THE IMPACT OF ECOROLL ON FUEL CONSUMPTION - USING LOOK AHEAD

(1)

THE IMPACT OF ECOROLL ON FUEL

CONSUMPTION - USING LOOK AHEAD

MUSTAFA ABDUL-RASOOL

Master’s Degree Project

Stockholm, Sweden June 2011

(2)

(3)

THE IMPACT OF ECOROLL ON FUEL

CONSUMPTION - USING LOOK AHEAD

MUSTAFA ABDUL-RASOOL

Master’s Thesis at Automatic Control Supervisor: Oskar Johansson

Examiner: Ather Gattami

(4)

(5)

Abstract

EcoRoll reduces fuel consumption with small development costs, since no additional hardware is required. It is a func-tion that enables a more efficient conversion of potential to kinetic energy, when travelling downhill. This is achieved by opening the powertrain, and let the engine run on idle to reduce engine losses. In this Master’s thesis, two control strategies were developed, where one is based on prevailing conditions and one utilizes Look-Ahead data. Compared to a vehicle with a conventional cruise control, the first strat-egy gave a fuel reduction of approximately 3.4% and the other 3.7%. This was simulated on the highway between Södertälje and Norrköping in Sweden.

(6)

Referat

EcoRoll är en funktion, med låga utvecklingskostnader, som reducerar bränsleförbrukning. Detta då den tillåter en effek-tivare konvertering av lagrad potentiell energi till kinetisk energi under en nedförsbacke. Funktionen öppnar drivlinan i nedförsbackar och låter därmed motorn gå på tomgång, vilket minskar motorförlusterna. I detta arbete har två oli-ka reglerstrategier utvecklats, där den ena är baserad på nuvarande tillstånd, medan den andra använder sig av in-formation om vägen framför fordonet, dvs. Look-Ahead. Si-mulering på vägen från Södertälje till Norrköping ger en bränslereducering på ungefär 3,4% för den första strategin och 3,7% för den andra, vid jämförelse med ett likadant fordon med en traditionell farthållare.

(7)

Acknowledgements

The research presented in this Master’s thesis is the final project in achieving my M.Sc. Degree in Electrical Engineering at the Automatic Control department at the Royal Institute of Technology, kth, in Stockholm, Sweden. The work has been carried out at the Driving Assistance Software unit, neca, at Scania cv ab in Södertälje, Sweden.

I would like to start expressing my gratitude to the senior manager of Control Strategy section, nec, at Scania, Magnus Staaf, who gave me the opportunity to do this thesis and believed in my idea. Also, I would like to thank Andreas Renberg, the head of neca, for his review and feedback.

I owe my deepest gratitude to my supervisor at Scania, Oskar Johansson, for his great inputs, patience, and guidance through this project. He made an effort even during his vacations. Many thanks to Maria Södergren who was my second supervi-sor. She followed the project continuously, giving tips and wise inputs. Thank you Oskar and Maria! Furthermore, it is a pleasure to thank my examiner Dr. Ather Gattami at kth, for his positive support and advice.

I am also indebted to all my colleagues from different departments at Scania, among them; Mikael Ögren, Anders Kjell, Peter Asplund, Magnus Svensson, Niklas Lerede, Niklas Pettersson, and Olof Lundström, to mention some of them. They all have contributed to this project by sharing invaluable knowledge or giving feedback on the report. I would also like to express my gratitude to my colleague and friend, Kuo-Yun Liang, for the valuable and interesting discussions we have had together. Most of all I would like to thank my parents for supporting and advising me. I owe them my life for achieving this stage of life and study. Last but absolutely not least, many thanks to my wonderful wife who encouraged me, reviewed the report and made the time during the project great.

Mustafa Abdul-Rasool

(8)

List of Abbreviations

AMT automated manual transmission APS air processing system

CC cruise control

CVT continuously variable transmission DHSC downhill speed control

DP dynamic programming ECU electronic control unit GPS global positioning system HDV heavy-duty vehicle

ICE internal combustion engine PI proportional-integral

(12)

(13)

List of Symbols

aER The estimated acceleration for an open powertrain. Av Front area of the vehicle.

CaF A coefficient used in Michelin model for rolling resistance. Cb A coefficient used in Michelin model for rolling resistance. Cd Air resistance coefficient.

Cr Rolling resistance coefficient.

CrrisoF A coefficient used in Michelin model for rolling resistance. decmax Maximum accepted deceleration.

Fair The force caused by air resistance. Fbrake The braking force acting on wheels. fcost The cost function.

Fdrive The driving force generated by the engine. Fgrav The longitudinal gravitational force.

FICElosses The total engine losses expressed as force losses.

Froll The force caused by rolling resistance. Ftot The total forces acting on a HDV. Ftrac The traction force on the wheels. g The gravitational constant. gnr Gear number.

if The gear ratio of the final drive. it Transmission’s gear ratio.

(14)

CONTENTS

Mv The mass of the vehicle. Mε The mass estimation error.

n An integer when multiplied with ∆p gives the distance from p to the

predicted position.

ncyl The number of cylinders in the engine. nr Number of revolutions per cycle. p Current position of the HDV.

∆p The distance between two positions.

qf The fuel consumption in [g/sec].

qfacc The accumulated fuel consumption in [liter].

qf_accCC The accumulated fuel consumed by the referenceHDV, during a

simu-lation cycle.

qf_accER The accumulated fuel consumed by a heavy-duty vehicle (HDV) equipped

with EcoRoll, during a simulation cycle.

qfaccsave The fuel saving of EcoRoll compared to the reference HDV.

qfinst The instantaneous fuel consumption per revolution, [g/cycle].

rw Wheel radius.

Tc The output torque of the clutch to the gearbox.

tCC The time it takes for the reference HDV, to travel through the road

that is used in simulation.

tchangemin The lower time limit for remaining within a main state; opened or

closed powertrain.

Tdrag The drag torque, caused by engine’s internal friction.

tER The time it takes for aHDV equipped with EcoRoll, to travel through

the road that is used in simulation.

Tf The output torque from the final drive. TICE The torque out from theICE.

T_ICE∗ The demanded ICE torque.

(15)

CONTENTS

tnewpred The time period for a new start of prediction.

Tpump The coolant pump losses expressed in torque loss. tsample The sampling time for speed prediction.

tsave The travel time savings of EcoRoll compared to the reference HDV. v The current vehicle speed.

v0 Initial vehicle speed when entering the state Active.

˙v The time derivative of the vehicle speed. ˆv Prediction of speed.

vair The air speed.

vCCmargin A parameter specifying how much the speed may drop below CC speed

with an open powertrain.

vCCset The CCset speed.

vDHSCmargin Margin from DHSCset speed.

vDHSCset The DHSCset speed.

vISO The nominal speed where Michelin model is linearized around. vmax Maximum speed where the EcoRoll can be active, when positive

ac-celeration is estimated.

vmin Minimum speed where the EcoRoll can be active, when positive

accel-eration is estimated.

voverspeed A parameter defining a downhill when the speed increases to its value. vpredmax The maximum predicted achievable speed.

α The prevailing road inclination.

αCP The inclination corresponding to the HDV, without any fuel injected,

keeping the desired speed with closed powertrain.

