Preprints, 8th 8th IFAC IFAC International International Symposium Symposium on on Preprints, Preprints, 8th IFAC International Symposium on Advances in Control Preprints, IFAC International on Advances 8th in Automotive Automotive Control Symposium Preprints, 8th IFAC International Symposium on Advances Automotive Control June 19-23,in 2016. Norrköping, Sweden Available online at www.sciencedirect.com Advances in Automotive Control June 19-23, 2016. Norrköping, Sweden Advances in2016. Automotive Control June 19-23, Norrköping, Sweden June June 19-23, 19-23, 2016. 2016. Norrköping, Norrköping, Sweden Sweden ScienceDirect IFAC-PapersOnLine 49-11 (2016) 657–664 Cycle Beating -- An Analysis of the Cycle Beating An Analysis of Cycle Beating An Analysis of the Cycle Beating - An Vehicle AnalysisTesting of the the Boundaries During Boundaries During Vehicle Testing Boundaries During Vehicle Testing Boundaries During Vehicle Testing ∗ ∗ Kristoffer Ekberg Ekberg ∗∗ ,, Lars Lars Eriksson Eriksson ∗∗ and and Martin Martin Sivertsson Sivertsson ∗∗∗ Kristoffer ∗ , Lars Eriksson ∗ and Martin Sivertsson ∗ Kristoffer Ekberg ∗ , Lars Eriksson ∗ and Martin Sivertsson ∗ Kristoffer Ekberg Kristoffer Ekberg , Lars Eriksson and Martin Sivertsson ∗ ∗ Vehicular Systems, Dept. of Electrical Engineering, Linköping ∗ Vehicular Systems, Dept. of Electrical Engineering, Linköping ∗ Vehicular Systems, Dept. of Electrical Engineering, Linköping ∗ University, Vehicular Dept. of Engineering, Linköping SE-581 83 Linköping, Sweden, {kristoffer.ekberg, University, SE-581 83 Linköping, Sweden, {kristoffer.ekberg, Vehicular Systems, Systems, Dept. of Electrical Electrical Engineering, Linköping University, SE-581 83 Linköping, Sweden, {kristoffer.ekberg, University, SE-581 83 Linköping, Sweden, {kristoffer.ekberg, lars.eriksson}@liu.se, sivertsson.martin@gmail.com lars.eriksson}@liu.se, sivertsson.martin@gmail.com University, SE-581 83 Linköping, Sweden, {kristoffer.ekberg, lars.eriksson}@liu.se, sivertsson.martin@gmail.com sivertsson.martin@gmail.com lars.eriksson}@liu.se, lars.eriksson}@liu.se, sivertsson.martin@gmail.com Abstract: Todays Todays vehicle vehicle industry industry is is strictly strictly controlled controlled by by environmental environmental legislations. legislations. The The Abstract: Abstract: Todays vehicle industry is strictly controlled by environmental legislations. The Abstract: Todays vehicle industry is strictly controlled by environmental legislations. The vehicle industry is spending much money on reducing the fuel consumption and fulfilling the vehicle industry is spending much money on reducing the fuel consumption and fulfilling the Abstract: Todays vehicle industry is strictly controlled by environmental legislations. The vehicle industry is spending much money on reducing the fuel consumption and fulfilling the vehicle industry is spending much money on reducing the fuel consumption and fulfilling the emission requirements to make sales possible in different regions in the world. Before introducing emission requirements to make sales possible in different regions in the world. Before introducing vehicle industry is spending much money on reducing the fuel consumption and fulfilling the emission requirements toitmake sales possible in different regions in thecycles world. Before introducing emission requirements sales possible regions in world. a vehicle on the market, is tested according to standardized driving to specify the vehicle a vehicle vehicle on on the market, market,to itmake is tested tested according to different standardized driving cycles toBefore specifyintroducing the vehicle vehicle emission requirements toit make salesaccording possible in in different regions in the thecycles world. Before introducing a the is to standardized driving to specify apollutant vehicle on on the market, market, it is tested tested according to standardized standardized driving cycles to to from specify the vehicle pollutant emissions and it fuel consumption. These cycles allow allowdriving some deviation deviation from thethe reference emissions and fuel consumption. These cycles some the reference a vehicle the is according to cycles specify the vehicle pollutant emissions and fuel consumption. These cycles allow some deviation from the reference pollutant emissions fuel consumption. These cycles deviation reference vehicle speed during tests, e.g. NEDC allows deviations of ±2 km/h and ±1 s. This paper vehicle speed speed duringand tests, e.g. NEDC allows allows deviations of some ±2 km/h km/h and from ±1 s. s.the This paper pollutant emissions and fuel e.g. consumption. Thesedeviations cycles allow allow some deviation from the reference vehicle during tests, NEDC of ±2 and ±1 This paper vehicle speed during tests, e.g. NEDC allows deviations of ±2 km/h and ±1 s. This paper uses dynamic programming to find fuel optimal velocity profiles, given the allowed deviations uses dynamic programming to find fuel optimal velocity profiles, given the allowed deviations vehicle speed during tests, e.g. NEDC allows deviations of ±2 km/h and ±1 s. This paper uses dynamic programming to find fuel optimal velocity profiles, given the allowed deviations uses programming to optimal velocity profiles, the deviations of ±2 km/h and ±1 from reference speed during drive cycle test. By taking advantage of of ±2 ±2dynamic km/h and and ±1 sss from from reference reference speed during drive cycle cycle test.given By taking taking advantage of the the uses dynamic programming to find find fuel fuel optimal velocity profiles, given the allowed allowed deviations of km/h ±1 speed during drive test. By advantage of of ±2 km/h km/h and ±1 ±1 s from from reference speed speed during drive cycle cycle test. By taking taking advantage of the the allowed deviation, the fuel consumption consumption canduring be reduced reduced by up uptest. to 16.56 16.56 % according according to model model allowed deviation, the fuel can be by to % to of ±2 and s reference drive By advantage of the allowed deviation, the fuel consumption can be reduced by up to 16.56 % according to model allowed deviation, the consumption can reduced up to % to results, NEDC if gear selections are unrestricted (i.e. using automatic gearbox), and results, running running NEDC if gear selections are be unrestricted (i.e. using automatic gearbox), and allowed deviation, the fuel fuel consumption can be reduced by by(i.e. up using to 16.56 16.56 % according according to model model results, running NEDC if gear