Title: Active roll control using reinforcement learning for a single unit heavy vehicle

Authors: Maria Jesus Lopez Boada, Beatriz Lopez Boada, Antonio Gauchia Babe, Jose Antonio Calvo Ramos, Vicente Diaz Lopez

Addresses: Mechanical Engineering Department, University of Carlos III of Madrid. Avd. de la Universidad, 30, 28911, Leganes (Madrid), Spain. ' Mechanical Engineering Department, University of Carlos III of Madrid. Avd. de la Universidad, 30, 28911, Leganes (Madrid), Spain. ' Mechanical Engineering Department, University of Carlos III of Madrid. Avd. de la Universidad, 30, 28911, Leganes (Madrid), Spain. ' Mechanical Engineering Department, University of Carlos III of Madrid. Avd. de la Universidad, 30, 28911, Leganes (Madrid), Spain. ' Mechanical Engineering Department, University of Carlos III of Madrid. Avd. de la Universidad, 30, 28911, Leganes (Madrid), Spain

Abstract: In this paper, a reinforcement learning algorithm using neural networks to improve the roll stability in a single unit heavy vehicle is proposed. The controller, consisting of active anti-roll bars, provides the adequate roll moments to the vehicle to improve its performance. The main advantages of the implemented neuro control are its good performance to control non-linear systems, it is a free-model control and it learns online, so that the system can adapt to changes produced in the environment. In this case, it is only necessary to measure a unique variable (the sprung mass roll angle) to control the vehicle so both the number of sensors and vehicle cost are reduced. Simulation results show the effectiveness of the proposed control system during different manoeuvres such as J-turn and lane-change.

Keywords: single unit heavy vehicles; roll control; neural networks; reinforcement learning; roll stability; active anti-roll bars; nonlinear control; sprung mass roll angle; simulation.

DOI: 10.1504/IJHVS.2009.027413

International Journal of Heavy Vehicle Systems, 2009 Vol.16 No.4, pp.412 - 430

Published online: 22 Jul 2009 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article