αER The inclination that lets the speed of the HDV being maintained with

open powertrain.

αlesssteep Inclinations where theHDV decelerates when the powertrain is open.

αsteep Inclinations where theHDV accelerates despite no fuel is injected. αuphill All positive inclinations.

(16)

CONTENTS

β A weight coefficient for the cost function. ηf The efficiency of final drive.

ηt The efficiency of the transmission.

ωc The angular speed on the output shaft of the clutch. ωf The output angular velocity from the final drive. ωICE Angular velocity of the ICE.

ωidle The angular velocity of the engine running on idle. ωw The angular speed of the wheels.

ρair The air density.

(17)

Chapter 1

Introduction

The technology behind hybrid vehicles is nowadays one of the main approaches in order to reduce fuel consumption. The hybrid system enables the feature of regen-erative braking, i.e., the battery gets charged whenever the vehicle is braking. This technology enables the energy to be utilized more efficiently. However, regenerative braking is more suitable for city driving, with frequent braking. This makes the function less suitable for long haulage vehicles.

What feature can then be utilized on a highway? One answer is hills. Potential energy is stored while driving uphill, which will be converted into kinetic energy when travelling downhill. Thus, the idea of EcoRoll was born in me; let the vehicle convert the potential energy as efficiently as possible.

The idea is to let the vehicle roll with disengaged gear in downhill, with the engine running only on idle to power steering-, brake servo, etc. This was presented to theR&Ddepartment at Scania cv ab, a Swedish heavy-duty vehicle (HDV)

man-ufacturer, together with the thesis application. Apparently the idea was not new. However, unlike Volvo and Mercedes trucks, Scania has chosen different solutions. One solution is to decrease the engine speed by lowering the gear ratio, when only a small amount of torque is needed, e.g., when travelling downhill. This reduces the losses with a closed powertrain.

A study made by Anders Jensen at Scania [8] concludes that without using any Look-Ahead data, the EcoRoll function reduces fuel consumption. However, it could also have a negative impact due to the risk of a hill being too long and steep, where braking is needed before the end of the hill. In this case, it is more beneficial to go into fuel-cut off mode, but it requires Look-Ahead data to identify the length of the hill. As known so far, no HDVswith EcoRoll utilize Look-Ahead data.

The Ph.D. thesis [5] concludes that the fuel consumption can be reduced by approximately 3%, but only on specially constructed road topographies. Otherwise, only 0.3% can be reduced on an authentic road. The results are generated from a control strategy based on dynamic programming (DP), which gives the optimal

result for the specific vehicle on the studied road. However, a deeper study on EcoRoll is needed for Scania to identify the potential of this functionality. Also,

(18)

CHAPTER 1. INTRODUCTION to compare two control strategies, where one is based on actual circumstances and the other on Look-Ahead data. Thus, this thesis studies EcoRoll with and without Look-Ahead data. The strategies are compared to conventional cruise control (CC),

(19)

Chapter 2

Background

This chapter presents a background describing EcoRoll, Look-Ahead, and the pur-pose of this thesis. Also, previous related studies in this area are presented.

2.1 EcoRoll

The concept of EcoRoll is to disengage the gear when traveling downhill. Its aim is to increase the benefit of the potential energy stored from driving uphill. This is converted into kinetic energy that propels the vehicle when travelling downhill. It is well known that energy conversions usually have losses in form of heat, friction, etc. One of the losses in a vehicle is the engine itself. It has friction that needs to be overcome. Furthermore, the friction increases with higher angular velocity. Therefore, the losses are decreased when disengaging the gear in a downhill. The speed is also increased further, since the impact of engine losses on the motion of the vehicle is eliminated, see Figure 2.1. By that the fuel consumption can be reduced. However, the internal combustion engine (ICE) cannot be turned off since it powers

the steering- and brake servo, etc, unless these are electronically powered systems. For HDVs, this is usually not the case. Hence, the ICE should be kept running at

idle.

Since this thesis focuses onHDVs, and the servo is presumed not to be

electron-ically powered, theICE will run on idle when driving in EcoRoll mode.

2.2 Look-Ahead

An experienced driver plans his driving by looking at the road ahead. Doing this, the correct gear can be selected early and the speed reduced before a downhill, providing fuel efficiency and better driving performance. However, this ability is missing in the automated functionalities in vehicles such as the automatic gearbox and theCC. The solution is to use Look-Ahead data.

The idea of Look-Ahead is to use a global positioning system (GPS) for

(20)

CHAPTER 2. BACKGROUND

Figure 2.1: The graphs shows the behavior of two HDVs traveling down a hill.

The darker graphs shows the speed and fuel consumption for aHDVequipped with

EcoRoll, and the other two graphs corresponds to a conventionalHDV.

for the possibilities of this technology. It creates great opportunities for improving most of the systems in a vehicle, regarding performance and fuel efficiency.

2.3 Problem Statement

The idea of EcoRoll is to reduce the fuel consumption by automatically disengaging the gear in a downhill to let the vehicle roll freely, and let the engine run on idle. This raises some questions:

• How much fuel can be reduced by EcoRoll?

• Can the results be improved by taking advantage of Look-Ahead data?

• Are the results sensitive to different disturbances or configurations?

The aim of this thesis is to answer these questions. This section introduces the goals of the study, the approach to reach them and the delimitations of this work.

(21)

2.3. PROBLEM STATEMENT

2.3.1 Goals

The main goal is to develop two control strategies for EcoRoll and study their fuel saving potential. The first strategy is based on prevailing conditions, while the second uses Look-Ahead data.

It is of course important to analyze the sensitivity of the results for robustness. Sensitivity analysis is done to the following areas:

• Rolling and air resistance models and their parameters. • Differences in vehicle mass estimation.

• Different road topographies.

• Engines with different characteristics and piston displacements.

Another goal is to validate the results by driving a HDV, manually requesting

disengaged gear when it seems beneficial. A computer is connected to log data, such as fuel consumption. Therefore, to get a reasonable comparison, the vehicle param-eters are chosen according to an available truck that will be driven. Furthermore, the same road topography that is simulated will be driven, in order to be able to compare the results.

2.3.2 Approach

The study is divided into four phases; pre-study of EcoRoll, conventional EcoRoll, EcoRoll utilizing Look-Ahead data and result analysis. Each phase complements each other.

Pre-study: A feasibility study is done in order to understand when it is beneficial to activate EcoRoll. In order to develop a reasonable control strategy, different simplified downhill slopes are studied where EcoRoll is activated in different parts of the hill.

Conventional EcoRoll: The aim of studying the conventional EcoRoll is to iden-tify its potential to reduce fuel consumption and its limitations.

Look-Ahead EcoRoll: The identified limitations of the conventional EcoRoll are used to create a control strategy that overcomes these limitations by using Look-Ahead data.

Analyses: The results are analyzed to study the sensitivity for different distur-bances and parameters, e.g., wind or different road topographies. Also, in order to validate that the results are reasonable, a HDV is driven and data is logged. The

(22)

CHAPTER 2. BACKGROUND

2.3.3 Delimitations

Several functions could require a closed powertrain to ensure the most efficient operating points with respect to fuel consumption. An example of this is the air processing system (APS). The impact of these systems will not be considered in the

basic Simulink model. Instead, these factors will be taken into account when the results are validated in the truck.

