Data-Driven Leading Vehicle Speed Forecast and Its Application to Ecological Predictive Cruise Control

Abstract

In this article, we propose an ecological predictive cruise control method for connected and automated vehicles (CAVs) using data-driven predicted leading vehicle speed in a car-following scenario. Many existing studies assume that the leading vehicle's behavior is known when planning the trajectory of the ego vehicle. Unfortunately, predicting the future behavior of adjacent vehicles is extremely uncertain and inaccurate for use in control. To overcome this, we adopt the vector autoregressive model (VAR) that is well suited for generating simultaneous forecasts of the response variables when predicting the short-term behavior of the vehicle ahead. As many human drivers behave similarly in a car-following situation, vehicle connectivity is specifically used to drop hints of the behaviors of the cars following the connected car. Once the leading vehicle's future trajectory is predicted, the ego vehicle is controlled in a way that minimizes energy consumption by optimizing its speed trajectory while remaining safe. Through simulation case studies, we demonstrate that our approach can achieve improved energy efficiency considerably than the conventional strategy that cannot predict the speed of vehicles ahead. Given the recent market penetration of vehicle connectivity technologies and advanced driving assistance systems (ADAS), the proposed method is expected to have a high commercialization potential.