A Probability-Based Short-Term Velocity Prediction Method for Energy-Efficient Cruise Control

Abstract

Numerous automotive control systems depend on the prediction of the velocity of the controlled vehicle or other road users in close proximity. A forecast is indispensable for the design and operation of model predictive control strategies which are utilized for example in order to optimize fuel economy in cruise control systems. In this article, a novel velocity time series prediction method, which depends exclusively on past velocity measurements, is presented and analyzed. It is based on conditional probabilities of acceleration and deceleration, which are estimated from real driving data. The proposed velocity prediction is used in a new model predictive control algorithm for an adaptive cruise control system for heavy duty vehicles, where the predicted future velocity of a preceding vehicle is needed, as well as its probability distribution. Simulation results show that an accurate velocity prediction is crucial for achieving energy-efficiency goals.