The position of the HDV received from the GPS is assumed to be accurate and

correct in all conditions. However, other modeling assumptions will be presented clearly in their respective section.

2.4 Related Work

There are already patents in this area by different manufactures, among them is Volvo which has had several since 2001. Some of these patents treat EcoRoll both for use withCCand with gas pedal, while others treat improvements of said strategies.

Anders Fröberg [5] has done a study in cooperation with Scania. Here, two control strategies were compared. One was based on current slope where EcoRoll were activated between two predefined angles. The second utilizes DPwith Look-Ahead

data. DP with an infinite horizon gives the optimal result regarding fuel efficiency,

but is a processor intensive strategy. Another study, made at Scania by Anders Jensen [8], compares EcoRoll with Scania’s LowRev. The concept of LowRev is to design the HDVs such that the engine runs at maximum of 1000 rpm when driving

at 80 km/h and when the driving resistance is small [3]. The conclusion was that the best option is to combine EcoRoll and LowRev by using Look-Ahead data, since the functionalities have advantages in different hills.

EcoRoll has also been studied for other circumstances. A Master’s Thesis by Arvid Rudberg [2] studied a fuel efficient deceleration to reach a speed limit decrease. EcoRoll was the most fuel efficient way to decelerate, but caused an increased travel time. Since this deceleration is slower than braking, it needs to start earlier which can affect the vehicles behind. Therefore EcoRoll, engine braking and brakes are combined to find an optimum regarding both fuel efficiency and driving time.

(23)

Chapter 3

Vehicle Model

The powertrain of aHDVconsists of an engine, clutch, transmission, propeller shaft,

final drive, drive shafts, and wheels, depicted in Figure 3.1. It is important that each part is modeled such that it is able to show reasonable results for the specific subject that is studied but at the same time not being too complex. A basic model of the longitudinal propulsion of the vehicle is well described in Vehicular Systems [10, ch. 8]. This will serve as premise for the Matlab Simulink model used in this study.

Figure 3.1: An overview of the main components a powertrain consists of.

3.1 Basic Model of Powertrain

As a start, an overview of the vehicle model will be presented followed by a deeper description of the important parts for this study. All parameter values for the vehicle model, that are used for this study, are given in Table A.1 in Appendix A.

The longitudinal motion of the vehicle is described by Newton’s second law,

(24)

CHAPTER 3. VEHICLE MODEL where Mv is the HDV mass, ˙v its acceleration, and Ftrac, Fair, Froll and Fgrav the

forces acting on the vehicle. The forces are visualized in Figure 3.2 and discussed in following subsections.

Figure 3.2: The longitudinal forces acting on aHDV.

3.1.1 Driving Resistances

The gravitational force is given by

Fgrav = Mvgsin (α) , (3.2)

where g is the acceleration caused by the gravity and α the current inclination, defined positive for an uphill, see Figure 3.2.

The air resistance increases quadratically to the difference between vehicle speed and air speed,

Fair= 1₂CdAvρair(v − vair)2, (3.3)

where Cd is the air force coefficient, Av denotes the front area of the vehicle, ρair

the air density, v the vehicle speed and vair the air speed. However, the rolling

re-sistance is more difficult to model. Scania has previously noticed problems with the models used today; calculations and simulations do not match the measurements. Therefore, a Master’s Thesis [6] studied this issue and developed a new model. The model is however too complex for this work, therefore Michelin’s rolling resistance model is used, Froll= Mvg q 1 + rw 2.7 CrrisoF + Cb(|v| − vISO) + CaF v2− v_ISO2 , (3.4)

where rw is the wheel radius, vISO the nominal speed and CrrisoF, Cb and CaF are

(25)

3.1. BASIC MODEL OF POWERTRAIN

Since Michelin is a company that develops and manufactures tires, their model is assumed to be sufficient for the purpose of this study. However, in order to consider the deviation of the model from real measurements, a sensitivity analysis is done in Chapter 8. This is achieved by comparing the results from Michelin’s model to another model. The chosen model is:

Froll= MvCrgcos (α) , (3.5)

where Cr is a coefficient and α the current road inclination.

3.1.2 Traction Force

The traction force Ftrac is the force acting on the wheels generated by the engine, Fdrive, and brakes, Fbrake:

Ftrac = Fdrive− Fbrake. (3.6)

The torque generated by the ICE TICE is transferred through the powertrain

with different gear ratios ending in the wheels, which gives a force Fdrive that sets

theHDVin motion.

The engine characterization is described deeper in the next subsection. In mean-time consider it as a black box with the torque, TICE, and the angular velocity, ωICE, as outputs to the powertrain. Furthermore, it uses the demanded torque, T_ICE∗ , as an input. The angular velocity of the ICE and wheels are related to each

other with a factor defined by the current gear ratio and final drive, see equation (3.8) and (3.9). The clutch is assumed to be stiff. Therefore the torque and angular velocity from theICE are the same that is delivered to the gearbox,

(

Tc = TICE

ωc = ωICE

, (3.7)

where Tc is the output torque of the clutch to the gearbox, and ωcis corresponding

angular velocity.

The transmission is a gearbox with different gear ratios denoted by it(gnr), where gnr is the gear number. Each gear has its efficiency, ηt(gnr).

The next part is the final drive that is connected through the propeller shaft. The propeller shaft is assumed to not have any friction. Furthermore, the final drive is a fix ratio if designed such that the engine angular velocity is within desired

operating points. The efficiency of the final drive is denoted by ηf. The obtained

final drive torque Tf, and its angular velocity ωf are given by

   Tf = TICEitif− |TICEitif(1 − ηtηf) | ωf = ωICE itif . (3.8)

The wheels are connected through drive shafts. They are also assumed to be ideal. It is also assumed that the two wheels on the shafts have the same speed.

(26)

CHAPTER 3. VEHICLE MODEL Therefore, the shafts can be modeled as one shaft, and the conversion between angular velocity of the wheel to vehicle speed is thereby given by

v= rwωw = rwωf = rw ωICE

itif

, (3.9)

where ωw is the angular speed of the wheels.

