- 3549 - Development of Ecological Driving System Using Model Predictive Control M.A.S. Kamal 1 , M. Mukai 2 , J. Murata 2 and T. Kawabe 2 1 Fukuoka Industry, Science and Technology Foundation, Fukuoka, Japan (Tel: +81-92-802-3691; E-mail: maskamal@ieee.org) 2 Department of Electrical and Electronic Systems Engineering, Kyushu University, Fukuoka, Japan (E-mail: {mukai, murata, kawabe}@ees.kyushu-u.ac.jp) Abstract: Ecological driving aiming at optimizing energy consumption is highly desirable for sustainable intelligent transportation systems. This paper presents a unique development of ecological vehicle driving system in model predictive approach. The vehicle’s fuel consumption model and the model based anticipation of future road-traffic situations are used in this rigorous reasoning approach of deriving the control input. A combination of Continuation and Generalized Minimum Residual Methods is used to optimize the sequence of vehicle control actions required in the prediction horizon aiming long run fuel economy while maintaining a safe driving. Performance of the proposed system is evaluated through simulations in AIMSUN NG microscopic transport simulator. The driving behavior with fuel saving aspects is graphically illustrated, compared and analyzed to signify the achievement of the developed system. Keywords: Ecological driving, model predictive control. 1. INTRODUCTION Maneuvering a vehicle in a changing road-traffic envi- ronment is a very complex task. Human driver can cope with such changing situations utilizing his complex and experience-based reasoning and attain his own typical be- havior in cruising or following a car. A car following or cruising behavior describes processes by which drivers follow each other in the traffic stream, and various mod- els have been proposed to approximate driving behavior since its research began about sixty years back [1]. The driving behavior, in which the vital issues are the speed and range clearance patterns, varies widely among the drivers, and a single behavior model neither fully repre- sents all the driving situations nor matches every driver [2, 3]. Ecological driving means controlling a vehicle in any traffic streams and road situations in such a way so that the fuel consumption in long run is minimum. Obvi- ously, the ecological driving behavior is much more com- plex since it is necessary to forecast the road-traffic situ- ation ahead. Besides many physical factors, driving styles have a great influence on vehicle emissions and energy con- sumption [4]. Proper driving styles may improve the travel economy or driving efficiency considerably. A re- cent experiment through ecological-driving contest con- ducted on urban roadway shows a reduction of fuel con- sumption by 25%, which yields an improvement in fuel economy (km/l) by 35% [5]. Generally, fuel economy is maximized when acceleration and braking are mini- mized. Therefore, a fuel-efficient or ecological strategy is to anticipate what is happening ahead, and drive in such a way so that it minimizes acceleration and braking, cruises at the optimal speed and maximizes coasting time at stops. Since the changing nature of the road-traffic af- fects the driving behavior, a simplified model that does not anticipate situations ahead cannot represent ecologi- cal driving behavior completely. Various speculative for- mulations of desired driving behavior for ecological driv- ing are well known. The existing ecological driving tips or assistance are a bit superficial and based on rough ve- hicle engine characteristics [6]. Some recent efforts uses optimal control approach in which only the model of the engine, in terms of speed, gear ratio and load, are con- sidered [7]. They do not have any rigorous reasoning by analyzing the instant road-traffic situation and its trends in future. Model predictive control is a potential control tech- nique for non-linear systems that suits vehicle driving. This paper presents a novel and unique approach of de- veloping an ecological driving system using model pre- dictive control that measures how the current situations may affect the fuel consumption in long run and deter- mines the optimal control input based on maximization of traveling distance per unit fuel consumption. This rig- orous approach of deriving optimal control input would make the proposed system more efficient, reliable and trustworthy. The system senses the status of the subjec- tive host and its surrounding vehicles, and the traffic sig- nal ahead through information technology. It anticipates the possible behavior of the surrounding vehicles in near future using some sort of simplified model. Based on this information and fuel consumption model of the vehicle, it generates an immediate control action that may lead to a long-range economy travel. It is expected that new sens- ing and communication technology be incorporated for the development of road-traffic infrastructure in coming decades. Through this advancement in intelligent trans- portation systems, the proposed ecological-driving can be realized. The model of the vehicle control system, used in this paper, expresses the dynamic relationship of the host ve- hicle, preceding vehicle and traffic signal system at any instant. The optimization of control inputs in the pre- diction horizon is conducted using Continuation method combined with generalized minimum residual method known as C/GMRES method [8, 9]. From the viewpoint ICROS-SICE International Joint Conference 2009 August 18-21, 2009, Fukuoka International Congress Center, Japan PR0002/09/0000-3549 ¥400 © 2009 SICE