A lightweight ANN based robust localization technique for rapid deployment of autonomous systems

Abstract

The capability to localize or identify position in the field of deployment is a primary requirement of future autonomous system in domains such as warehouse transportation, ambient-assisted living/ health care systems, search and rescue, motion monitoring, etc. Although reliable indoor localization in the order of few centimeters can be achieved with the existing localization systems in Line-of-Sight (LOS) conditions, the localization under Non-line-of-Sight (NLOS) conditions is an open area of research. In range-based localization systems, distance estimation is a pre-requisite for location estimation. Time of Arrival (ToA) is considered to be the most accurate technique for distance estimation when compared to Time Difference of Arrival (TDoA) or Received Signal Strength Indication (RSSI). Most of the work available as literature on indoor localization under NLOS conditions is based on the profiling of the indoor deployment area under various NLOS conditions and mitigating NLOS affected timestamps from the ToA measurements. However, it is not practically possible to obtain a comprehensive data set containing all possible conditions of NLOS in indoor environments. In this paper, an Artificial Neural Network based Location Estimation Unit (ANN-LEU) based scheme is proposed to estimate the two-dimensional (2-D) location of an object under LOS and NLOS conditions. One of the unique features of the novel location estimation scheme is that the training of the system is required to be performed only under LOS conditions, thus facilitating the quick deployment in new environments. The proposed ANN-LEU is robust as it identifies the presence of NLOS if any, in the ToA measurements and thus removing false position estimations if any. The Mean Average Error (MAE) error in position estimated during the performance analysis of the proposed system was restricted to lesser than 20 cm, if the object is in range of three beacons in LOS, and also for the scenarios in which one of the three beacon nodes are in NLOS. The proposed scheme eliminates false position identification. The proposed scheme requires lesser number of beacons for localization when compared to the available indoor localization systems, thus also improving the cost and energy efficiency.