Prototype of a System for Tracking Transit Service Based on IoV, ITS, and Machine Learning

Abstract

The transit service in a city should be the most efficient, least polluting, most accessible, and sustainable means of transportation for its citizens. However, serious shortcomings have been detected, mainly in medium-sized cities in developing countries. These shortcomings are related to a lack of user information, insecurity, low service availability, and repeated stops in inappropriate and/or unauthorized places. Some of these shortcomings contribute to high accident rates and traffic congestion. The development of tools to improve the characteristics and conditions of transit service in cities has become an imperative need to improve the quality of life of citizens and city sustainability. Transit service tracking is relevant in aspects such as online location information to travelers and control by transport companies for compliance with speed limits, schedules, routes, and stops. This research proposes a transit vehicle tracking system based on the Internet of Vehicles (IoV) in Vehicle-to-Roadside (V2R) classification. The proposed system is ideal for the use of electric vehicles due to the low power consumption of the tracking device. This system uses Intelligent Transportation Systems (ITS) tracking service architecture, Long Range (LoRa) communication technology, and its LoRa Wide Area Network (LoRaWAN) protocol. Additionally, the system offers real-time location prediction in the absence of position data. The IoV tracking device integrates a GPS-LoRa module card with an Inertial Measurement Unit (IMU). A location prediction algorithm was implemented to train and store a prediction model with previously collected data from tracking devices. To evaluate the developed model, a case study in the city of Popayán (Colombia) was implemented, using three routes for testing. The results of the system implementation were satisfactory, obtaining an average coverage of 60.4% of the routes in the final field tests through LoRa communication. For the remaining 39.6% of the routes, location data prediction was used, with an average accuracy of 177 m with respect to the real location. Considering the obtained results, a tracking system such as the one proposed in this article can be used in the transit systems of medium-sized cities in developing countries to improve service quality and fleet control.

Keywords

• transit service
• transit vehicle tracking
• location prediction
• LoRaWAN
• machine learning