selections are unrestricted automatic gearbox), and results, running NEDC if gear selections are unrestricted (i.e. using automatic gearbox), and up to 5.90 % if changing gears according to the specifications in the drive cycle. Two different up to 5.90 % if changing gears according to the specifications in the drive cycle. Two different results, running NEDC if gear selections are unrestricted (i.e. using automatic gearbox), and up to 5.90 % if changing gears according to the specifications in the drive cycle. Two different up to if gears to specifications in drive optimization goals are investigated, minimum amount of mass fuel consumed and best mileage. optimization goals are investigated, investigated, minimum amount of mass mass fuel fuel consumed and Two best different mileage. up to 5.90 5.90 % % goals if changing changing gears according according to the the specifications in the the drive cycle. cycle. Two different optimization are minimum amount of consumed and best optimization goals goals are are investigated, investigated, minimum minimum amount amount of of mass mass fuel fuel consumed consumed and and best best mileage. mileage. optimization mileage. © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Keywords: Keywords: Dynamic Dynamic Programming, Programming, Cycle Cycle Beating. Beating. Keywords: Dynamic Programming, Cycle Beating. Keywords: Dynamic Programming, Cycle Keywords: Dynamic Programming, Cycle Beating. Beating. 1. E/ECE/TRANS/505/Rev.1/Add.82/Rev.5 1. INTRODUCTION INTRODUCTION E/ECE/TRANS/505/Rev.1/Add.82/Rev.5 (2015), (2015), the the drdr1. INTRODUCTION E/ECE/TRANS/505/Rev.1/Add.82/Rev.5 (2015), the dr1. E/ECE/TRANS/505/Rev.1/Add.82/Rev.5 (2015), the drive cycle is displayed in Figure 1. The NEDC defines 1. INTRODUCTION INTRODUCTION ive cycle is displayed in Figure 1. The NEDC defines E/ECE/TRANS/505/Rev.1/Add.82/Rev.5 (2015), the drive speed cycle is is displayedthat in isFigure Figure 1. The The NEDC defines ive cycle displayed in 1. NEDC defines the trajectory supposed to followed and the speed speed trajectory that isFigure supposed to be be followed and ive cycle is displayedthat in is 1. The NEDC defines Todays concern of the environment has resulted in difthe trajectory supposed to be followed and Todays concern of the environment has resulted in difthe speed trajectory that is supposed to be followed and which gears that should be selected. The cycle consists Todays concern of the environment has resulted in difwhich gears that should be selected. The cycle consists the speed trajectory that is supposed to be followed and Todays concern of the environment has resulted in different standards and regulations concerning vehicle emiswhich gears that should be selected. The cycle consists ferent standards standards andthe regulations concerning vehicleinemisemisTodays concern of environment has resulted dif- which gears that be The consists of standstills and accelerations. ferent and regulations concerning vehicle of constant constant speeds, standstills and constant constant accelerations. which gearsspeeds, that should should be selected. selected. The cycle cycle consists ferent standards and concerning vehicle emissions, regulations are for and deof constant speeds, standstills and constant accelerations. sions, these these regulations are drivers drivers for improving improving and de- of ferent standards and regulations regulations concerning vehicle emisconstant speeds, standstills and constant accelerations. The behavior of the cycle can be hard to follow precisely in sions, these regulations are drivers for improving and deThe behavior of the cycle can be hard to follow precisely in of constant speeds, standstills and constant accelerations. sions, these regulations are drivers for improving and developing better vehicles. The regulations have a large imThe behavior of the the cycle can be befrom hardregions to follow follow precisely in veloping better vehicles.are Thedrivers regulations have aa large large imsions, these regulations for improving and imde- The behavior of cycle can hard to precisely in a vehicle due to rapid changes with constant veloping better vehicles. The regulations have a vehicle due to rapid changes from regions with constant The behavior of the cycle can be hard to follow precisely in veloping better vehicles. The regulations have a large impact on the technical development concerning automotive a vehicle vehicle due due to to rapid rapid changes changes from regions regions with constant pact on on the the technical development concerning automotive veloping better vehicles. The regulations have automotive a large im- aacceleration from with speed. The question pact technical development concerning acceleration totoregions regions with constant constant speed. with The constant question aacceleration vehicle dueto rapid changes from regions with constant pact on the technical development concerning automotive industry (Eriksson and Nielsen (2014)). To sell vehicles to regions with constant speed. The question industry (Eriksson and Nielsen (2014)). To sell vehicles pact on the technical development concerning automotive acceleration to regions with constant speed. The question raised here is, how much can much can you gain in industry (Eriksson and Nielsen (2014)). To sell vehicles raised here is, how much can much can you gain in terms terms acceleration to regions with constant speed. The question industry (Eriksson and Nielsen (2014)). To sell vehicles in regions, they have fulfill regional reraised here is, how how much can much much can you you gain in in terms in different different regions,and theyNielsen have to to fulfill the the regional re- raised industry (Eriksson (2014)). To sell vehicles here is, much can can gain terms of fuel consumed by the vehicle, if taking advantage of the in different regions, they have to fulfill the regional reof fuel consumed by the vehicle, if taking advantage of the raised here is, how much can much can you gain in terms in different regions, they have to fulfill the regional requirements regarding how much pollutant emissions the of fuel fuel consumed consumed byinthe the vehicle, if taking taking advantage of the the quirements regarding how have muchtopollutant pollutant emissions the in different regarding regions, they fulfill theemissions regional the re- of by vehicle, if advantage of allowed deviation vehicle speed? The investigation is quirements how much allowed deviation in vehicle speed? The investigation is of