Against the backdrop of rapid development in China’s construction and infrastructure sectors,discrepancies between project budgets and actual costs have become pronounced,manifesting in project overruns and suspensio...Against the backdrop of rapid development in China’s construction and infrastructure sectors,discrepancies between project budgets and actual costs have become pronounced,manifesting in project overruns and suspensions,posing significant challenges.To address inaccuracies in investment targets and operational complexities,this study focuses on a beam-bridge construction project in a district of Shijiazhuang city as a case study.Drawing upon historical analogs,the project employs a Work Breakdown Structure(WBS)to decompose the engineering works.Building on theories of Cost Significant(CS)and Whole Life Costing(WLC),the study constructs Cost Significant Items(CSIs)and develops a CNN-BiLSTM-Attention neural network for nonlinear prediction.By identifying significant cost drivers in engineering projects,this paper presents a streamlined cost estimation method that significantly reduces computational burdens,simplifies data collection processes,and optimizes data analysis and forecasting,thereby enhancing prediction accuracy.Finally,validation with real-world cost fluctuation data demonstrates minor errors,meeting predictive requirements across project execution phases.展开更多
In massive multiple-input multiple-output(MIMO)systems utilizing frequency division duplexing,optimizing system performance requires user equipment(UE)to compress downlink channel state information(CSI)and transmit it...In massive multiple-input multiple-output(MIMO)systems utilizing frequency division duplexing,optimizing system performance requires user equipment(UE)to compress downlink channel state information(CSI)and transmit it to the base station(BS).As the number of antennas increases,there is a significant rise in the overhead related to CSI feedback,posing considerable challenges to the precise acquisition of CSI by the BS.Existing approaches to CSI feedback utilizing deep learning techniques face challenges such as significant feedback overhead and limited precision in the reconstruction process.This study presents a novel lightweight CSI feedback framework known as the dual attention neural network(DANet).Within the DANet architecture,a dual attention module(DAM)is designed to enhance the network's performance.This DAM includes both channel attention blocks and spatial attention blocks.The channel attention blocks direct the model's focus toward channel features rich in information content while simultaneously suppressing less significant features.This approach enables the extraction of temporal correlations within the CSI matrix.The spatial attention block aids in extracting the correlation between the delay domain and the angle domain in the CSI matrix.By enhancing neural network performance,the DAM reduces information dispersion while enhancing the representation of global interactions.Simulation results demonstrate that DANet exhibits superior normalized mean square error and cosine similarity with comparable complexity compared to existing advanced CSI feedback methods.展开更多
With the growing advancement of wireless communication technologies,WiFi-based human sensing has gained increasing attention as a non-intrusive and device-free solution.Among the available signal types,Channel State I...With the growing advancement of wireless communication technologies,WiFi-based human sensing has gained increasing attention as a non-intrusive and device-free solution.Among the available signal types,Channel State Information(CSI)offers fine-grained temporal,frequency,and spatial insights into multipath propagation,making it a crucial data source for human-centric sensing.Recently,the integration of deep learning has significantly improved the robustness and automation of feature extraction from CSI in complex environments.This paper provides a comprehensive review of deep learning-enhanced human sensing based on CSI.We first outline mainstream CSI acquisition tools and their hardware specifications,then provide a detailed discussion of preprocessing methods such as denoising,time–frequency transformation,data segmentation,and augmentation.Subsequently,we categorize deep learning approaches according to sensing tasks—namely detection,localization,and recognition—and highlight representative models across application scenarios.Finally,we examine key challenges including domain generalization,multi-user interference,and limited data availability,and we propose future research directions involving lightweight model deployment,multimodal data fusion,and semantic-level sensing.展开更多
