基于迁移学习的室内动态环境定位算法

Indoor dynamic environment localization algorithm based on transfer learning

传统室内指纹定位系统的精度受指纹库中参考位置节点的密度和室内环境特征等多方面因素的制约。室内环境动态变化时RSS波动较大,通常不满足同分布的假设条件,故传统指纹定位方法难以满足高精度需求。针对室内环境动态变化导致传统算法无法精准定位问题,设计并实现了一种基于室内指纹库的迁移学习动态环境定位算法,该算法采用迁移学习的思想把不同分布的数据集嵌入对齐到潜在特征空间中,从而有效缓解了环境动态变化对系统造成的不利影响。本文算法实验数据均来自于真实的环境,通过仿真得到该算法的平均定位误差是1.23 m。

The accuracy of the traditional indoor fingerprint localization system is limited by many factors, such as the density of the reference location node in the fingerprint database and the characteristics of the indoor environment. When the indoor environment changes dynamically, the RSS fluctuates, and usually does not meet the assumption of the same distribution. Therefore, it was difficult to obtain high-precision requirements for conventional fingerprint positioning method. Aiming at the problem that the traditional algorithm couldn＇t locate accurately, an algorithm based on the indoor fingerprint database was designed and implemented. The algorithm adopted the idea of migration learning to embed different data sets into the latent feature space, and the adverse effects of environmental changes on the system were mitigated. The simulation results show that the average positioning error of this algorithm is 1.23 m.