Indoor localization through trajectory tracking using neural networks

Abstract

Currently deployed wireless and cellular positioning techniques are optimized for outdoor operation and cannot provide highly accurate location information in indoor environments. Meanwhile, new applications and services for mobile devices, including the recent Enhanced 911 (E911), require accurate indoor location information up to the room/suite level. In this work, a new system for improving indoor localization of mobile users is presented by exploiting trajectory tracking techniques using neural networks (NNs). The motion trajectories of indoor mobile users are tracked using conventional positioning algorithms, then a NN is applied to identify the current room location of a mobile user based on the tracked motion trajectory. Simulation results show that the trajectory-based NN is able to provide indoor location information at the room level with much higher accuracy in different scenarios, with an enhancement of up to 49% in correct room identification, as compared to positioning techniques based only on a single-point location estimate. In addition, miss-classification of the NN system will result in selecting one of the immediate neighboring rooms instead with at least 30% probability.