The forces acting on the wheel where described in (3.1), where Ftrac is given by

(3.6), which is the difference between Fdrive and Fbrake. Fdrive is Tf given in (3.8)

divided by rw: Fdrive = Tf rw = 1 rw (T ICEitif − |TICEitif(1 − ηtηf) |) . (3.10) HDVs have in exception of the usual service brakes, different auxiliary brake

systems, e.g., a retarder and an exhaust brake. The auxiliary brakes are often used inHDVs in order to minimize the wear of the service brakes, but also to avoid

brake fading when the service brakes are used extensively and continuously. Since simulations are done on highways, a model of the retarder is enough for this study. It is modeled as a desired braking torque from the transmission through the final drive,

Fbrake =

Tbrakeif rw

. (3.11)

Finally, theHDV system from engine to vehicle motion is described by:

                                 v = rw ωICE itif ˙v = 1 Mv (Ftrac

− F_air− F_roll− F_grav) = = 1 Mv 1 rw (TICE itif − |TICEitif(1 − ηtηf) |) − Tbrakeif rw − − 1 2CdAvρairv2−

Mvg CrrisoF + Cb(|v| − vISO) + CaF v2− v2ISO

1000q 1 + rw 2.7 − − Mvgsin (α) . (3.12)

3.2 Important Parts of the Powertrain

3.2.1 Engine

Since the engine losses TICElosses are central for this work, it is necessary to look

deeper into its characteristics. There is the so called drag torque Tdrag, which is the

internal friction caused by moving parts in the engine. The drag torque increases by increased engine angular velocity. Another loss is the coolant pump Tpump that

(27)

3.2. IMPORTANT PARTS OF THE POWERTRAIN

is cooling the engine. The higher angular velocity the engine operates at, the more the engine needs to be cooled, which increases the losses,

TICElosses = Tdrag + Tpump. (3.13)

A total fuel map based on measurements is used to calculate the fuel consump-tion. An example of a typical fuel map for an ICE is shown in Figure 3.3. The

fuel map has TICE and ωICE as inputs and returns the fuel consumption qf in

[g/sec]. An accumulated fuel consumption qfacc, and an instantaneous consumption

per revolution qfinst are then given by:

       qfinst = 2πnr ncylwICE qf [g/cycle] qfacc = 1 ρf uel Z qfdt [liter] , (3.14)

where nr is number of revolutions per cycle, ncyl is number of cylinders in engine,

and ρf uel denotes the density of the fuel that is used.

0 500 1000 1500 2000 2500 −20 0 20 40 60 80 100 −20 0 20 40 60 80 100 engine speed [rpm] Fuel Map normalized torque [%]

normalized fuel consumption [%]

Figure 3.3: An authentic example of a typical fuel map for anICE.

There is a limit of how much torque the engine can deliver. The amount varies for different angular velocities. The engine has an inertia, which limits the speed of

(28)

CHAPTER 3. VEHICLE MODEL a change in torque. However, the dynamics of the vehicle is significantly slower than the dynamics of the engine. Therefore, the inertia of the engine can be neglected.

3.2.2 Gear Shifting

A gear shift is assumed to take one second, during which no torque is delivered from the transmission to the wheels. Therefore, the vehicle motion described in (3.1) becomes, ˙v = 1 Mv (−Fair − F_roll− F_grav) = = 1 Mv  − 1 2CdAvρairv2− Mvg q 1 + rw 2.7

CrrisoF + Cb(|v| − vISO) + CaF v2− v2_ISO

1000 −

− Mvgsin (α)

.

(3.15) However, the dynamics for the angular speed of the ICE during the gear shift is

not regarded. A request for a gear shift is a very complicated action in the current automated manual transmission (AMT). Therefore, a basic shifting model is done

based only on the engine’s angular velocity, which is specific for each gear ratio. When EcoRoll is activated, the gearbox is set to neutral. Thus, no torque is delivered from the ICE, and is modeled as in (3.15). Furthermore, the engine is

run on idle speed ωidle, where the standard speed is 500 rpm for 6-8 cylindrical

engines. When disengaging the powertrain, the drop of ωICE to ωidle is not done

instantaneously. Likewise, when the powertrain is to be engaged, the increase of

ωICE does also require time and fuel. However, these dynamics are not regarded

since the fuel consumption needed to engage the gear is assumed to be small and can thus be neglected, compare the discussion around (3.13). Besides, the injected fuel to increase ωICE before engaging the powertrain, is compensated when the

powertrain is disengaged, since no fuel is injected when ωICE is decreased.

3.2.3 Cruise Control and Downhill Speed Control

Due to the mass, theHDVshave more advanced functionalities for maintaining speed

than cars. A HDV is usually equipped with a traditional CC, which maintain the

speed according to the speed reference set by the driver. However, theCCdoes not

use any brakes. And due to the large mass of aHDV, it can accelerate rapidly when

travelling downhill. Therefore, a downhill speed control (DHSC) has been developed

forHDVs. It enables the driver to set a maximum speed that theHDV is allowed to

reach.

In this study, it is assumed that the driver always uses theCCand DHSC, where

(29)

3.3. MODELING SOFTWARE

3.3 Modeling Software

Matlab Simulink is the simulation software that is used. It is a simulation and model-based design software for dynamic and embedded systems [9]. It also has a powerful tool, Matlab Stateflow, which is especially suitable for the control strategy that is developed. This, as the control strategy is based on rules defining transitions between two discrete states; either open or close the powertrain.

(30)

(31)

Chapter 4

Pre-study of EcoRoll

An interesting part of EcoRoll is that it is possible to implement in a vehicle equipped with an automatic gearbox, only by upgrading the software. There is no need for any hardware changes, which is discussed in the following section. Fur-thermore, it is important to consider possible safety risks and advantages when implementing a new function. These aspects are discussed briefly in this chapter, but are only guidelines for future work since the purpose of this thesis is to study the fuel consumption. However, this chapter is studying hills with different slopes, in order to determine when it is beneficial to open the powertrain. This chapter is concluded with a summary and a discussion of the results.

4.1 Possibility of Implementation

Unlike cars, the automaticHDVs are often equipped with an AMT. The concept of

anAMTis a manual gearbox that is automatically controlled by a electronic control

unit (ECU) with actuators and sensors. The typical automatic gearbox, continuously

variable transmission (CVT), that is used in cars is rarely used in HDVs. The main

reason is that aCVThas lower efficiency than a manual transmission. However, the AMTmatches or even improves the efficiency of the manual transmission [7], hence

it is preferred.

When a gear change is demanded, a signal is sent to the clutch to open the powertrain between the ICE and the gearbox in order to be able to disengage the

gear, see Figure 3.1. Then, the clutch is engaged. Engaging the new gear is done by torque control; the engine speed is synchronized with the powertrain angular veloc-ity, regarding the new gear ratio. However, the angular velocity of the powertrain is directly depended on vehicle speed, which can be changed during the gear shift, due to the environmental forces (3.2), (3.3) and (3.4). So, a good prediction of the vehicle speed is necessary.

Thus, disengaging a gear requires only a request by existing signals. Therefore, EcoRoll only needs software implementation, and no extra hardware is needed.

(32)

CHAPTER 4. PRE-STUDY OF ECOROLL

4.2 Safety Aspects

Engine stops and slippery roads are possible risks for EcoRoll. If the engine suddenly stops during activated EcoRoll, the steering servo will be turned off. This can be solved by either closing the powertrain to help the engine to start rotating again, and thereby turning on the steering servo, or by powering the servo electrically. Besides, an electrical steering servo as such, is more fuel efficient than a conventional one [11]. However, this is a large change that initially could have high costs, while the first solution of closing the powertrain is just a matter of control signals. It depends though on how the powertrain has been opened. It is easy to close the clutch by a control signal, but a disengaged gear requires a controlled engine speed, which is impossible with the engine turned off. Due to possible lubrication problems, disengaging gear may be preferable to clutch opening. If so, a system for detecting when the engine is turned off during activated EcoRoll, and that it tries to start it again, should be developed.

The brake system on the other hand are, for HDVs, based on air pressure. If

the tanks already have high pressure, the brakes will work even though the engine is off. However, the APS that charges the tanks is dependent on a running engine.

