一种基于多维标度和区域细化的无线室内定位方法

An Indoor Localization Algorithm Based on Multidimensional Scaling and Region Refinement

移动用户的位置信息成为许多基于位置应用的内在驱动力.基于指纹的WLAN定位方法因在定位精度、定位复杂性以及定位成本等方面的优势,近年来成为室内定位研究领域的热点.然而,由于需要在训练阶段采集大量的参考点指纹信息,这类方法在大规模环境下的应用受到了很大的限制.该文提出一种基于多维标度(Multidimensional scaling,MDS)技术的室内定位算法,能够有效减少训练阶段指纹采集的开销.该方法利用不同位置之间的信号强度差异来表征对应的物理空间距离,并以此为基础利用多维标度技术计算目标对象的位置.提出一种区域细化的方法来进一步提高定位精度,在获取目标对象的近似位置后,通过缩小目标对象所在的可能区域并利用离其最近的参考点来对其位置进行迭代计算,有效提高了定位精度.在多个实际场景中的测试结果表明,相比于已有工作,该文所提出的算法有效降低了所需的参考点个数,并达到了与已有算法类似甚至更高的定位精度.

The position information of mobile users becomes the internal motivation to many locations based applications,such as indoor navigation,travel application and mobile advertising service.Fingerprinting-based indoor localization has been a hot topic in localization research in recent years due to its high accuracy,low complexity and low cost.This method,however,requires collecting a large number of reference point information in the offline training phase,which greatly limits its application in large environments.This paper proposes a new indoor localization algorithm based on multidimensional scaling（MDS）,which utilizes the distance between two points＇ received signal strength vectors to approximately measure their distance in physical environment,and then calculates the target object＇s location with the distance constraints with MDS.We further propose a region refinement method to improve the localization accuracy,in which only reference points contained in a small rectangle that contains the estimated position of the target object are used to calculate the target＇s location.Experiment results in two real environments and large scale simulations demonstrate the usefulness of our approach.The results show that,compared with the Horus algorithm,the proposed algorithm can achieve similar or even higher accuracy and reduce the number of required reference points by almost one order of magnitude.The effectiveness of the proposed algorithm is validated via large-scale simulation.