fuel consumed by the vehicle, if taking advantage of the quirements regarding how much pollutant emissions the vehicle releases. To fulfill the emission legislations today, allowed deviation in vehicle vehicle speed? The investigation is vehicle releases. releases. To fulfill fulfill the emission legislations today, quirements regarding howthe much pollutant emissions the allowed deviation in speed? The investigation is made using both specified gear changes and freely selected vehicle To emission legislations today, made using both specified gear changes and freely selected allowed deviation in vehicle speed? The investigation is vehicle releases. To fulfill the emission legislations today, electric hybrids are introduced on the market, using both made using both specified gear changes and freely selected electric hybrids are introduced on the market, using both vehicle releases. To introduced fulfill the emission legislations today, both gear selected gears, to the from gear selections on electric hybrids are on the the market, using both made gears, using to investigate investigate the impact impact from and gear freely selections on made using both specified specified gear changes changes and freely selected electric hybrids are on both an machine and engine. By gears, to investigate the impact from gear selections on an electric electric machine and aaa combustion combustion engine. using By downdownelectric hybrids are introduced introduced on the market, market, using both gears, to investigate the impact from gear selections on the savings in terms of fuel consumption. During vehicle an electric machine and combustion engine. By downthe savings in terms of fuel consumption. During vehicle gears, to investigate the impact from gear selections on an electric machine and a combustion engine. By downsizing the internal combustion engine and adding a electhe savings savings in in terms terms ofNEDC fuel consumption. consumption. During vehicle sizing the internal internal combustion engine and and adding elec- the an electric machinecombustion and a combustion engine. By aadownfuel During simulations, full (full low speed sizing the engine adding elecsimulations, both fullof NEDC (full cycle) cycle) and and lowvehicle speed the savings inboth terms ofNEDC fuel consumption. During vehicle sizing the internal combustion engine and adding aa electric machine with aa battery, energy can be regenerated simulations, both full (full cycle) and low speed tric machine with battery, energy can be regenerated sizing the internal combustion engine and adding elecsimulations, both full NEDC (full cycle) and low section of NEDC (city cycle) are investigated. Low speed tric machine with a battery, energy can be regenerated section of NEDC (city cycle) are investigated. Low speed simulations, both full NEDC (full cycle) and low tric machine with aa Hybrid battery, energy can be regenerated and used if needed. Electric Vehicles (HEV) are section of NEDC (city cycle) are investigated. Low speed and used if needed. Hybrid Electric Vehicles (HEV) are tric machine with battery, energy can be regenerated of NEDC are investigated. Low is to visualize different to and used used if needed. needed. Hybrid of Electric Vehicles (HEV) are section is displayed displayed to cycle) visualize the different solutions solutions to section of NEDC (city (city cycle) arethe investigated. Low speed speed and if Hybrid Electric Vehicles (HEV) are used to the released emissions. Wang section is displayed to visualize the different solutions to used used to reduce reduce the amount amount of released emissions. Wang and if needed. Hybrid of Electric Vehicles (HEV) are section is displayed to visualize the different solutions to the problem, depending on which optimization goal that is used to reduce the amount released emissions. Wang the problem, depending on which optimization goal that is section is displayed to visualize the different solutions to used to reduce the amount of released emissions. Wang and Lukic (2012) have used dynamic programing to find the problem, problem, depending on which optimization goal that is and Lukic Lukic (2012) have used of dynamic programing toWang find the used to reduce thehave amount released emissions.to depending on which optimization goal that is selected and if choosing gears according to cycle or freely and (2012) used dynamic programing find selected and if choosing gears according to cycle or freely the problem, depending on which optimization goal that is and Lukic (2012) have used dynamic programing to find optimal control strategies for different configurations of selected and and if choosing choosing gears according to cycle cycle or freely freely optimal control strategies fordynamic differentprograming configurations of selected and Lukic (2012)strategies have usedfor to find if to or selecting gears. The cycle is in 1, optimal control different configurations of selectingand gears. The full fullgears cycleaccording is displayed displayed in Figure Figure 1, selected if choosing gears according to cycle or freely optimal control for different of HEV’s. programing been used selecting gears. The full cycle is displayed in Figure 1, HEV’s. Dynamic Dynamic programing have beenconfigurations used by by several several optimal control strategies strategies for have different configurations of selecting gears. The full cycle is displayed in Figure 1, one city cycle is displayed in Figure 2, in both figures the HEV’s. Dynamic programing have been used by several one city cycle is displayed in Figure 2, in both figures the selecting gears. The full cycle is displayed in Figure 1, HEV’s. Dynamic programing have been used by several authors (for example Luu et al. (2010) and Wollaeger one city city cycle cycle is allowed displayed in Figure Figure 2, in in both both figuresgear the authors Dynamic (for example example Luu et et have al. (2010) (2010) and by Wollaeger HEV’s. programing been used several one is displayed in 2, figures the reference speed, speed deviation specified authors (for Luu al. and Wollaeger reference speed, allowed speed deviation and specified gear one city cycle is allowed displayed in Figure 2, inand both figuresgear the authors (for example Luu et al. (2010) and Wollaeger et al. (2012)) to solve vehicle optimization problems conreference speed, speed deviation and specified et al. (2012)) to solve vehicle optimization problems conauthors (for example Luu et al. (2010) and Wollaeger speed, allowed speed deviation