arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2512.02462eess.SP
收藏

基于贝叶斯概率融合的多接入点协同感知移动网络

Bayesian Probability Fusion for Multi-AP Collaborative Sensing in Mobile Networks

School of Information Science and Engineering, Southeast University(东南大学信息科学与工程学院);Purple Mountain Laboratories(紫金山实验室)

Shengheng Liu, Xingkang Li, Yongming Huang, Yuan Fang, Qingji Jiang, Dazhuan Xu, Ziguo Zhong, Dongming Wang, Xiaohu You

AI总结:

本文提出基于贝叶斯概率融合的多接入点协同感知框架,通过正交频分复用波形实现目标参数估计,提升通信效率与感知精度。

摘要:

集成感知与通信被广泛认为是下一代移动网络的基础技术。与单站感知相比,多接入点(AP)协同感知为移动网络提供了更广泛、更准确和更稳健的感知能力,这对于多样化的基于位置的领域至关重要。本文聚焦于多AP网络中的协同感知,并提出了一种基于正交频分复用波形的贝叶斯概率融合框架,用于目标参数估计。该框架将多AP接收信号建模为概率分布,以捕捉来自信道噪声和散射系数的随机观测。然后,将先验信息纳入联合概率密度函数,将问题转化为约束最大后验估计。为了解决高维优化问题,我们开发了一种先验约束梯度上升(PCGA)算法,该算法解耦相关参数并执行高效的梯度更新,以引导目标先验。理论分析涵盖了全局信噪比最大化下的最优融合权重、PCGA收敛性和估计器的克里弗-罗下界,这些见解适用于更广泛的融合方案。广泛的数值模拟和使用商用设备的现实世界实验表明,该框架将传输开销减少了90%(相比信号融合)并降低了41%的估计误差(相比参数融合)。值得注意的是,实地测试在典型毫米波AP覆盖范围内实现了95%的亚米精度。这些改进突显了通信效率和估计精度之间有利的平衡,适用于实际的多AP感知部署。数据集已发布供研究使用,并可在http://pmldatanet.com.cn/dataapp/multimodal上公开获取。

英文摘要:

Integrated sensing and communication is widely acknowledged as a foundational technology for next-generation mobile networks. Compared with monostatic sensing, multi-access point (AP) collaborative sensing endows mobile networks with broader, more accurate, and resilient sensing capabilities, which are critical for diverse location-based sectors. This paper focuses on collaborative sensing in multi-AP networks and proposes a Bayesian probability fusion framework for target parameter estimation using orthogonal frequency-division multiplexing waveform. The framework models multi-AP received signals as probability distributions to capture stochastic observations from channel noise and scattering coefficients. Prior information is then incorporated into the joint probability density function to cast the problem as a constrained maximum a posteriori estimation. To address the high-dimensional optimization, we develop a prior-constrained gradient ascent (PCGA) algorithm that decouples correlated parameters and performs efficient gradient updates guided by the target prior. Theoretical analysis covers optimal fusion weights for global signal-to-noise ratio maximization, PCGA convergence, and the Cramer-Rao lower bound of the estimator, with insights applicable to broader fusion schemes. Extensive numerical simulations and real-world experiments with commercial devices show the framework reduces transmission overhead by 90% versus signal fusion and lowers estimation error by 41% relative to parameter fusion. Notably, field tests achieve submeter accuracy with 50% probability in typical coverage of mmWave APs. These improvements highlight a favorable balance between communication efficiency and estimation accuracy for practical multi-AP sensing deployment. The dataset is released for research purposes and is publicly available at: http://pmldatanet.com.cn/dataapp/multimodal

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