Therefore, long use of brakes drops the pressure. If no action is taken theHDVwill

remain unable to brake.

On slippery roads, it is important to not use the auxiliary brakes. This, since these brakes are acting only on the driving axles. The EcoRoll in Volvo trucks is activated only if the retarder is in automatic mode [1]. This is necessary though, as the activation of EcoRoll results in an increased acceleration. This, since the increased speed needs to be braked if the maximum speed is reached. However a disengaged powertrain is optimal to avoid skidding during a slippery turn, according to vehicle dynamics [10, ch. 12.3]. The reason is that the tires can take maximum side force when no longitudinal force is acting on them. Otherwise, e.g., when engine braking with closed powertrain, the ability to take side force is limited even more. This may give unbalance between wheels, that causes the skid in a slippery turn. Low speed is thus required before the turn, and thereby it is more optimal to have a closed powertrain and slow down the vehicle speed before the turn, and open the powertrain only during the turn.

4.3 Road Topography and Strategies

The most important part of the feasibility study is to identify the characteristics of a hill that needs to be fulfilled, such that it is advantageous to activate EcoRoll. Furthermore, it can also be more or less efficient to activate EcoRoll before, in and after a downhill. This section will discuss different hills and strategies followed by a section where the simulation results are presented and discussed.

(33)

4.3. ROAD TOPOGRAPHY AND STRATEGIES

4.3.1 The Studied Hills

There are three different downhill slopes that are of interest; an ideal, steeper, and less steep hill. These hills are defined and described here and used for the study in next section.

Ideal Downhill

The definition of an ideal downhill for EcoRoll in this work is a hill with a slope such that the sum of the environmental forces is zero,

Fgrav + Froll+ Fair= 0, (4.1)

i.e., the vehicle with an open powertrain can keep its speed constant. This hill is interesting since a fuel reduction for this situation indicates that the reduction would be even greater for an acceleratedHDV with an open powertrain. However,

that is not obvious for all accelerating situations. This will be described in the following subsection.

If the powertrain is closed in the ideal hill, the drag torque of theICEis added as

a resistance force and thereby the vehicle decelerates if no more fuel is injected. The additional fuel is needed since the drag torque increases for higher angular velocity as described in Section 3.2.

The inclination of an ideal downhill with activated EcoRoll, αER, is derived

using (4.1), where the gravity force is given in (3.2). The resulting expression is then, αER= − arcsin _F roll+ Fair Mvg . (4.2)

To get corresponding inclination for the closed powertrain, αCP, the losses of

theICE is added to (4.1) obtaining

FICElosses + Fgravity+ Froll+ Fair= 0, (4.3)

where FICElosses is the engine losses given by an engine specific table, see Section 3.2.

The ideal inclination for a closed powertrain is expressed as:

αCP = − arcsin

_F

ICElosses+ Froll+ Fair

Mvg

. (4.4)

Obviously a hill with inclination αCP is steeper than αER. Thereby, if the

condition for α in,

αCP < α < αER, (4.5)

is fulfilled, the vehicle accelerates when EcoRoll is activated and decelerates when the powertrain is closed.

(34)

CHAPTER 4. PRE-STUDY OF ECOROLL Steeper Downhill

Here, the negative slope αsteep is larger than αCP,

αsteep< αCP. (4.6)

Therefore, theHDV will accelerate with a closed powertrain even though no fuel is

injected. According to the PhD Thesis [5], it is always more beneficial to close the powertrain, since the vehicle accelerates even though it enters fuel cut-off mode. This is though not obvious, since by opening the powertrain the acceleration be-come even higher. With the increased speed, the fuel consumption is decreased per distance, since the fuel consumption of an engine running at idle is constant per time interval. EcoRoll eliminates the drag losses from the ICE, which means that

EcoRoll is activated longer, while a conventional HDV will need to deliver torque

sooner, seen in Figure 2.1.

However, the problem occurs when the length of the hill is not known, since the speed will increase more with activated EcoRoll and the maximum speed set by theDHSCcould be reached earlier and the HDVwill then brake. Increased braking

means more losses.

Less Steep Downhill

Due to a smaller negative slope, αlesssteep,

αER< αlesssteep <0, (4.7)

the HDV will decelerate even if the powertrain is open. Here, it is interesting to

study whether it is beneficial to accept some deceleration by EcoRoll mode followed by a peak in torque from theICEto maintain the speed. If the required torque peak

is brief the accumulated fuel consumption could still be small. However, the time is increased.

It is also interesting to study the advantage of activating EcoRoll in such hills, if the current speed is above the set speed, and let the vehicle decelerate slowly down to the set speed.

Uphill

Despite the inclination, αuphill, being positive,

αuphill>0, (4.8)

the EcoRoll mode could be beneficial in specific situations. As in Less Steep Hill, when the current speed is above the CC set speed, the drag torque from ICE is

eliminated on the vehicle motion with activated EcoRoll, even though the gravity force is opposite to the speed direction,

FICElosses+ Fgravity+ Froll+ Fair> Fgravity+ Froll+ Fair>0. (4.9)

(35)

4.3. ROAD TOPOGRAPHY AND STRATEGIES

4.3.2 Activation Time of EcoRoll

According to [5], it is beneficial to activate EcoRoll only when the inclination of the hill fulfills (4.5). Even so, it is still interesting to study hills with inclinations outside the given range. So in this phase of the study, EcoRoll will be activated even in hills with other slopes.

Another aspect is to study where in the hill it is beneficial to activate EcoRoll. Of course, during the hill is the first thing that comes in mind, call the segment

In Hill. When the vehicle is accelerated during the hill, it is interesting to study

the benefit of decelerating with EcoRoll after the hill untilCC set speed is reached,

With AfterHill.

There are studies within Look-Ahead where the speed is reduced before a down-hill to decrease the need of braking. An attendant question is whether it is more efficient to reduce the speed before the hill with EcoRoll, ER BeforeHill, or with closed powertrain, CP BeforeHill.

4.3.3 Constructed Simplified Hills

The simplified hills are 15 km long and consist of three parts with equal lengths, see Figure 4.1. The first part is plane followed by a downhill, and in order to maintain stationary conditions, the last part is also plane. The slope of the second part is varied to study the impact of different inclinations. The steepest hill is constructed also with a short version in order to study the influence of the length of hill.

Figure 4.1: The constructed simplified hill. Note that the inclination is small, but seems to be high due to scaling. The inclination is -0.98%, corresponding to αER

(36)

CHAPTER 4. PRE-STUDY OF ECOROLL The first part makes it possible to study the BeforeHill function. The last part enables study of the AfterHill function. It also allows both the vehicle that is equipped with EcoRoll and the reference vehicle to reach the CC set speed, to get

comparable results.

An even more simple hill was constructed without any plane part, only a constant slope. Here, it is chosen to study the reduction of fuel consumption when travelling in an inclination of αER.

4.4 Quadratic Cost Function

The control strategies are based on rules rather than a cost function. However, since the reduction of the fuel consumption usually increases the travel time, a cost function is used to regard both the travel time and fuel consumption, in order to determine a fair result.

