Keynote Address 3 - Lars Nielsen PDF

Look-Ahead Fuel-optimal Control of Heavy Trucks
Professor Lars Nielsen

lars@isy.liu.se
Vehicular Systems
Department of Electrical Engineering
E-COSM, December 1, 2009
Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 1 / 48

Objective
Scenario
Long-haulage heavy truck on open road
An on-board road topography map and GPS (i.e. Look Ahead)
Controls
Engine and brake torque (continuous)
Gear ratio selection (discrete)

Objective
Scenario
Long-haulage heavy truck on open road
An on-board road topography map and GPS (i.e. Look Ahead)
Controls
Engine and brake torque (continuous)
Gear ratio selection (discrete)
Minimum-fuel strategy for a drive mission with a given maximum trip time
minimize M
subject to T ≤ T0
Motivation : Cost
Fuel 29.3%
Salaries 30.9%
Life cycle cost

Taxes
& Insurances 9.0%
Oil 1.5% Interest

& Depreciation 14.8%
Maintenance
& Service 7.3%
Tires 7.2%
An average class 8 (i.e. heavy truck for long haulage)

travels 150,000 km per year
consumes 32.5 liter per 100 km in Europe according to (Schittler, 2003)

Motivation : Environment
Class 8 trucks
typically travels at operating points with high efficiency
on the other hand, they
consume 68 % of all commercial truck fuel used
70 % of this amount is spent traveling on open road with a trip length
of more than 100 miles (161 km)
In the U.S. according to (Bradley, 2000)

Motivation : Environment
Class 8 trucks
typically travels at operating points with high efficiency
on the other hand, they
consume 68 % of all commercial truck fuel used
70 % of this amount is spent traveling on open road with a trip length
of more than 100 miles (161 km)
In the U.S. according to (Bradley, 2000)
Motivations
Fuel is a large share of the life cycle cost.
According to industry any technology that improves fuel efficiency of
long-haul class 8 trucks with 0.5% is worth pursuing.

Outline
Introduction and Motivation
Outline and Initial Observations
An Efficient Algorithm
Experimental Evaluation
Extended analysis
Non-linear fuel maps
Gear-shifts
Hybrid vehicles
Concluding Remarks

History
Just some samples to show background
Vehicles - First wave
A.B. Schwarzkopf and R.B. Leipnik, Control of Highway Vehicles for
Minimum Fuel Consumption over Varying Terrain, 1977.
J.N. Hooker, A.B. Rose and G.F. Roberts, Optimal Control of
Automobiles for Fuel Economy, 1983.
J.N. Hooker, Optimal Driving for Single-vehicle Fuel Economy, 1988.
V.V Monastyrsky and I.M. Golownykh, Rapid Computations of
Optimal Control for Vehicles, 1993.
Trains - Still important (energy, residual cost)

R. Franke, M. Meyer and P. Terwiesch, Optimal Control of the
Driving of Trains, 2002.
R. Liu and I.M. Golovitcher, Energy-efficient Operation of Rail
Vehicles, 2003.
History
Just some samples to show background
Vehicles - Second wave
S. Terwen, M. Back and V. Krebs, Predictive Powertrain Control for
Heavy Duty Trucks, 2004.
F. Kirschbaum, M. Back and M. Hart, Determination of the
Fuel-Optimal Trajectory for a Vehicle along a Known Route, 2002.
F. Lattemann, K. Neiss, S. Terwen and T. Connolly, The Predictive
Cruise Control - A System to Reduce Fuel Consumption of Heavy
Duty Trucks, 2004.
A. Sciarretta and L. Guzzella, Fuel-Optimal Control of Rendezvous
Maneuvers for Passenger Cars, 2005.
D. J. Chang and E. K. Morlok, Vehicle Speed Profiles to Minimize
Work and Fuel Consumption, 2005.
L. Johannesson and B. Egardt, Approximate Dynamic Programming
Applied to Parallel Hybrid Powertrains, 2008.
Our own development
Road map for this talk

First developments of on-line algorithms
Successful demonstrations in vehicle
Observed behavior in experiments and simulations
Understand behavior
Improved developments of on-line algorithms
Analysis being foundation for tuning
Analysis being foundation for robustness
Extended analysis
Energy formulation
Residual cost
Gear-shifts
Hybrid vehicles

Objective
Minimum-fuel strategy for a drive mission with a given maximum trip time
minimize M
subject to T ≤ T0
A computationally efficient approach is to consider a truncated

horizon, a receding horizon approach.

