Exploiting a real-time non-geolocation data to classify a road type with different altitudes for strengthening accuracy in navigation

Abstract

Most location-based applications for navigation purposes use geolocation data, i.e., a pair of a latitude and a longitude, to determine a real-time location of a handheld device (e.g., smartphones or tablets) that runs the applications. This can be implemented basically by requesting a pair of a latitude and a longitude from the device’s sensor that receives geolocation data from satellites. However, telling a device’s location by GPS sensor is sometimes impractical, especially when the device is in a vehicle on a road that shares exactly the same geolocation with other roads. Particularly, this is a scenario that there is a groundlevel road along with another elevated road (e.g., a turnpike) which is very common in cities like Bangkok, Singapore, or Hong Kong. The geolocation data yield no clue whether or not a vehicle is running on a ground-level road. Since a pair of a latitude and a longitude can no longer be used in such scenario, we proposed a methodology to identify the correct location of both a device and a vehicle without any involvement of geolocation data by using a Random Forest classifier and realtime traffic data that are able to be captured by a handheld device as training features to train a classification model. A completed experiment and results after testing the model were reported in this article.