面向多信息融合的自适应无线室内定位方法

Wireless indoor localization has become a research hotspot in recent years. Existing works usually aim at static environments, and they cannot achieve adaptive and precise positioning in complex and dynamic environments. This project investigates how to perform adaptive and highly precise wireless localization in complex and dynamic indoor environments from a multiple information fusion perspective. The project covers three topics. 1) Design of precise indoor localization methods based on multiple information fusion, including fingerprinting methods based on multi-information fusion and hybrid methods based on combination of fingerprinting and range-based localization. The target of fusing multiple information is to improve localization accuracy in complex and dynamic environments. 2) Design of localization information expansion mechanisms, which aim to generate more information for localization and improve accuracy, especially in environments where localization devices are very sparsely deployed and hence it is difficult to collect enough information for localization. Two approaches will be investigated, including new localization models considering effects of multi-path signals and fingerprinting data expansion based on matrix completion. 3) Design of cross environments precise localization algorithms, which adopt incremental learning to perform adaptive localization model updating and fingerprint mapping to perform fingerprinting updating. The target is to achieve adaptive and precise localization across different environments without recollection of fingerprints. The outputs of this project will help to realize high precision positioning and enhance the adaptivity in the complex indoor environments, which has important research meanings and practical values.

室内无线定位问题一直是近年来的研究热点。现有工作通常针对固定环境，在复杂变化的环境下难以实现自适应精确定位。本课题研究复杂室内环境下的自适应无线定位方法，从多信息融合、定位信息扩充和跨场景信息复用的角度研究如何提高定位算法的精度和适应性，包括：1）基于多信息融合的精确室内定位方法，包括基于多类型信息融合的指纹定位、基于测距定位和指纹定位相结合的方法等，以期解决复杂环境下基于单类型信息定位方法精度不高的问题；2）定位信息扩充机制和方法，包括多径信号增强机制以及基于矩阵填充的指纹数据扩充机制，以解决定位设备稀疏部署时难以获取到足够定位信息从而导致定位精度下降的问题；3）跨场景高精度定位算法设计，基于增量学习和指纹映射机制来进行定位模型更新和指纹动态更新，从而实现跨场景的高精度自适应定位。项目的研究成果将有助于在复杂室内环境下实现高精度、自适应的无线定位，具有重要的研究意义和应用价值。

智能感知问题一直是近年来的研究热点。现有工作通常针对固定环境，在复杂变化的环境下难以实现自适应精确感知。本课题研究了复杂环境下的自适应智能感知方法，包括：1）设计无线信号视距路径实时识别算法，以减少由非视距路径产生的测量误差，从而提升感知精度；2）跨场景高精度行人感知算法，设计基于分布对齐的正则化方法来解决行人再识别中的领域泛化的问题；3）通过引入基于注意力机制的句嵌入方法，使模型消除无关的伴随动作对识别的干扰；4）综述了深度学习在图像领域的工作。项目的研究成果将有助于在复杂环境下实现高精度、自适应的智能感知，具有重要的研究意义和应用价值。

内容获取失败，请点击重试