and specified gear selections are displayed. et al. al. (2012)) (2012)) to solve vehicle vehicle optimization problems con- reference selections are displayed. reference speed, allowed speed deviation and specified gear et to solve optimization problems concerning speed trajectories and fuel consumption. Mensing selections are are displayed. displayed. cerning speed to trajectories and optimization fuel consumption. consumption. Mensing et al. (2012)) solve vehicle problems con- selections cerning speed trajectories and fuel Mensing cerning speed trajectories and fuel et dynamic to optimal et al. al. (2011) (2011) uses dynamic programing to find find Mensing optimal selections are displayed. cerning speed uses trajectories andprograming fuel consumption. consumption. Mensing et al. (2011) uses dynamic programing to find optimal et al. uses programing to optimal ECO-driving Vehicles are according to 1.1 Contributions Contributions ECO-driving trajectories. Vehicles are tested tested according to 1.1 et al. (2011) (2011) trajectories. uses dynamic dynamic programing to find find optimal 1.1 Contributions ECO-driving trajectories. Vehicles are tested according to 1.1 ECO-driving trajectories. Vehicles are tested according to standardized drive cycles to make it possible to compare 1.1 Contributions Contributions standardized drive cycles to make it possible to compare ECO-driving trajectories. Vehicles are tested according to standardized drive cycles to to make it possible possible toincompare compare standardized drive cycles it different manufactures. Vehicles tested specific This paper paper uses uses aaa dynamic dynamic programing programing approach approach to to show show different vehicle vehicle manufactures. Vehicles testedto incompare specific This standardized drive cycles to make make it possible toin This paper uses dynamic programing approach to show different vehicle manufactures. Vehicles tested specific This paper uses aa dynamic programing approach to show different vehicle manufactures. Vehicles tested in specific how fuel economy is affected by smart driving, within drive cycles can today be driven by robots to achieve how fuel economy is affected by smart driving, within drive cycles can today be driven by robots to achieve This paper uses dynamic programing approach to show different vehicle manufactures. Vehicles tested in specific how fuel economy is affected by smart driving, within drive cycles can today be driven by robots to achieve fuel by within drive cycles today robots to the limits the drive The contribution is good following, but NEDC (New European Drive the legal legal limits of of is theaffected drive cycle. cycle. The driving, contribution is good reference reference following, but driven NEDC by (New European Drive how how fuel economy economy is affected by smart smart driving, within drive cycles can can today be be driven by robots to achieve achieve the legal limits of the drive cycle. The contribution is good reference following, but NEDC (New European Drive the legal limits of the drive cycle. The contribution is good reference following, but NEDC (New European Drive a quantification how much the fuel consumption can Cycle) allows some deviation in vehicle speed during vehia quantification how much the fuel consumption can Cycle) allows some deviation in vehicle speed during vehithe legal limits of the drive cycle. The contribution is good reference following, but NEDC (New European Drive a quantification quantification of how how much the the fuel consumption can Cycle) allows some deviation deviation in vehicle vehicle speed during vehi- abe of much consumption can Cycle) allows some in during vehireduced taking advantage of the deviation cle Regulations concerning the drive NEDC bequantification reduced by by taking advantage of fuel the allowed allowed deviation cle testing. testing. Regulations concerning the speed drive cycle cycle NEDC of how much the fuel consumption can Cycle) allows some deviation in vehicle speed during vehi- abe reduced by taking advantage of the allowed deviation cle testing. Regulations concerning the drive cycle NEDC be reduced by taking advantage of the allowed deviation cle testing. Regulations concerning the drive cycle NEDC speed during drive paper allows aa deviation in speed following of km/h in reference reference speed during drive cycle cycle testing, the paper allows deviation in reference reference speedthe following of ±2 ±2NEDC km/h in be reduced by taking advantage of thetesting, allowed the deviation cle testing. Regulations concerning drive cycle in reference speed during drive cycle testing, the paper allows a deviation in reference speed following of ±2 km/h speed drive testing, paper allows in following of km/h also show gear if and ±1 during decelerations constant alsoreference show optimal optimal gear selections selections if gears gears arethe selected and ±1aa sssdeviation during accelerations, accelerations, decelerations and constant in reference speed during during drive cycle cycle testing,are theselected paper allows deviation in reference reference speed speed followingand of ±2 ±2 km/h in also show optimal gear selections if gears are selected and ±1 during accelerations, decelerations and constant also show optimal gear selections if gears are selected and ±1 s during accelerations, decelerations and constant freely. Two different optimization goals are analyzed, High speed traveling, see E/ECE/324/Rev.1/Add.82/Rev.5freely. Two different optimization goals are analyzed, High speed traveling, see E/ECE/324/Rev.1/Add.82/Rev.5also show optimal gear selections if gears are selected and ±1 s during accelerations, decelerations and constant freely. Two Two different different optimization optimization goals goals are are analyzed, analyzed, High High speed traveling, traveling, see see E/ECE/324/Rev.1/Add.82/Rev.5E/ECE/324/Rev.1/Add.82/Rev.5- freely. speed speed traveling, see E/ECE/324/Rev.1/Add.82/Rev.5- freely. Two different optimization goals are analyzed, High Copyright © 2016, 2016 669 2405-8963 © IFAC (International Federation of Automatic Control) Copyright 2016 IFAC IFAC 669 Hosting by Elsevier Ltd. All rights reserved. Copyright © 2016 IFAC 669 Copyright © 2016 IFAC 669 Peer review under responsibility of International Federation of Automatic Copyright © 2016 IFAC 669Control. 