Receding Horizon - Some notation
Drive mission s ∈ [0, S], S > H

Look-ahead horizon s ∈ [0, H]
The criterion is defined for s ∈ [0, S]: J = min M(S)

but we consider it for s ∈ [0, H]: ≈ min M(H) + R̃(x(H))
where R̃ is an estimate of the residual cost and a function of the terminal
state x(H).
Objective defined
The problem is to find the fuel-optimal control law for a finite horizon

A Generic Analysis - Constant speed optimal
Fd (s, v) m Fp
dv
mv = Fp − Fd (s, v )
s ds
Minimize the Propulsive Energy

Z S Z S
dv
min W = Fp ds = mv + Fd (s, v ) ds
v (s) 0 0 ds
Z S
ds
subject to ≤ T0
0 v
Minimum for Constant Speed

By calculus of variations, the solution is constant speed if
∂Fd
≥0 and Fd (s, v ) = f1 (s) + f2 (v ).
∂v

Key Features
However
The engine power compared to
vehicle mass makes
moderate slopes significant
⇒ Constant speed not possible
to keep.
gear shifts necessary

Key Features
However
The engine power compared to
vehicle mass makes
moderate slopes significant
⇒ Constant speed not possible
to keep.
gear shifts necessary
Some challenges trying to reduce fuel consumption

Already highly fuel-efficient
Real-time requirements

Outline
Extended analysis
Gear-shifts
Hybrid vehicles
Concluding Remarks

Where do we stand today?
Current main challenge is to find the solution efficiently
We propose an algorithm based on Dynamic Programming (DP)

Motivation for DP
Exponential increase of complexity in DP due to continuous variables
No issue here since the dimension is low
Favorable linear growth with increasing horizon
A rather long horizon is needed
Alternative methods typically give a complex combinatorial problem.
Allows general modeling

Motivation for DP
Exponential increase of complexity in DP due to continuous variables
No issue here since the dimension is low
Favorable linear growth with increasing horizon
A rather long horizon is needed
Alternative methods typically give a complex combinatorial problem.
Allows general modeling
Approaches to improve computational tractability

Shorter horizon
Better estimate of the residual cost at the end of the horizon
Fewer grid points
Lowering the dimension and reducing the search space
Coarse grid and interpolation together with simple integration
An analysis of errors due to discretization and interpolation shows that
an energy formulation of the dynamics is beneficial
Residual Cost
fueling u (g/cycle cylinder)
mass M = ncyl u (g/cycle)
Fueling u versus torque T is
approximately an affine
function with gradient k
Scaling gives the relation
u = f (T, . . .) ≈ kT + c between fuel mass M and
work W . The gradient is
engine torque T (Nm)

ncyl
γ=k (g/J)
work W = 2πnr ηT (J/cycle) 2πnr η
Equivalence factor γ between fuel and work

∆M ≈ γ∆W

Residual Cost using same Equivalence factor
A change in kinetic energy ∆e at the end of the horizon is

approximately proportional to a fuel mass
∆M ≈ γ∆e
This reflects that kinetic energy at the end of the horizon can be used
to save fuel in the future
With this assumption, the residual cost is
R̃ = C − γe
where C can be chosen to zero (optimization is relative).

We have also shown that this is accurate in the unconstrained case
For the general case, numerical results show that this is reasonable

Lowering The Dimension : Reformulation
J = min M (P1)
T =T0 J = min M + βT (P2)
 0 0
t v
v  = f (s, v , g , u) = f˜(s, v , g , u)
g
g
For a given β, the solution for (P2) gives a trip time T = T (β). If we find
β such that T (β) = T0 , then we have the solution for (P1) as well.
Time is no longer necessary as a state, instead we have to find β.

Lowering The Dimension : Finding β
The function T (β) is monotonically decreasing
β could be found by, e.g., simple shooting methods
We derive an approximate value of β
A vehicle travels at speed v̂ on level road for ∆s. Then, typically

Fd (s, v ) = Fd (v ) and we have J(v̂ ) = M + βT where
∆s
M = γ∆W = γFd (v̂ )∆s and T =
v̂
In a stationary point, J 0 (v̂ ) = 0, it holds that β = γv̂ 2 Fd0 (v̂ ), where
typically Fd0 (v̂ ) = Fair
0 (v̂ )
Continued use of equivalence factor (time vs fuel)

β = 2γPair (v̂ ) (g/s)

Algorithm : Summary
Drive mission s ∈ [0, S], S > H
Look-ahead horizon s ∈ [0, H]
We consider (P2) for s ∈ [0, H], i.e.,

n o
min M + βT + R̃
where R̃ = −γe(H) and β = 2γPair (v̂ ).

