Global and quadrantal kernel information for mobile localization

To deal with the non-line of sight effect on mobile localization, the kernel information of surroundings, which can be extracted by supervised learning, is exploited to correct the location bias. The proposed kernel-based location method takes advantage of the dimension augment of feature space, which is produced by a nonlinear mapping on the original measurements and the ordinary location estimate. In the sense of minimum training error, the kernel matrices are learned from a set of training data provided by the advanced mobiles, which can locate themselves more accurately and coexist with the old mobile stations. Furthermore, quadrant information is used to conform the kernel information to different parts of the surrounding. Simulation results illustrate the applicability of global and quadrantal kernel information.