According to [4], a quadratic cost function is preferred to a linear cost function. This, since the quadratic function values points of the same distance to the origin equal. The origin is though not achievable, since it means that the HDV travels

with infinity speed by no fuel consumption. However, the aim is to move in this direction. The quadratic cost function is given by,

fcost = β _t ER tCC 2 + (1 − β) qfaccER qf_accCC !2 , (4.10)

where travel time and accumulated fuel consumption, for the HDV equipped with

EcoRoll, are normalized by corresponding time and fuel consumption for a reference

HDV with conventional CC. Furthermore, β is a weighting parameter. In [4], β is

determined such that it is optimal to drive in constant speed on plane road. The value of beta, for aCCset speed of 80 km/h, is determined in [4] to be

β = 0.4. (4.11)

It is worth to mention that a decreased fuel consumption for EcoRoll does not in-crease the travel time, while it usually does in a conventionalHDV, see Section 4.3.1.

However, the cost function is based on an increased time when fuel is reduced, but since EcoRoll is compared to a conventionalHDV, (4.10) is a reasonable cost

func-tion.

The results will be presented in percentage, where a negative result corresponds to saving when EcoRoll is utilized,

             fcost% = 100 (fcost−1) tsave = 100 _t ER tCC −1 qfaccsave = 100 qf_accER qf_accCC −1 ! . (4.12)

(37)

4.5. SIMULATION RESULTS

4.5 Simulation Results

For the studied vehicle with parameters according to Table A.1, the ideal slopes are calculated according to (4.2) and (4.4). The results are:

(

αER= −0.0098 rad ≈ −0.98% αCP = −0.0116 rad ≈ −1.16%

. (4.13)

The fuel saving on a hill with the constant slope αERis as much as 73% without

any change in travel time. The result is amazing! However, that is for an ideal hill, which is only a small part in authentic roads.

A simulation is made where two vehicles, a conventional vehicle and one equipped with EcoRoll, driving through the simplified hills, see Figure 4.1. The results pre-sented in Table 4.1 is for when the EcoRoll is activated only during the segment In

Hill. The first column gives the slope of the second part of the hill. The cost in

second column is calculated according to (4.10), based on the time- and fuel saving results presented in column three and four. A negative value indicates reduction of time or fuel respectively.

To study the impact of the length of a hill, a shorter hill of the steepest hill were also studied.

Table 4.1: The results of aHDV equipped with EcoRoll using the function In Hill,

i.e., opening the powertrain when the HDV is in the downhill. A Negative sign

indicates a saving compared to a reference vehicle equipped with a conventionalCC.

Slope Cost Time Fuel

-0.78% -3.06% 2.94% -4.65% -0.88% -8.06% 1.35% -7.94% -0.98% -7.27% 0.01% -6.26% -1.08% -6.01% -1.15% -4.34% -1.18% -4.78% -1.98% -2.71% -1.28% -2.47% -1.46% -1.10% -1.38% -0.33% -0.79% 0.25% -1.48% 0.18% -0.32% 0.36% -1.48% short -1.53% -0.29% -1.09%

Table 4.2 shows the simulation results using the With AfterHill function. Hills that does not give any acceleration are not of interest for this function and are therefore omitted. The results of using the Only AfterHill function is presented in Table 4.3. Here only hills where the HDVreaches, at least nearly, the DHSC set

speed are of interest.

The last study is the Before Hill function on the steepest hill where the DHSC

set speed is reached. Here are CP BeforeHill and ER BeforeHill combined with

(38)

CHAPTER 4. PRE-STUDY OF ECOROLL Table 4.2: The results of a HDV equipped with EcoRoll using the function With

AfterHill, meaning that the powertrain is opened both in the hill and after the end

of the hill, untilCCset speed is reached again. However, this study requires a slope

giving an acceleration. Therefore, some hills are omitted.

-1.08% -6.39% -1.16% -4.67% -1.18% -5.56% -2.01% -3.38% -1.28% -3.27% -1.49% -1.75% -1.38% -1.27% -0.83% -0.51% -1.48% -0.76% -0.36% -0.39% -1.48% short -2.21% -0.33% -1.64%

Table 4.3: The results of a HDV equipped with EcoRoll using the function Only

AfterHill, i.e., the powertrain is only opened after the end of the hill. This is

studied for steep slopes where the DHSCset speed is reached in the hill.

-1.28% -0.47% -0.01% -0.39% -1.38% -0.88% -0.03% -0.71% -1.48% -0.94% -0.04% -0.76% -1.48% short -0.46% -0.02% -0.37%

4.6 Discussions on the Feasibility Study

The ideal hill, with the 73% fuel saving, indicates that large losses in the ICE can

be reduced by opening the powertrain. However, such hills are rare on authentic roads. Thus, a real indication of fuel saving for aHDVin duty can only be obtained

by constructing a good control strategy and simulate on known data of an authentic road. These simulations are done for the two EcoRoll strategies that are presented in Chapter 5 and 6.

The results in Table 4.1 show that it is advantageous to activate EcoRoll even for steeper slopes than αCP. However, the limit of how steep the hill can be to

be beneficial to open the powertrain, is dependent on the length of the hill and the difference between CC set speed and DHSC set speed. That is proved by a

comparison of the regular and short hill with inclination -1.49%. TheHDVdoes not

reach theDHSCset speed when travelling down the short hill, while it does reach it

in the regular one. In the table, both fuel and time is saved in the short hill using EcoRoll compared to conventional CC. On the longer hill, only time is saved with

(39)

4.6. DISCUSSIONS ON THE FEASIBILITY STUDY

Table 4.4: In too steep and long hills, it is beneficial to decrease the speed before the start of the hill to avoid braking. This table compares two ways of decreasing the speed before a hill; with closed powertrain (CP) and open powertrain (ER). It is tested in combination of With AfterHill or Only AfterHill.

With AfterHill -1.48%-1.48% ER BeforeHill -5.95%-0.76% -0.36% -0.39%1.07% -5.84% -1.48% CP BeforeHill -4.82% 1.01% -4.81%

Only AfterHill -1.48%-1.48% ER BeforeHill -5.42%-0.94% -0.04% -0.76%2.40% -6.34% -1.48% CP BeforeHill -4.61% 2.35% -5.58%

Comparing the InHill and With AfterHill (Table 4.1 and 4.2), the latter has better results in all different hills regarding both fuel and time. Some hills are however not included in Table 4.2, since the negative inclination is not high enough to get theHDV to accelerate by activating EcoRoll, and thereby no deceleration is

needed.

It is already shown that the function In Hill is beneficial, unless the HDV does

reach the DHSC set speed. Furthermore, adding AfterHill increases the savings

forHDVthat reaches higher speed than the givenCCspeed, which is the case when

reaching theDHSCset speed. It can be concluded that Only AfterHill should be used

for hills where theDHSCset speed is reached. The results are presented in Table 4.3.

Comparing the results for the hill with the inclination -1.48% and regular length, the best results where, as expected, reached using Only AfterHill, regarding fuel saving and total cost.

