RSS difference-aware graph-based semi-supervised learning (RG-SSL) RSS smoothing method for crowdsourcing indoor localization

In order to realize the rapid deployment of indoor localization systems, the crowdsourcing method has been proposed to reduce the collection workload. However, compared to conventional methods, the reduced number of received signal strength (RSS) values lends greater influence to noises and erroneous measurements in RSS values. In this paper, a graph-based semi-supervised learning (G-SSL) method is used to exploit the correlation of RSS values at nearby locations to infer an optimal RSS value at each location in terms of error. The RSS difference between different locations is used as a part of cost function to improve the performance of G-SSL. Experimental results show that the proposed method results in a smoother radio map and improved localization accuracy.