Localizing an unknown number of targets with radio tomography networks

Radio tomography network (RTN) is an emerging technology which deploys wireless sensor nodes around an area of interest and uses variation of received signal strength (RSS) to localize objects in that area. Existing localization methods for RTN all require a priori knowledge of the number of objects to be localized. This paper presents a novel localization method which can localize an unknown number of objects with RTN using RSS measurements at a single time step. Moreover, the method can adapt to different environments without changing parameters. First, we utilize the kernel density estimation (KDE) to approximate the RSS distribution of a link and determine whether the link is affected or not. Then, a localization strategy is proposed to detect all possible target regions in the network area, and a clustering method is applied to give the final decision. Experiments with a 16-node RTN deployed around a 16 square meters area in an office room show that both number and positions of multiple targets can be accurately estimated.