As mentioned earlier, an avoidable braking is considered as a loss, which is desired to be reduced. By decreasing speed before a steep hill, braking can be decreased or even eliminated. By that, fuel is saved at the expense of increased travel time. However, comparing Table 4.3 and 4.4, shows that the total cost is reduced radically. With respect to fuel consumption, it is even better to reduce the speed by activating EcoRoll before a hill rather than reducing it with a closed powertrain. However, activating EcoRoll in the segment BeforeHill shows better results when it is combined with With AfterHill than Only AfterHill, since for the studied hill, the DHSC set speed is not reached when the speed is reduced before

hill.

Another observation from the results is that neither the time nor the fuel savings are linear for a linear variation of inclination. This is mainly caused by the HDV

motion being affected by two components of the environmental forces that are non-linearly dependent on the speed, see the equations (3.3) and (3.4). Therefore, the acceleration is nonlinearly varied during a slope with constant inclination, meaning that the variations of the speed and the time are also nonlinear. As mentioned

(40)

CHAPTER 4. PRE-STUDY OF ECOROLL earlier, for open powertrain, the fuel consumption decreases with increased speed. Therefore, the fuel savings are also nonlinear.

(41)

Chapter 5

Conventional EcoRoll

The decisions of the conventional EcoRoll are rules based on prevailing conditions. The strategy is described in this chapter, followed by simulation results. The simula-tions are done for twoHDVs, a conventionalHDVand one utilizing EcoRoll

function-ality. The HDVs were simulated to drive through two different roads; the highway

from Södertälje to Norrköping and from Linköping to Jönköping, in Sweden. All val-ues of the parameters of the control strategy are given in Table A.2 in Appendix A.

5.1 Control Strategy

The control strategy consists of rules for when EcoRoll should be activated. The rules considers the results from the feasibility study, based on prevailing road in-clination and vehicle speed, and on estimated acceleration of the HDV with open

powertrain.

The strategy has two different main states; Active and Inactive, see Figure 5.1. The active state has three sub states for different conditions; Accelerating Hill,

Decelerating Hill and After Hill, see Figure 5.2. The EcoRoll is active when being

in any of these sub states. The differences between the states are the conditions defined by transitions and from which state it is able. However, all conditions can be defined by only using the main states with two transitions, but dividing in sub states provides better visualization and easier understanding. The transitions and their conditions are described below. However, the signals that the controller is based on should be described first.

5.1.1 Input Signals to the Controller

This study is meant to indicate the achievable result when driving with EcoRoll in reality. Therefore, the signals into the controller should be available in the ECU

where it could be implemented in the future. The controller is based on five signals; current speed,CCset speed,DHSCset speed, inclination and estimated acceleration

(42)

CHAPTER 5. CONVENTIONAL ECOROLL

Figure 5.1: The main states of the controller; Active and Inactive, i.e., opened or closed powertrain. A transition occur when some specific conditions regarding speed and acceleration are fulfilled.

available through different systems. However, the acceleration is not known, nor measured. Even though theHDVis equipped with an accelerometer, the

accelerom-eter measures the current acceleration regardless of the powertrain being open or close. The desired acceleration is for aHDV with open powertrain aER. Therefore,

it should be estimated, e.g., by dividing the environmental forces by the vehicle mass,

aER= −

Froll+ Fair+ Fgrav Mv

. (5.1)

However, the aim is that the controller should be ready for implementation in a

HDV. Therefore, the estimation of the acceleration should either be based on an

estimation of the components in (5.1) or on measured signals. A drawback with the first case is that it requires accurate models of the environmental forces.

For the second case, the current acceleration can be measured by an accelerome-ter. Thus, the total force Ftotacting on theHDVcan be calculated. Ftotcorresponds

to the force caused by the engine and brake (Ftrac) with subtraction of the

environ-mental forces,

Ftot = Ftrac− Fair− Froll− Fgrav. (5.2)

The total environmental forces can then be estimated by subtracting Ftrac from Ftot. Hence, the estimation of aERis finally obtained by dividing the resulting force

by Mv,

aER=

Ftot− Ftrac Mv

. (5.3)

This method is used since all parameters in (5.3), i.e., Ftot, Ftrac, and Mv can

either be measured or estimated by using already existing signals in HDVs. This

(43)

5.1. CONTROL STRATEGY

Figure 5.2: The state Active consists of several sub states depending on the prefer-ences of the current circumstances. For instance, the sub state After Hill is entered only if conditions in Algorithm 3 are fulfilled and the prior active sub state is

Ac-celerating Hill. This, since the powertrain is desired to still be open only after an

accelerating hill. The solid lines corresponds to the main required transitions, where the dashed and dotted lines are added to improve the controller, taking in account the interaction between the states.

5.1.2 Accelerating Hill

An accelerating hill is a hill where the vehicle accelerates if the EcoRoll is activated. Although, this is desirable since theHDVaccelerates with low fuel consumption, the

limitation is when the DHSC set speed is reached. However, a maximum speed

vmax with margin from theDHSCset speed vDHSCmargin is declared. It is no longer

possible to activate EcoRoll, if the HDV reaches the maximum speed and has a

positive acceleration.

The margin is set for two reasons. The first one is for the time elapse to engaging the gear, before reaching the DHSC set speed. The second is since it is, according

(44)

CHAPTER 5. CONVENTIONAL ECOROLL that the DHSCset speed is going to be reached. However, Look-Ahead data is not

available for this control strategy, but it can be considered that the risk is high to reachDHSCset speed.

Imagine a large uphill which lets the HDV decelerate despite that maximum

torque is delivered. The speed can decelerate radically, depending on engine’s power rate and mass of theHDV. So, if there is no lower speed limit, and an accelerating

hill is followed by this large uphill, the EcoRoll will be activated no matter how small this acceleration could be. This gives longer travel time. Therefore, a lower speed limit, vmin, is another condition for this state.

The conditions for Accelerating Hill are summarized in Algorithm 1. Algorithm 1 Accelerating hill

vmax = vDHSCset− vDHSCmargin

vmin = vCCset− vCCmargin

if aER≥0 && v < vmax && v > vmin then

return open powertrain end if

5.1.3 Decelerating Hill

If the downhill is not steep enough, the HDVwill decelerate with open powertrain.

However, it could still be advantageous to utilize the benefit of EcoRoll’s fuel saving. It is proved in the pre-study that fuel saving is more than the time loss, according to the cost function (4.10). However, the conditions have to be strict since much speed reduction is not preferable. Therefore, a maximum deceleration limit decmax is set.

Furthermore, unlike the Accelerating Hill, the negative inclination condition should be explicitly defined. This, since a deceleration could be caused by an uphill, which is not preferred if the initial speed before entering the Active state v0 is below theCC

set speed. However, if the state is already entered, the speed is allowed to decrease below the CC set speed by a margin of v_CC_margin. The conditions for Decelerating

Hill are finally summarized in Algorithm 2.

Algorithm 2 Decelerating hill

if aER<0 && aER> decmax && α < 0 && v0 > vCCset then

5.1.4 After Hill

At the end of an accelerating hill, the speed of theHDV is above theCC set speed.

So, deceleration is necessary after the hill. From the pre-study it was shown that it is always (unless the speed is not too low, which usually is not the case on highways)

(45)

5.1. CONTROL STRATEGY

more advantageous to decelerate by activating EcoRoll than by engine braking, both regarding travel time and fuel consumption. The condition for this sub state is then an estimated negative acceleration with open powertrain, and speed above the CC

set speed, see Algorithm 3.