通过压缩信道状态信息(Channel Status Information,CSI)传输码字降低大规模多输入多输出(Multiple-Input Multiple-Output,MIMO)系统的CSI反馈开销,可以有效减少计算资源的使用和信息传输时间的消耗。针对如何使用轻量化模型准确估计...通过压缩信道状态信息(Channel Status Information,CSI)传输码字降低大规模多输入多输出(Multiple-Input Multiple-Output,MIMO)系统的CSI反馈开销,可以有效减少计算资源的使用和信息传输时间的消耗。针对如何使用轻量化模型准确估计低压缩比条件下CSI反馈的问题,通过设计的轻量化迭代交叉网络(Iterative Cross Network,ICNet)模型,在用户端使用设计的迭代压缩模块压缩CSI反馈,基站端使用设计的迭代重建模块估计CSI反馈,以较高的准确率和较低的时间消耗估计了低压缩比条件下的CSI反馈。在COST2100模型生成的数据样本下评估了ICNet在低压缩比条件下的鲁棒性,实验表明,在较小的1/64压缩比条件下,ICNet的归一化均方误差比次优值降低了8.48%,ICNet的参数量降低了35%左右。展开更多
文摘Against the backdrop of rapid development in China’s construction and infrastructure sectors,discrepancies between project budgets and actual costs have become pronounced,manifesting in project overruns and suspensions,posing significant challenges.To address inaccuracies in investment targets and operational complexities,this study focuses on a beam-bridge construction project in a district of Shijiazhuang city as a case study.Drawing upon historical analogs,the project employs a Work Breakdown Structure(WBS)to decompose the engineering works.Building on theories of Cost Significant(CS)and Whole Life Costing(WLC),the study constructs Cost Significant Items(CSIs)and develops a CNN-BiLSTM-Attention neural network for nonlinear prediction.By identifying significant cost drivers in engineering projects,this paper presents a streamlined cost estimation method that significantly reduces computational burdens,simplifies data collection processes,and optimizes data analysis and forecasting,thereby enhancing prediction accuracy.Finally,validation with real-world cost fluctuation data demonstrates minor errors,meeting predictive requirements across project execution phases.
基金National Natural Science Foundation of China(12005108)。
文摘In massive multiple-input multiple-output(MIMO)systems utilizing frequency division duplexing,optimizing system performance requires user equipment(UE)to compress downlink channel state information(CSI)and transmit it to the base station(BS).As the number of antennas increases,there is a significant rise in the overhead related to CSI feedback,posing considerable challenges to the precise acquisition of CSI by the BS.Existing approaches to CSI feedback utilizing deep learning techniques face challenges such as significant feedback overhead and limited precision in the reconstruction process.This study presents a novel lightweight CSI feedback framework known as the dual attention neural network(DANet).Within the DANet architecture,a dual attention module(DAM)is designed to enhance the network's performance.This DAM includes both channel attention blocks and spatial attention blocks.The channel attention blocks direct the model's focus toward channel features rich in information content while simultaneously suppressing less significant features.This approach enables the extraction of temporal correlations within the CSI matrix.The spatial attention block aids in extracting the correlation between the delay domain and the angle domain in the CSI matrix.By enhancing neural network performance,the DAM reduces information dispersion while enhancing the representation of global interactions.Simulation results demonstrate that DANet exhibits superior normalized mean square error and cosine similarity with comparable complexity compared to existing advanced CSI feedback methods.
基金supported by National Natural Science Foundation of China(NSFC)under grant U23A20310.
文摘With the growing advancement of wireless communication technologies,WiFi-based human sensing has gained increasing attention as a non-intrusive and device-free solution.Among the available signal types,Channel State Information(CSI)offers fine-grained temporal,frequency,and spatial insights into multipath propagation,making it a crucial data source for human-centric sensing.Recently,the integration of deep learning has significantly improved the robustness and automation of feature extraction from CSI in complex environments.This paper provides a comprehensive review of deep learning-enhanced human sensing based on CSI.We first outline mainstream CSI acquisition tools and their hardware specifications,then provide a detailed discussion of preprocessing methods such as denoising,time–frequency transformation,data segmentation,and augmentation.Subsequently,we categorize deep learning approaches according to sensing tasks—namely detection,localization,and recognition—and highlight representative models across application scenarios.Finally,we examine key challenges including domain generalization,multi-user interference,and limited data availability,and we propose future research directions involving lightweight model deployment,multimodal data fusion,and semantic-level sensing.