10.1016/j.ifacol.2016.08.095 IFAC AAC 2016 658 June 19-23, 2016. Norrköping, Sweden (1) Low Fuel as possible, using manual gearbox with gear selections specified by the drive cycle. (2) High Mileage as possible, using manual gearbox with gear selections specified by the drive cycle. (3) Low Fuel as possible, using automatic gearbox with free possibility to select gears. (4) High Mileage as possible, using automatic gearbox with free possibility to select gears. Cycle speed and Speed limits 150 Velovity [km/h] Kristoffer Ekberg et al. / IFAC-PapersOnLine 49-11 (2016) 657–664 Vref Vmin 100 Vmax 50 0 0 200 400 6 Gear [-] 600 800 1000 1200 Time [s] Gear selections 2.1 Vehicle model To calculate the fuel consumption for a specific vehicle driving according to the drive cycle, a vehicle model is implemented. The model calculates the amount of fuel needed for a certain energy requirement at the wheels. During simulations, vehicle data according to Table 1 is used. The gearing setup γ is used for both the specified and unspecified gear changes. 4 2 0 0 200 400 600 800 1000 1200 Time [s] Fig. 1. Full NEDC, displays speed in km/h, time in seconds and gear number. Speed trajectory is displayed in top figure and which gears to select is displayed in bottom figure. Gear number 0 corresponds to neutral gear. Cycle speed and Speed limits Velovity [km/h] 60 Vref Vmin 40 Vmax 20 0 0 50 3 Gear [-] 100 150 200 150 200 Time [s] Gear selections 2 1 0 0 50 100 Time [s] Fig. 2. Section from NEDC, city cycle, displays speed in km/h, time in seconds and gear number. Speed trajectory is displayed in top figure and which gears to select is displayed in bottom figure. Gear number 0 corresponds to neutral gear. Mileage and Low Fuel, with and without restrictions on the gear changes made in the vehicle. Drive cycle simulations are performed on both low speed part of NEDC cycle and full NEDC cycle. 2. CYCLE BEATING Table 1. Vehicle data used during simulations. Parameter Value Unit ρair ρf uel Cd Cr A g m rw Jw γ ηgearbox Je Te,max Ne,max Vd e pme0 Hl 1.18 737.2 0.32 0.015 2.31 9.81 1500 0.30 0.6 0 13.0529 8.1595 5.6651 4.2555 3.2623 0.98 0.2 115 5000 1.497 × 10−3 0.4 1 × 105 44.6 × 106 kg/m3 kg/m3 − − m2 m/s2 kg m kg m2 − − kg m2 N m rpm m3 − Pa J/kg The forces acting on a vehicle traveling on flat road are described in equations (1a), (1b) and (1c) (see Guzzella and Sciarretta (2007)): 1 (1a) Fdrag = ρair Cd Av 2 2 Jw (1b) Facc = a(m + 2 ) rw Froll = Cr mg (1c) The energy dissipated during one time step, from time tn to tn+1 , with the assumption that the vehicle acceleration a is constant, is described by: ! tn+1 F vdt (2) E= tn Cycle beating refers to performing well while fulfilling the specifications of the cycle. By using a vehicle model to simulate the forces acting on the vehicle during driving, the required fuel consumption is calculated. This paper uses dynamic programing to find the least fuel consuming driving strategies. The analysis consists of four cases. The cases have different setups and different optimization goals. Optimizing the speed trajectory within the limit of ±2 km/h and ±1 s from reference speed, to achieve as 670 If using equations (1a), (1b) and (1c) together with (2), following energy equations can be derived: 1 2 Edrag = ρair Cd A(vn+1 + vn2 )(vn+1 + vn )∆t (3a) 8 Jw a (3b) Eacc = (m + 2 )(vn+1 + vn )∆t 2 rw 1 Eroll = Cr mg(vn+1 + vn )∆t (3c) 2 IFAC AAC 2016 June 19-23, 2016. Norrköping, Sweden Kristoffer Ekberg et al. / IFAC-PapersOnLine 49-11 (2016) 657–664 Energy required at the wheels is Ereq = Edrag + Eacc + Eroll . 659 ω1 θ1 ω2 θ2 T 2.2 Gearbox and clutch model To extend the analysis of the drive cycle, a gearbox model and clutch model are implemented. These models are used to analyze the use of free gear selections. The models takes following actions into account: (1) Changes in kinetic energy due to engine rotation. (2) Dissipated energy in clutch during gear changes. (3) Start/stop model, restricts the engine speed to 800 rpm during starts and stops. The total amount of energy required by the combustion engine is described in equation (4). Eice = Egearbox + Eengine + Eclutch (4) Energy required by gearbox (Egearbox ), with gearbox efficiency ηgearbox and gear ratio γ (se values in Table 1) is displayed in equation (5). Ereq Egearbox = (5) ηgearbox The kinetic rotational energy inside the engine is included in the analysis, to enable the use of engine braking, the kinetic rotational energy in the engine at current and next time step is expressed in equations (6a) and (6b). Je vcurr 2 , where ωice0 = Eeng0 = ωice0 ginit (6a) 2 rw vnext Je 2 , where ωice1 = Ggrid (6b) Eeng1 = ωice1 2 rw Where Ggrid represents all possible gear changes from the current time step with gear, ginit to next time step. The cost to perform a gear change, is both the change in engine rotational speed (which can be both positive and negative) and the energy dissipated in the clutch. The kinetic energy needed to change the engine speed from current time step to next time step (Eengine ) is described in equation (7). Eengine = Eeng1 − Eeng0 (7) The energy dissipated in the clutch (Eclutch ) is described in equation (14). Eclutch will always be positive, since the clutch, independent of if there is a upshift or downshift maneuver, will dissipate heat when performing clutch lock maneuver. A graphical representation of the clutch model is displayed in Figure 3. The description of the energy dissipated in the clutch during a clutch maneuver is described in Bhandari (2007). The rotational speeds in the clutch can be described by the incoming and outgoing shaft torques, see equation (8). T1 T2 θ̈1 = and θ̈2 = (8) J1 J2 The rotational speeds θ̇i are received by integrating the above functions from time t = 0 to time t = tclutch . The initial rotational speeds att t = 0 are ω1 and ω2 . ! tclutch T1 T1 θ̇1 = dt = tclutch + ω1 (9a) J J1 1 !0 tclutch T2 T2 dt = tclutch + ω2 (9b) θ̇2 = J J2 2 0 671 J1 J2 Fig. 3. Clutch discs, index 1 corresponds to disc connected to driveline, index 2 corresponds to engine disc. The torque T indicates the torque delivered between the two clutch discs. The rotational speed difference between the incoming and outgoing disc in the clutch is described as: ∆θ̇ = θ̇1 − θ̇2 =⇒ T2 T1 tclutch + ω1 − tclutch − ω2 ∆θ̇ = J1 J2 Torque balance T = −T1 = T2 gives: J1 + J2 )T tclutch + ω1 − ω2 ∆θ̇ = −( J1 J2 The clutch lockup is completed when the rotational difference ∆θ̇ equals 0, equation (11) gives: J1 + J 2 0 = −( )T tclutch + ω1 − ω2 =⇒ J1 J2 ω 1 − ω2 J 1 J2 tclutch = T J1 + J2 (10a) (10b) (11) speed (12a) (12b) The energy dissipated in the clutch during a clutch lockup is described in equation (13b). ! tclutch