基于位置指纹技术的WiFi室内定位是未来最具广泛应用潜力的室内定位导航系统。然而，目前困扰其普及的最大技术瓶颈就是Radio Map建立的方式耗时费力，不利于广泛推广。本项目提出基于群智信息感知模式来构建Radio Map，基本思想是在不影响用户正常使用的前提下利用智能移动终端各种传感设备(WiFi网卡和惯导器件)在后台感知周围环境，并将感知信息上传给服务器最终形成Radio Map，从而在保证定位精度的前提下大大降低Radio Map构建成本。本项目交叉融合机器学习理论、凸优化理论、流形学习理论、模式识别以及控制学科等各领域知识，主要解决群智信息感知模式下构建Radio Map过程中出现的各种问题，包括终端多样性测量数据融合、信号空间填充与恢复、位置空间特征点的最佳匹配问题以及稀疏位置空间的估计与预测。本项目的研究将为未来WiFi室内定位系统的推广提供值得借鉴的有效方法。

WiFi indoor positioning based on fingerprint technology is the most potential indoor positioning and navigation systems in the future. However, its popularity is currently troubled by the technical bottleneck which is Radio Map is constructed in a time-consuming way so that it is not conducive to be widely used. Our project proposes a method to construct the Radio Map based on crowdsourcing. The basic idea is the Radio Map is constructed by large amounts of mobile devices in the background under the premise that no interference happens. By sensing the environment with the help of various sensing devices (WiFi card and inertial hardware), the mobile devices upload the information to the sever which is responsible for Radio Map construction. This will ensure the localization accuracy and at mean time greatly reduce the cost of Radio Map construction. Our project integrates alternately the machine learning theory, convex optimization theory, manifold learning theory, pattern recognition and control theory, aiming to resolve the problems appear in the Radio Map construction in the way of crowdsourcing, such as signal integration for mobile diversity, signal space completion and recovery, feature points optimization matching in location space and sparse location estimation and prediction. This research project will contribute effective ways to the future WiFi indoor localization.