FUEL EFFICIENT HIGH-DENSITY PLATOONING USING FUTURE CONDITIONS PREDICTION USING MACHINE LEARNING

Abstract

A promising application of cooperative driving is high density platooning, which main 
goal is to reduce fuel consumption by driving with inter-vehicle distances below ten meters. The 
prediction of factors influencing the platoon capability to drive with such inter-vehicle distances the 
derived safe inter-vehicle distances, drives the potential fuel saving. Our aim is to study the influence 
of the prediction, especially the prediction horizon, on the achieved fuel saving as a function of 
different maneuver parameters. The contributions of this paper are: introducing the concept of 
maneuver reference to distribute the effort of maneuvering in truck platooning; linking the fuel 
consumption to a compensation time, that is the time during which the platoon will counter-balance 
the fuel consumption by benefiting from the reduced air drag; presenting an optimization method for 
maximizing the fuel saving depending on some predictive quality of service parameters. To model the 
fuel consumption and the duration of the maneuvers, we use a lasso regression on data obtained from 
simulation. We then use these regression models in our optimization framework, which is based on 
particle swarm optimization. We show that to benefit from high-density platooning, the magnitude 
order of the prediction horizon required by a five -truck platoon is minimum hundred seconds.