本文研究了一类在非高斯噪声影响下的状态饱和系统的分布式隐私保护熵滤波问题。首先,引入变量分解策略结合系统状态饱和的特征设计了一个分布式滤波器,其中考虑的系统遭受的乘性噪声以及非高斯噪声。接下来,结合已有的符合约束条件的时变动态密钥以及分解规则,得到私有误差协方差矩阵与公共协方差矩阵之间的关系以及它们的上界。在此基础上,最小化基于S-t核函数的最大熵准则下的代价函数得到待设计的滤波器增益。最后,通过对比窃听者滤波误差协方差和正常滤波误差协方差,分析了所设计的滤波算法的隐私性。最后利用仿真算例验证了算法的有效性以及安全性。This paper studies the distributed privacy-preserving entropy filtering problem for a class of state-saturated systems under the influence of non-Gaussian noise. Firstly, a distributed filter is designed by introducing a variable decomposition strategy and considering the characteristics of system state saturation, where the multiplicative noise and non-Gaussian noise suffered by the system are taken into account. Next, by combining the existing time-varying dynamic keys that meet the constraint conditions and the decomposition rules, the relationship between the private error covariance matrix and the public covariance matrix and their upper bounds are obtained. On this basis, the filter gain to be designed is obtained by minimizing the cost function under the maximum entropy criterion based on the S-t kernel function. Finally, the privacy of the designed filtering algorithm is analyzed by comparing the eavesdropper’s filtering error covariance with the normal filtering error covariance. The effectiveness and security of the algorithm are verified by simulation examples.
超宽带(Ultra Wide Band, UWB)技术因其高精度和强抗干扰能力,在室内定位领域中有着广泛的应用。然而,在复杂的室内环境中,UWB信号易受多径效应和非视距条件的影响,使得定位精度下降。为此,文章提出了一种基于深度学习的UWB室内定位方法。通过引入双向门控循环单元(Bidirectional Gated Recurrent Unit, BiGRU)与Bahdanau注意力机制的结合模型,充分挖掘UWB信号的时序特征和关键信息。BiGRU利用其在时序数据处理中的优势,有效捕捉UWB信号的动态特征,而Bahdanau注意力机制通过动态权重分配,增强模型对关键特征的关注,从而提高定位精度。实验结果表明,文章提出的模型平均定位误差为6.9 cm,相较于传统的循环神经网络(Recurrent Neural Network, RNN)、长短时记忆(Long Short-Term Memory, LSTM)网络和门控循环单元(Gated Recurrent Unit, GRU),误差减少了约29.59%至42.98%。研究结果表明,结合BiGRU与Bahdanau注意力机制的深度学习模型在复杂环境下具有更高的鲁棒性和定位精度。Ultra Wide Band (UWB) technology is widely used in indoor positioning due to its high accuracy and strong anti-interference capability. However, in complex indoor environments, UWB signals are susceptible to multipath effects and non-line-of-sight conditions, which degrade positioning accuracy. To address this issue, this paper proposes a deep learning-based UWB indoor positioning method. By introducing a combined model of the Bidirectional Gated Recurrent Unit and Bahdanau attention mechanism, the method effectively exploits the temporal features and key information of UWB signals. BiGRU leverages its advantages in handling sequential data to capture the dynamic characteristics of UWB signals, while the Bahdanau attention mechanism enhances the model’s focus on critical features through dynamic weight allocation, thus improving positioning accuracy. Experimental results show that the average positioning error of the proposed model is 6.9 cm, which represents a reduction of approximately 29.59% to 42.98% compared to traditional Recurrent Neural Network, Long Short-Term Memory Network, and Gated Recurrent Unit. The results demonstrate that the deep learning model combining BiGRU and the Bahdanau attention mechanism offers higher robustness and positioning accuracy in complex environments.
暂无评论