Algorithm 3 After hill

if aER<0 && v > vCCset then

5.1.5 A Low-Pass Filter

The gearbox requires time to engage and disengage the gear. Therefore, a lower time limit, tchangemin, is required to remain within a main state. So, if the conditions to

activate EcoRoll are fulfilled, the controller will first check that it has been located in Inactive mode for at least tchangemin. Since, the state Active consists of several

sub states, another sub state, Wait, is placed there, in order to collect from other sub states that requests to deactivate EcoRoll, and then check that the lower time limit has elapsed. This time limit filters rapid decision changes, which may happen due to that theHDV does not know anything about the road ahead.

5.1.6 The Transitions between the States

The transition from the state Inactive to Active, consists unfortunately of two parts, see Figure 5.1 and 5.2. This depends on whether Algorithm 1 or Algorithm 2 that has fulfilled conditions. Therefore, there are two inputs in the detailed Fig-ure 5.2. Furthermore, it can be seen, in same figFig-ure, that all sub states have to go through the sub state Wait if closed powertrain is demanded, which was described in Section 5.1.5. As mentioned before, it is beneficial to activate EcoRoll after an accelerating hill. Therefore, a transition goes from Accelerating Hill to After Hill.

Besides the solid arrows, that are the basic part of the controller, there are dashed and dotted transition arrows. The dashed arrows were added, between

Accelerating Hill and Decelerating Hill, since inclination of a hill may vary such

that some parts of it accelerate theHDV while other parts decelerate it. Moreover,

the dotted arrows are added from the sub state Wait to give the controller ability to undo a decision when it occurs within tchangemin. This, since it is unfavorable,

from the driving performance point, to instead let the powertrain close first and open again after additional tchangemin, when the controller already wanted to undo

the decision of closing powertrain within the first tchangemin.

Note that conditions for some transitions may occur simultaneously. Therefore, a prioritization of which transition is preferred if this happens should be chosen. This is given by the number in the beginning of the arrow in Figure 5.2, where 1

(46)

CHAPTER 5. CONVENTIONAL ECOROLL gives the highest priority. The priorities are chosen such that EcoRoll is activated maximum possible time.

5.2 Simulations

The control strategy that has been described in the previous section is used in simulation of anHDVin two roads; Södertälje-Norrköping and Linköping-Jönköping.

This is compared to a reference HDV, and based on the cost function (4.10) the

results are presented in Table 5.1. The results of Södertälje-Norrköping are amazing. However, even though simulations on Linköping-Jönköping also give positive results, the difference between the results are large.

The simulations can be compared in Figure 5.3 and 5.4. The figures show the speed of bothHDVs(the one with EcoRoll and the reference), the road topography

and the control signal. Value one on the control signal indicates a request to open the powertrain. Comparing the speed graphs and the road topographies, it can be seen that the large difference in results can be explained by the amount of slopes. Södertälje-Norrköping has a lot more slopes that can be utilized by EcoRoll. Further analyze of road topography is made in Chapter 7. Moreover, different disturbances are added to analyze the sensitivity in Chapter 8. Prior to these, another strategy is developed in the next chapter, utilizing Look-Ahead data in order to improve the results.

Table 5.1: The simulation results using the conventional EcoRoll.

Road Cost Time Fuel

Södertälje-Norrköping -4.2648% -0.2943% -3.4164% Linköping-Jönköping -0.8622% -0.1572% -0.6157%

(47)

5.2. SIMULATIONS 0 2 4 6 8 10 12 x 104 60 80 100 distance [m] speed [km/h] v_CC set v_DHSC set HDV with EcoRoll HDV with CC 0 2 4 6 8 10 12 x 104 0 50 100 150 distance [m] altitude [m] 0 2 4 6 8 10 12 x 104 0 0.5 1 distance [m] control signal

Figure 5.3: Simulations of the conventional EcoRoll, traveling from Södertälje to Norrköping. 0 2 4 6 8 10 12 14 x 104 70 80 90 100 distance [m] speed [km/h] v_CC set v_DHSC set HDV with EcoRoll HDV with CC 0 2 4 6 8 10 12 14 x 104 0 100 200 300 distance [m] altitude [m] 0 2 4 6 8 10 12 14 x 104 0 0.5 1 distance [m] control signal

Figure 5.4: Simulations of the conventional EcoRoll, traveling from Linköping to Jönköping.

(48)

(49)

Chapter 6

EcoRoll utilizing Look-Ahead

The main reason to develop a strategy that utilizes Look-Ahead data, is the ability to identify hills whereDHSCset speed can be reached with activated EcoRoll. This,

in order to minimize fuel consumption even further.

To get results that are comparable with the previous strategy, the structure of the controller and its parameters will be held the same as much as possible. However, some modifications are done, where the main change is the speed prediction.

6.1 Control Strategy

This strategy predicts the speed of the downhill ahead. Based on the predicted achievable speed, the controller makes a decision; whether to open the powertrain or not. The main states are the same as before, see Figure 5.1, but with modifications within the states. However, the speed prediction is calculated in parallel to these states. The prediction is described in the following section, followed by presenting the modifications of the algorithm.

6.1.1 Speed Prediction

In addition to the required signals that where described in Section 5.1.1, the data of the topography ahead should be available, and a signal of the current position received from, e.g., a GPS. Based on the known current position, inclination and

the inclination of the road ahead, the environmental forces that will act on theHDV

can be predicted using models like in (3.2), (3.3) and (3.4), to calculate aER by

using (5.1). However, the available data for the topography is discontinuous, with a distance of ∆p between known points. It is assumed that ∆p is small enough, giving neglected dynamics between two positions. Therefore, the acceleration between two positions is assumed to be constant. With this background, the speed for next position, p + (n + 1) ∆p, is given by

THE IMPACT OF ECOROLL ON FUEL CONSUMPTION - USING LOOK AHEAD

THE IMPACT OF ECOROLL ON FUEL

CONSUMPTION - USING LOOK AHEAD

MUSTAFA ABDUL-RASOOL

Master’s Degree Project

Stockholm, Sweden June 2011

THE IMPACT OF ECOROLL ON FUEL

CONSUMPTION - USING LOOK AHEAD

Abstract

Referat

Acknowledgements

Contents

List of Abbreviations

List of Symbols

Chapter 1

Introduction

Chapter 2

Background

2.1

EcoRoll

2.2

Look-Ahead

2.3

Problem Statement

2.4

Related Work

Chapter 3

Vehicle Model

3.1

Basic Model of Powertrain

3.2

Important Parts of the Powertrain

3.3

Modeling Software

Chapter 4

Pre-study of EcoRoll

4.1

Possibility of Implementation

4.2

Safety Aspects

4.3

Road Topography and Strategies

4.4

Quadratic Cost Function

4.5

Simulation Results

4.6

Discussions on the Feasibility Study

Chapter 5

Conventional EcoRoll

5.1

Control Strategy

5.2

Simulations

Chapter 6

EcoRoll utilizing Look-Ahead

6.1

Control Strategy