随着万物互联时代的到来，物联网中海量数据的融合面临着准确性低与时效性差等巨大挑战。本项目针对物联网资源受限的特点，以实现准确且及时的数据融合为目标，提出“可重构智能表面赋能数据融合”的创新思路，为物联网数据融合建立完整的信道估计、资源优化、以及动态决策理论框架。研究内容主要包括：1）利用可重构智能表面关联信道的稀疏特性，提出低开销的串联信道估计方案，并设计基于循环神经网络的快速信道估计算法；2）以提升数据融合的准确性为目标，提出基于可重构智能表面和空中计算的数据融合方案，并设计基于凸差规划的高性能算法和基于黎曼流形的低复杂度算法；3）以提升数据融合的及时性为目标，充分利用可重构智能表面的能效优势，提出基于信息年龄的数据融合方案，并设计基于深度强化学习的动态决策机制。本项目旨在设计高效的信道估计和数据融合方案与算法，大幅提升物联网数据的利用价值，并推动物联网领域的理论创新和技术突破。

With the advent of Internet of everything, the aggregation of massive data in Internet of things (IoT) networks is facing significant challenges in terms of accuracy and timeliness. To achieve accurate and timely wireless data aggregation (WDA), this research proposes the novel idea of reconfigurable intelligent surface (RIS) empowered WDA, and constructs a complete theoretical framework of channel estimation, resource optimization, and dynamic decision making, while taking into account the resource constraints of IoT networks. This research mainly focuses on the following three research problems. First, by utilizing the sparsity of RIS-related wireless channels, we propose a low-cost cascaded channel estimation scheme, and develop an efficient recurrent neuron network (RNN) based fast channel estimation algorithm. Second, to improve the accuracy of WDA, we propose an RIS and over-the-air computation (AirComp) based WDA scheme, and develop two resource optimization algorithms based on difference-of-convex programming and Riemannian manifold, respectively. Third, to enhance the timeliness of WDA, we propose an age of information (AoI) based WDA scheme by taking advantage of the energy efficiency of RIS, and develop a deep neuron network (DNN) based dynamic decision-making mechanism. This research aims to design efficient schemes and algorithms for channel estimation and WDA, thereby significantly enhancing the value of IoT data and promoting the theoretical innovation and technical breakthrough for IoT.