Annotation and Position Recall from Low Grade Sensorial Data in the Context of Topological Railway Maps

Abstract

Describing high quality track maps is an important prerequisite for the localization algorithms, which depend on relatively unreliable and unsafe triangulation technologies, such as the Global Navigation Satellite System solutions available on modern mobile devices. The state of the art studies and experiments concerning the maps annotation methodologies have been focused on using large quantities of crowdsourcing sensorial data and open source topological maps. The objective is to produce a low-cost platform for regional railway operators, to visualize the track's geometry and to generate a track representation, which can be then recalled across multiple journeys. The paper aims to develop and experiment with a sensorial signature representation toward providing a flexible solution for signal alignment based on Genetic Algorithms. The alignment solution intends to describe the sensorial signals of the track line in a relevant manner that makes localization algorithms possible based on the nonlinear state estimation.