T ∆θ̇dt =⇒ (13a) Eclutch = 0 Eclutch 1 J 1 J2 = (ω1 − ω2 )2 2 J1 + J2 (13b) 2 By introducing J1 = Jv = γ12 (mrw + Jw ), J2 = Je . A gear change is assumed to take place during one time step, ∆t, therefore the rotational speeds ω1 = ωice0 and ω2 = ωice1 , when a gear change is performed. If no gear change is performed, Eclutch = 0. " 0 if ∆θ̇ = 0 Eclutch = 1 Jv Je (14) 2 (ω − ω ) if ∆θ̇ ̸= 0 ice0 ice1 2 Jv +Je 2.3 Combustion engine model Total mass fuel consumed during one time step can be calculated using an engine model based on Willans approximation (see Guzzella and Sciarretta (2007)), the mass fuel consumed during one time step is described in equation (15). wice pme0 Vd ∆mf = (Tice + + Je ω̇ice )∆t (15) Hl e 4π Equation (15) is rewritten with equation (16) to form the expression in equation (17a). In both equations (17a) and (17b) the inertia Je in equation (15) is included in Eice . The mass fuel consumed during one time step from tn to tn+1 , is used to form the cost function in the dynamic programming algorithm. IFAC AAC 2016 660 June 19-23, 2016. Norrköping, Sweden Kristoffer Ekberg et al. / IFAC-PapersOnLine 49-11 (2016) 657–664 ωice1 − ωice0 2ω̇ice , where ω̇ice = 2 2 ωice1 − ωice0 ∆t (16) During starts, stops and standstills, the engine speed ωice0 and ωice1 are restricted to a constant speed of 800 rpm. During starting maneuvers, the rotational speed is kept at 800 rpm until the rotational speeds of the engine and gearbox (clutch incoming and outgoing rotational speeds) equals each other. This restriction is implemented to include the energy dissipated during starting maneuver when clutch is slipping and first gear is engaged. 2 w2 − wice0 pme0 Vd ∆mf = ice1 ) (Acceleration) (Tice + Hl e2ẇice 4π (17a) wice0 pme0 Vd (Tice + )∆t (Constant speed) ∆mf = Hl e 4π (17b) Tice = Eice 2.4 Dynamic Programing Solving Optimization problem The drive cycle is driven with the specified gearbox and clutch model in all four cases, two of these where the gears are selected freely by the optimization, and two where the gear changes are according to the specified drive cycle. Dynamic programing (Bertsekas (2000)) is used to find the best driving pattern to reduce the fuel consumption as much as possible.Two different optimization goals are investigated, one where the total mass fuel is minimized and one where the mean fuel consumption is minimized. When the free gear selection mode is simulated, the vehicle is free to select gears as long as the cycle requirements and vehicle restrictions are fulfilled. The problem size changes from 1D to 2D, when changing from specified gear changes to free gear changes, a graphical representation of the different gear and speed possibilities are displayed in Figure 4. A backward algorithm is used, solving the problem from the final states at time tend to the states at starting time tstart . The restrictions in the vehicle model are listed in Table 2. Table 2. Constraints on the vehicle model. Constraints γ ̸= 0 if vavg > 0 ω̇e < ω̇e,max Tice < Tice,max Ne ≤ Ne,max If γωw < ωe,min =⇒ γ = 1 st gear. !T JHM = min ! 0T JLF = min 2.5 Optimization problem The two different optimization problems solved during the simulation cases are displayed in equations (HM) and (LF). 672 vopt dt 0 " T ṁf dt (HM) (LF) 0 where mf is the mass of fuel consumed, v is the vehicle speed. Equation (HM) for High Mileage and equation (LF) is the problem to solve for Low Fuel during the cycle. The problem consists of three states, vehicle speed, driven distance and selected gear. The states driven distance and vehicle speed are linked to each other. The vehicle speed is defined by the drive cycle and the allowed deviation from the speed reference, the driven distance is solved with the knowledge of the vehicle speed and the time step length. Since the total driven distance is not specified by the drive cycle, the state driven distance is unconstrained. By solving the problem with dynamic programing, a twodimensional grid at each time step is needed, consisting of all possible gears and all possible vehicle speeds from tn to tn+1 . 3. RESULTS Results for the european city cycle are displayed in Figures A.1, A.3 and A.5. Results for the full cycle are displayed in Figures A.2 , A.4 and A.6. The figures displays the optimal speed trajectory for each case, fuel flow to combustion engine, selected gear and deviation from cycle reference speed. The results in Figures A.1 and A.2 (when using specified gear changes) are used as references when calculating the savings in terms of fuel consumption. All the simulations are summarized and displayed in Table 3, where the savings in fuel consumption is calculated according to equation (18), where ∅ref is fuel consumption (l/100 km) for reference simulation and ∅case is fuel consumption for the case that is to be compared. ηsavings Fig. 4. Graphical representations of the dynamic programing algorithm. At each time step from tn to tn+1 , the cost for each opportunity of selecting speed, v, and gear γ, is calculated. ṁf dt case = 100 ∅ref − ∅case ∅ref (18) The vehicle speed trajectory is optimized within the limit of ±2 km/h and ±1 s from reference speed, to achieve as Low Fuel consumption as possible, and as High Mileage as possible, using both manual and automatic gearbox with free possibility to select gears. The Low Fuel consumption optimization corresponds to comparing the fuel consumed by the nominal cycle distance in the drive cycle, with the fuel actually consumed. Since the vehicle speed and nominal cycle speed are unconnected, the solution ends up with minimizing the total amount of fuel used during the drive cycle. The High Mileage optimization uses the driven distance and amount of fuel consumed, since the vehicle speed is allowed to deviate from the cycle speed reference, also the driven distance will vary. All solutions IFAC AAC 2016 June 19-23, 2016. Norrköping, Sweden Kristoffer Ekberg et al. / IFAC-PapersOnLine 49-11 (2016) 657–664 where a deviation in vehicle speed is allowed, have one thing in common, the optimal solution is to extend the time in fuel-cut (e.g. where the engine brakes) compared to reference simulations. Due to the speed variance from the reference speed, the total driven distances are different for the different simulations. 