Development and Comparison of Algorithms to Derive Vehicle Location from Speed Profile Data

Abstract

The rapid increases in the quantity of data being gathered regarding technological systems such as railways can promote improvements in their design and operation. Combining information from different datasets allows more in-depth analysis, such as using train location data to enhance the analysis of speed profiles and energy consumption. Positioning systems such as GPS are frequently used to obtain this information, but are not necessarily always available, such as in underground metro systems. The focus of this paper is therefore the development of algorithms to derive train location information from measured speed profile data and network topology. Two different algorithms were developed to extract individual station-to-station journeys from an example consisting of a dataset of speed profiles and energy consumption from an urban rail system, and four classification algorithms were developed to identify the station pairs associated with each journey. It was found that the best-performing approach for this task was to compare the cumulative distance of a group of several consecutive journeys against a database of station-to-station distances to find the best match. This was more resilient than constructing sequences of consecutive journeys from possible matches in a database of station-to-station distances and orders of magnitude faster than heuristic algorithms.