂的空间电磁环境对远距离星−地传输链路的可靠性和有效性形成威胁。当链路中存在多个特定来向的干扰时,运用传统的自适应滤波功率倒置准则无法有效滤除压制式干扰以外的其他干扰信号。该文面向智能天线应用及地面测控站抗干扰接收场景,提出一种多空域约束的功率倒置准则及自适应滤波算法硬件加速实现方案。在波束成形中,通过对功率倒置准则添加零陷方向约束,实现功率域和空间域的非期望信号抑制;在实现时采用矩阵LDL分解完成低复杂度矩阵求逆,同时通过定点−浮点转换提升计算精度,进一步利用脉动阵列计算单元实现高效矩阵运算。搭建了测控信号接收系统,将上述自适应滤波算法部署在硬件上进行实验验证,结果表明:所提方法对指定来向的带内干扰和任意来向的压制式带内干扰均能产生深零陷,且对测控信号无失真影响;计算得到的自适应滤波最优权值向量相对误差在10−6~10−7量级。

     

    Abstract: In aerospace telemetry, tracking, and control (TT&C) systems, the increasingly complex space electromagnetic environment threatens the reliability and effectiveness of the satellite-earth communication links. When there are multiple interference signals arriving from specific directions in the communication link, the conventional beamforming power inversion criterion fails to effectively suppress interference signals other than suppressive jamming. Focusing on smart antenna applications and anti-interference reception scenarios in TT&C ground stations, this paper proposes a multi-spatial-domain constrained power inversion criterion and a hardware-accelerated implementation scheme for adaptive filtering algorithms. In beamforming, nulling direction constraints are incorporated into the power inversion criterion to eliminate undesired signals in both power and spatial domains. In implementation, LDL decomposition is adopted to achieve low-complexity matrix inversion; while fixed-point to floating-point conversion is applied to improve computational precision. Furthermore, systolic array units are leveraged to enable highly efficient matrix operations. For verification purpose, a TT&C receiver is constructed and the overall adaptive filtering algorithm mentioned above is deployed on hardware. Experimental results demonstrate that the proposed method generates deep nulls for both in-band interference from specified directions and suppressive in-band interference from arbitrary directions, introducing no distortion to TT&C signals; the relative error of the optimal weight vector for adaptive filtering is on the order of 10−6~10−7 in magnitude.

     

/

返回文章
返回