3.1 Low speed part of NEDC Speed deviations not allowed (optimizing for Low Fuel) If the deviations in vehicle speed had been forbidden, the gear changes had been the only degree of freedom if using automatic gearbox. In Figure A.1 the solutions for specified gear changes and optimized gear changes are displayed. The optimal solution when gear selections are unrestricted, is to select highest gear possible, to reduce the fuel flow and thereby reduce fuel consumed (the fuel consumption is reduced from 6.79 to 5.49 l/100 km). Optimizing for High Mileage When vehicle speed is allowed to deviate from reference speed, the problem has two degrees of freedom (when gear changes are unrestricted), gear changes and vehicle speed. In Figure A.3 the solutions for the High Mileage optimizations are displayed. The optimal solution when the gear selections are unrestricted, is to extend the driven distance, from 1018 m (in reference case) to 1042 m. When using specified gear changes, the optimal solution is to shorten the driven distance, from 1018 m (in reference case) to 971 m, the driven distance is only extended when using fuel cut during engine braking. Also the solver uses highest gear possible when optimized gear changes are used. Optimizing for Low Fuel When vehicle speed is allowed to deviate from reference speed, the problem has two degrees of freedom (when gear changes are unrestricted), gear changes and vehicle speed. In Figure A.5 the solutions for the Low Fuel optimizations are displayed. It is visible that the optimal solution is to keep the vehicle speed as low as possible, to keep the fuel flow low. The lowest allowed speed is kept along the cycle, except during engine braking, where the possibility to run on low or no fuel is present. 3.2 Full NEDC The low speed part of NEDC is previously displayed, to show the different optimal solutions found by the dynamic programming algorithm. The full NEDC cycle is also simulated, to investigate how the allowed deviation in vehicle speed may be used during a full drive cycle test. Speed deviation not allowed (optimizing for Low Fuel) In Figure A.2 it is visible that the optimal solution for the low speed parts of the drive cycle looks similar to the city cycle simulation (see Figure A.1). The solver uses the ability to change gears, by selecting the highest gear possible, to reduce the fuel flow and thereby the fuel consumed. Investigating the gear selections pattern in Figure A.2, at around 900s, the solver changes from 5th to 4th and down to 3rd gear for a short moment and then up to 5th gear again, during this time the vehicle does not require any fuel flow. The optimal solution at that moment is therefore to store kinetic energy in the engine (changing down to 3rd 673 661 gear), then use the kinetic energy stored in the engine when needed (changing from 3rd to 5th gear). Optimizing for High Mileage Studying the vehicle velocity profile in Figure A.4, it shows that the optimal solution for both specified and optimized gear selections is to increase the driven distance by driving faster than the reference profile at some occasions (such as accelerating earlier than the reference profile). It is also visible that the driven distance when using specified gear changes is shorter (10790 m) compared to optimal gear selections (11306 m). The optimal solution when gear selections are unrestricted, is to choose the highest gear possible while fulfilling the vehicle constraints, the use of higher gear reduces the fuel flow. Optimizing for Low Fuel As for the city cycle optimized for LF, the optimal vehicle speed is following the low speed boundary, except during roll outs when engine braking is possible (which cuts the fuel flow). Even though the highest gear is selected when possible (running the case when gear changes are unrestricted), the optimal vehicle speed stays below the reference speed (on the lower vehicle speed boundary condition). This is visible in Figure A.6. 3.3 Summary Results Table 3 summarizes all the results for the different cycles. Table 3. Results from all the different simulation cases. The case ”LF (vref )” follows the specified vehicle speed exactly. Case Description Savings % City Cycle Reference LF HM City Cycle LF (vref ) LF HM Full Cycle Reference LF HM Full Cycle LF (vref ) LF HM Manual 4.71% 8.40 % Automatic 19.15 % 24.60 % 29.60 % Manual 3.61 % 5.90 % Automatic 9.67 % 13.28 % 16.56 % Mass Fuel g Mean Fuel l/100 km Distance m 50.95 44.14 44.53 6.79 6.47 6.22 1018 925 971 41.22 35.52 36.72 5.49 5.12 4.78 1018 941 1042 495.57 450.32 456.63 6.10 5.88 5.74 11021 10394 10790 448.01 408.25 424.28 5.51 5.29 5.09 11021 10469 11306 3.4 Parameter sensitivity analysis A parameter sensitivity analysis has been performed, investigating the change in fuel consumption, if changing parameters vehicle mass m, tire rolling resistance cr and vehicle frontal area A. The analysis is performed for the reference case in Figure A.2, with the full NEDC cycle, where both the vehicle speed and gear changes are specified. The result of the parameter sensitivity analysis is displayed in Table 4. The result is displayed as the difference in fuel consumption, calculated according to Equation Kristoffer Ekberg et al. / IFAC-PapersOnLine 49-11 (2016) 657–664 Table 4. Parameter sensitivity analysis. The top percentage shows the size of the change in the parameters A, m and cr . The result shows the relative difference in fuel consumption, compared to the reference case in Figure A.2, using specified gear selections (increased fuel consumption in negative numbers). Parameter 10% 5% A m cr -1.743 -4.504 -2.410 -0.872 -2.250 -1.205 −5% 0.871 2.249 1.204 −10% 1.742 4.495 2.409 4. CONCLUSION Results show that by optimizing the driving pattern within the allowed deviation from the drive cycle reference velocity of ±2 km/h and ±1 second, the fuel consumption of the driven vehicle can be clearly reduced. A dynamic programming, backward algorithm has been used, to find the optimal solutions for High Mileage and Low Fuel over a specified drive cycle, with allowed deviations in vehicle speed. The use of free possibility to change gears was also examined. The results in table 3, show that the fuel consumption can be clearly reduced by taking advantage of the allowed deviation in vehicle speed. Optimization results shows that it is optimal to use engine braking when possible, to enable fuel cut, independent of if the optimization goal is High Mileage or Low Fuel. The possibility to deviate from reference speed leads to different driven distances for the different cases. To increase High Mileage, the driven distance is increased, while fuel flow is kept low. When optimizing for Low Fuel, the vehicle speed is made as low as possible, to lower the fuel flow in the