CIMLoc: 面向定位的众包室内地图构建系统

CIMLoc: 面向定位的众包室内地图构建系统

CIMLoc: A crowdsourcing indoor digital map construction system for localization

 

       室内地图作为许多室内定位导航系统的重要基础,在有些室内环境中有时很难获得。由于缺少室内地图数据以及手动构建室内地图的成本较高,因此急需一种廉价、高效的自动构建室内地图的方法。随着具有多传感器的移动设备的普及,我们可以低成本的通过众包用户轨迹自动的构建室内地图。与其他众包数据一样,搜集的用户轨迹往往包含噪声和低保真度(low fidelity),这对构建精确的地图带了了很大挑战。为了解决这个问题,作者提出了一种CIMLoc,即面向定位的众包室内地图系统。通过利用从不同移动设备搜集的真实轨迹数据对这个系统进行了评估。通过真实地图与所构建的地图来计算定位结果误差,进而量化构建地图的误差。实验结果表明,CIMLoc能够构建出精确的地图,可以有效的提高定位结果。CIMLoc能够在室内地图不可用的环境中提供一种有效的室内定位解决方案。

 

Abstract—Indoor maps, as crucial prerequisites for many indoor localization and navigation systems, are sometimes inaccessible. The absence of an indoor map database and the high cost of manually constructing an indoor map produce a need for an inexpensive and efficient way to dynamically construct indoor maps. The ubiquity of sensor-equipped mobile devices enables us to crowdsource user trajectories, out of which indoor digital maps can be automatically constructed at low costs. Similar to other crowdsourced data, the collected user trajectories are often noisy and of low fidelity, which poses a challenge to the accurate map construction. To alleviate this problem, we propose CIMLoc – a crowdsourcing indoor map construction system for localization. The system is evaluated with real-world trajectories collected from different mobile devices. We quantify the construction errors by computing the localization errors achieved with the constructed map and the real map. Experimental results reveal that CIMLoc is able to construct accurate maps that significantly improve localization results. We believe that CIMLoc provides an effective solution to the indoor localization problems where the indoor maps are unavailable.

System Architecture

 

An Example of the Indoor Map Construction Procedure

 

Zhang X, Jin Y, Tan H X, et al. CIMLoc: A crowdsourcing indoor digital map construction system for localization[C]// 2014 IEEE Ninth International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP). IEEE, 2014:1-6.