A time-series approach for estimated time of arrival prediction in autonomous vehicles

Autonomous Vehicles are expected to play a pivotal role in the future of transportation. While huge accomplishments have been conducted in this field during the recent years, the main field of relevant research is primarly private transportation. Public transport (PT), especially in urban areas, although equally important, is lagging behind when the technological breakthroughs are considered. One of the main PT desiderata is punctual arrival. Therefore, this paper deals with the Estimated Time of Arrival (ETA) issue, tackled as a time-series problem, using contemporary Gradient Boosting (GB) methods, in order to benefit both commuters and stakeholders. The GB methods used are eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machines (LightGBM), and Categorical Boosting (CatBoost). This study proposes a competitive ETA methodology for Autonomous Shuttles, providing encouraging results both in the field of Research & Innovation for Autonomous Vehicles and Public Transport, being able to be expanded for commercial purposes.