engine. Comparing the savings of fuel consumption in Table 3 with the changes of fuel consumption in Table 4, it shows that the savings when taking advantage of the allowed speed deviation is generally larger than a 10% decrease of any of the parameters frontal area A, rolling resistance cr or vehicle mass m used during simulations. Eriksson, L. and Nielsen, L. (2014). Modeling and Control of Engines and Drivelines. John Wiley and Sons Ltd, United Kingdom. Guzzella, L. and Sciarretta, A. (2007). Vehicle Propulsion Systems. Springer, New York. Luu, H.T., Nouvelière, L., and Mammar, S. (2010). Dynamic programming for fuel consumption optimization on light vehicle. In Prep. IFAC symp. Advances in Automotive Control. AAC2010. Mensing, F., Trigui, R., and Bideaux, E. (2011). Vehicle trajectory optimization for application in eco-driving. In Vehicle Power and Propulsion Conference (VPPC), 2011 IEEE, 1–6. doi:10.1109/VPPC.2011.6042993. Wang, R. and Lukic, S. (2012). Dynamic programming technique in hybrid electric vehicle optimization. In Electric Vehicle Conference (IEVC), 2012 IEEE International, 1–8. doi:10.1109/IEVC.2012.6183284. Wollaeger, J., Kumar, S., Onori, S., Filev, D., Ozguner, U., Rizzoni, G., and Di Cairano, S. (2012). Cloudcomputing based velocity profile generation for minimum fuel consumption: A dynamic programming based solution. In American Control Conference (ACC), 2012, 2108–2113. doi:10.1109/ACC.2012.6314931. Appendix A. OPTIMIZATION RESULTS 20 1 20 40 60 100 120 140 160 180 Specified Gear Optimal Gear 0 0 20 40 60 80 100 120 140 160 180 140 160 180 160 180 Gear selections 6 4 2 0 0 20 40 60 80 100 120 Difference from reference velocity 1 Velocity [km/h] 80 Fuel Consumption Spec/Opt 6.79/5.49 l/100 km 0.5 Gear [-] 674 0 ×10-3 This work was supported by the Vinnova Industry Excellence Center: LINK-SIC Linköping Center for Sensor Informatics and Control. Bertsekas, D.P. (2000). Dynamic Programming and Optimal Control, volume 2. Athena Scientific, Bellmonth, Massachusetts. Bhandari, V.B. (2007). Design of Machine Elements. Tata McGraw-Hill. E/ECE/324/Rev.1/Add.82/Rev.5E/ECE/TRANS/505/Rev.1/Add.82/Rev.5 (2015). Regulation no. 83 uniform provisions concerning the approval of vehicles with regard to the emission of pollutants according to engine fuel requirements. Boundary Boundary Specified Gear Optimal Gear 40 0 ACKNOWLEDGEMENTS REFERENCES Driven distance Spec/Opt 1018/1018 m 60 Velocity [km/h] (18). If comparing best saving in fuel consumption in Table 4 with for example Full Cycle - Manual - LF in Table 3, one of the smaller savings due to the allowed speed deviation corresponds the biggest change of mass in table 4. Fuel Flow [kg/s] IFAC AAC 2016 662 June 19-23, 2016. Norrköping, Sweden 0.5 0 -0.5 -1 0 20 40 60 80 100 120 140 Time [s] Fig. A.1. Exact speed reference following with cycle specified gear changes and automatic gear changes. Optimizing for LF with specified vehicle speed and both specified and optimal gear changes. It shows in figure that fuel cut is used with both specified and optimal gear changes. IFAC AAC 2016 June 19-23, 2016. Norrköping, Sweden Kristoffer Ekberg et al. / IFAC-PapersOnLine 49-11 (2016) 657–664 Velocity [km/h] 100 50 0 0 200 Fuel Flow [kg/s] 400 600 800 1 0 200 400 800 200 400 600 800 0 0 200 Velocity [km/h] 0 400 600 800 600 0 200 400 600 -5 200 400 600 Velocity [km/h] 80 100 120 140 160 20 40 60 100 120 140 160 0 Gear [-] 20 40 60 80 100 120 40 140 160 Velocity [km/h] Velocity [km/h] -5 60 80 100 120 140 120 140 160 180 0 20 40 60 80 100 120 140 160 180 140 160 180 160 180 Gear selections 2 0 20 40 60 80 100 120 Difference from reference velocity 0 40 100 4 0 180 5 20 80 Specified Gear Optimal Gear Difference from reference velocity 0 60 Fuel Consumption Spec/Opt 6.47/5.12 l/100 km 6 2 0 20 0.5 180 4 0 1 Gear selections 6 Gear [-] 80 0 ×10-3 Specified Gear Optimal Gear 0 20 0 0.5 0 Boundary Boundary Specified Gear Optimal Gear 40 180 Fuel Flow [kg/s] Velocity [km/h] Fuel Flow [kg/s] 60 Driven distance Spec/Opt 925/941 m 60 Fuel Consumption Spec/Opt 6.22/4.78 l/100 km ×10-3 1 40 1000 Fig. A.4. Optimizing for HM, full cycle, when velocity deviations are allowed. Both specified and optimal gear changes. The optimal gear changes case has longer driven distance than specified gear case. Driven distance Spec/Opt 971/1042 m 20 800 Time [s] Boundary Boundary Specified Gear Optimal Gear 0 1000 0 0 20 0 800 5 1000 Fig. A.2. Exact speed reference following with cycle specified gear changes and automatic gear changes. Optimizing for LF with specified vehicle speed and both specified and optimal gear changes. It shows in figure that fuel cut is used with both specified and optimal gear changes. 40 1000 2 Time [s] 60 800 Difference from reference velocity -0.5 200 400 4 0 1000 0.5 0 1000 Gear selections Difference from reference velocity 1 800 1 6 2 600 0.5 1000 4 0 400 Fuel Consumption Spec/Opt 5.74/5.09 l/100 km Specified Gear Optimal Gear 2 Gear [-] Gear [-] 600 200 1.5 Gear selections 6 -1 0 ×10-3 0.5 0 50 0 1.5 0 100 1000 Specified Gear Optimal Gear 2 Boundary Boundary Specified Gear Optimal Gear Fuel Consumption Spec/Opt 6.1/5.51 l/100 km ×10-3 Velocity [km/h] Driven distance Spec/Opt 10790/11306 m Boundary Boundary Specified Gear Optimal Gear Fuel Flow [kg/s] Velocity [km/h] Driven distance Spec/Opt 11021/11021 m 663 160 180 5 0 -5 0 Time [s] 20 40 60 80 100 120 140 Time [s] Fig. A.3. Optimizing for HM, city cycle, when velocity deviations are allowed. Both specified and optimal gear changes. The optimal gear changes case has longer driven distance than specified gear case. 675 Fig. A.5. Optimizing for LF, city cycle, when velocity deviations are allowed. Both specified and optimal gear changes. The optimal gear changes case has longer driven distance than specified gear case. IFAC AAC 2016 664 June 19-23, 2016. Norrköping, Sweden Kristoffer Ekberg et al. / IFAC-PapersOnLine 49-11 (2016) 657–664 Velocity [km/h] Driven distance Spec/Opt 10394/10469 m Boundary Boundary Specified Gear Optimal Gear 100 50 0 0 Fuel Flow [kg/s] ×10-3 200 400 600 800 1000 Fuel Consumption Spec/Opt 5.88/5.29 l/100 km Specified Gear Optimal Gear 2 1.5 1 0.5 0 0 200 400 6 Gear [-] 600 800 1000 800 1000 Gear selections 4 2 0 0 200 400 600 Velocity [km/h] Difference from reference velocity 5 0 -5 0 200 400 600 800 1000 Time [s] Fig. A.6. Optimizing for LF, full cycle, when velocity deviations are allowed. Both specified and optimal gear changes. The optimal gear changes case has longer driven distance than specified gear case. 676