The high cost of maintaining radio map is a major hurdle for wide application of WLAN fingerprint-based indoor localization. The development of mobile crowd sensing provides new possibilities, however the features of normal users such as moving freely and non -professional bring new challenges. In this paper, a radio map self-updating algorithm is proposed to resolve three key problems: the localization accuracy, determination of fingerprints need to be updated, and capture of new fingerprints. First we design the localization matrix mechanism and periodic adaptive estimate algorithm to ensure the localization accuracy. Second we propose the fingerprint integrity assessment algorithm to detect the access points changed and the periodic adaptive estimate algorithm to decide the update period for each reference point. Finally we design the active fingerprint collecting mode to update the radio map efficiently. The algorithm proposed has been deployed for real-world testing over 30 days, our studies show that it detects the network changes in indoor environment correctly in 98% cases, and automatically judges the localization accuracy in 95% cases. Meanwhile, the localization accuracy is stable and improved by over 40% even after long terms of deployment, and the overhead of user terminals is reduced over 40%.

无线电地图的高昂维护成本是WLAN指纹室内定位广泛应用的主要障碍。移动人群感知的发展提供了新的可能性，但普通用户自由移动、非专业等特点带来了新的挑战。本文提出了一种无线电地图自更新算法来解决三个关键问题：定位精度、需要更新指纹的确定和新指纹的捕获。首先我们设计了定位矩阵机制和周期性自适应估计算法来保证定位精度。其次，我们提出指纹完整性评估算法来检测改变的接入点，并提出周期性自适应估计算法来决定每个参考点的更新周期。最后我们设计了主动指纹采集模式来有效地更新无线电地图。所提出的算法已部署用于实际测试超过 30 天，我们的研究表明，它在 98% 的情况下正确检测了室内环境中的网络变化，并在 95% 的情况下自动判断了定位精度。同时，即使经过长期部署，定位精度也保持稳定，提升40%以上，用户终端开销降低40%以上。