| 摘要: |
| 基于模糊函数(Ambiguity Function,AF)赋型的恒模波形设计是雷达系统中的一项关键技术。该问题可构造为一个非线性的复四次问题(NP-hard)。现有的方法可分为两类:第一类方法通过松弛原问题来进行求解,但不可避免地会引入近似误差；第二类方法直接对该问题进行求解,但该类方法的超参数选取较为困难,同时面临较大的计算复杂度。我们注意到深度神经网络是一个天然的非线性系统,与上述的非线性问题模型高度契合。因此,本文提出了一种基于深度学习残差网络的方法来对AF进行赋型,该方法无需对该问题进行任何松弛。首先,我们将该问题转化为了一个无约束的相位优化问题。然后,该无约束问题的非凸目标函数构造为了网络的损失函数。最后,使用残差网络直接对波形的相位进行优化。仿真结果表明,与文献[9]和文献[12]中的方法相比,所提出的方法的信干比(Signal-to-Interference Ratio,SIR)提高了70.47dB和26.86dB,并且有着更好的目标探测性能。 |
| 关键词: 深度学习 模糊函数 残差网络 恒模约束 波形设计 |
| DOI： |
| 分类号:TN958.2 |
| 基金项目:国家自然科学基金(NO.62231006)；国家重点研发计划(2023YFF0717400)；衢州市人民政府(NO.2023D040、2023D009、2022D009、2022D013和2022D033)；四川省科技计划项目(编号:2023YFG0176) |
|
| An Ambiguity Function Shaping Method Based on Deep Learning Residual Network |
|
|
|
|
| Abstract: |
| Waveform design under constant modulus constraint (CMC) for ambiguity function (AF) shaping is a crucial technique in radar systems. This problem is formulated as a nonlinear complex quartic optimization problem (NP-hard) with respect to the transmitted waveform. Existing methods can be divided into two categories: the first one solves the problem by relaxing the original problem, but inevitably introduces approximation errors; the second one solves the problem directly, but the selection of hyper-parameters in this category is more difficult, while facing a larger computational complexity. Firstly, we transform the problem into an unconstrained phase optimization problem. Subsequently, we formulate the non-convex objective function of the problem as the loss function of the network. Finally, a residual network is employed to directly optimize the phase of the waveform. Simulation results demonstrate that the proposed method surpasses the methods in [9] and [12] in terms of Signal-to-Interference Ratio (SIR) by 70.47dB and 26.86dB, respectively, while achieving superior target detection performance. |
| Key words: Deep learning Ambiguity function Residual network Constant modulus constraint Waveform design |