A formulation and validation of rail-bound navigation

Abstract

The advancement of Mobile Mapping Systems (MMS) has accelerated both commercial activities and academic research by efficient and flexible data acquisition. The Position and Orientation System (POS), which is a core component supporting direct georeferencing of mapping data, usually provides superior navigation performance by integrating the Inertial Navigation System (INS) and the Global Navigation Satellite System (GNSS). Its navigation accuracy, however, is primarily dependent on sensor quality and GNSS conditions, which leads to the necessity for additional data inspection and post-processing. Accordingly, the MMS are currently restricted to finite applications where GNSS is reliably available. In railway environments, mobile mapping technology has considerable potential for supporting railway safety and management in real-time, but the conventional POS encounters a challenge of accuracy loss in GNSS-denied areas, especially in underground railways. To minimise the impact of this problem, an alternative configuration is presented to replace the GNSS component of a POS, with that of the railway track alignment to ensure navigation and geo-referencing accuracy is constantly maintained. The concept of Rail-bound Navigation (RBN) is introduced in this thesis that directly substitutes the GNSS with the track constraints. Since a train is always bounded by the physical track under normal conditions, its relatively position and orientation are fundamentally constrained by the track alignment. With a valid rail-bound condition, the nominal position and orientation of train can be continuously determined, which provide alternative error control for a typical POS. In this thesis, a generalised Track Alignment Positioning (TAP) method is established to realise the concept of track constraints, while the RBN solution is formulated with the integration of INS and TAP and the discussion for corresponding practical issues. To validate the RBN concept, a prototype RBN system has been built with consumer grade Inertial Measurement Units (IMU) for conducting a number of model-based and real world experiments. Despite the absolute position errors caused by poor sensor quality, results have demonstrated a significant improvement in attitude and velocity in that the error accumulation has been greatly constrained without additional measurements and external control data. Through the analysis of repeated measurements, the potential performance of RBN have been illustrated.