大型室内复杂环境下基于智能手机定位的关键技术研究

智能手机室内定位是基于位置服务、物联网和人工智能应用的核心技术之一。现有智能手机室内定位算法在大型室内复杂环境中部署时存在环境普适性差、用户体验差和部署成本高的问题，难以实际推广应用。本课题拟突破现有问题的制约，以环境普适性强、低成本高精度为目标，应用深度学习理论，研究智能手机室内定位的关键理论和技术。应用半监督深度生成模型构建低成本高精度指纹库，应用半监督深度核学习模型实现指纹库的自适应更新，突破在大型室内定位中指纹库构建和维护的高成本制约；设计新型适用于室内定位的低成本无源WiFi标签，和手机IMU相结合设计低成本高精度室内定位算法；研究智能手机多传感器数据和定位技术的深度融合模型，提出多阶段融合的层次型定位架构和融合定位算法；搭建智能手机室内定位的实验平台，实现原型系统，在真实室内环境下对所提算法和技术进行实验验证。为在大型室内复杂环境部署智能手机定位系统提供理论和技术支撑。

Smartphone indoor localization is one of the core technologies based on location services, Internet of Things and artificial intelligence applications. However, the existing indoor localization algorithm of smart phones is difficult to be promoted in large scale indoor environment，because of the poor universality and user experience and high deployment cost. This project intends to break through the constraints of existing problems. With the goal of environmental universality, low cost and high precision, we will apply deep learning theory to study the key theories and technologies of indoor localization of smart phones. The semi-supervised depth generation model will be used to construct a low-cost and high-precision fingerprint database. The semi-supervised deep kernel learning model will be applied to realize the adaptive update of the fingerprint database, which breaks the high cost constraint of fingerprint construction and maintenance in large indoor environment localization. The low-cost passive WiFi tag and a low-cost and high-precision indoor localization algorithm which combined with the mobile phone IMU will be designed for indoor localization. We also will study for the deep fusion models of the multi-sensor data and localization technology. Then, based on the models, we will propose the multi-stage fusion hierarchical localization architectures and the fusion localization algorithms. The experimental platform for indoor localization of smart phones will be built, and the prototype system will be realized. The proposed algorithm and technology are experimentally verified in the real indoor environment. The completion of this project can provide theoretical and technical support for the deployment of smart phone localization systems in large indoor complex environments.