Using DP in energy formulation, i.e. v 2 as state rather than v .
Energy formulation of the dynamics makes it possible to use coarse
grids together with linear interpolation and Euler forward integration
E. Hellström, J. Åslund, and L. Nielsen.
Design of an efficient algorithm for fuel-optimal look-ahead control.
Solicited for Control Engineering Practice.
Outline
Extended analysis
Gear-shifts
Hybrid vehicles
Concluding Remarks

Evaluation Setup
Demonstrator vehicle in collaboration with scania
GPS and road database on board
Louice
About 40,000 kg
5 cylinder, 9 liter engine
Max. 1550 Nm, 310 Hp
12-speed transmission
84 km/h ± 5 km/h
1500 m horizon
Search for Look-ahead Control
E. Hellström, M. Ivarsson, J. Åslund, and L. Nielsen.
Look-ahead control for heavy trucks to minimize trip time and fuel consumption.
Control Engineering Practice, 17(2):245–254, 2009.
Evaluation Setup : Interface
Feed the cruise GPS

Set speed
controller with Position
set speeds
CAN bus
Road
Road slope
Model the DP algorithm
database
automatic gear
shift system Current velocity and gear

Trial Route
120 km highway segment

Trial Route : Topography
Moderate slopes
Light to moderate traffic
120
100
Altitude [m]
80
60
40
20
0 20 40 60 80 100 120
Södertälje Position [km] Norrköping
2
Slope [%]
−2
−4
0 20 40 60 80 100 120
Södertälje Position [km] Norrköping

Average Results
Average over four runs back and forth
fuel consumption [L/100km] trip time [seconds] number of gear shifts [−]
−3.54 % −0.03 % −42.0 %
36.33 4905.12 4905.38 20
35.03
12
LC CC LC CC LC CC
Recall that 0.5% is worth pursuing.

The Hållet Segment

The Hållet Segment : Characteristics
Cruise controller (CC): ∆fuel = −8.17 % Look−ahead controller (LC): Cruise controller (CC): ∆fuel = −6.04 % Look−ahead controller (LC):
29.72 dm3/100km, 162.7 s ∆time = −1.78 % 27.29 dm3/100km, 159.8 s 47.79 dm3/100km, 164.3 s ∆time = −1.82 % 44.90 dm3/100km, 161.3 s
45
Altitude [m]
Altitude [m]
50
40
40 35
30
30 25
0
58.77° N, 17.01° E 500 1000 1500 2000 2500 3000 3500N, 16.97° E 58.75° N, 16.97°
58.75° 0 E 500 1000 1500 2000 2500 3000 58.77°
3500N, 17.01° E
Velocity [km/h]
Velocity [km/h]
90 90
80 80
LC LC
70 70
CC CC
0 500 1000 1500 2000 2500 3000 3500 0 500 1000 1500 2000 2500 3000 3500
LC acc 1 1
LC brake
LC acc
Level [−]
Level [−]
CC acc
0.5 0.5 LC brake
CC brake
CC acc
CC brake
0 0
0 500 1000 1500 2000 2500 3000 3500 0 500 1000 1500 2000 2500 3000 3500
Gear number [−]
Gear number [−]
1
Fuel use [dm3]
Fuel use [dm3]

LC gear LC gear 1.5
CC gear CC gear
12 12 1
0.5
LC fuel use LC fuel use 0.5
11 CC fuel use 11 CC fuel use
0 0
0 500 1000 1500 2000 2500 3000 3500 0 500 1000 1500 2000 2500 3000 3500
Position [m] Position [m]

The Järna Segment

The Järna Segment : Characteristics
Cruise controller (CC): ∆fuel = −0.62 % Look−ahead controller (LC): Cruise controller (CC): ∆fuel = −7.21 % Look−ahead controller (LC):
50.67 dm3/100km, 116.2 s ∆time = −1.03 % 50.36 dm3/100km, 115.0 s 28.04 dm3/100km, 110.6 s ∆time = −0.54 % 26.02 dm3/100km, 110.0 s
55
45
Altitude [m]
Altitude [m]
50
40
45
40 35
59.14° N,017.65° E 500 1000 1500 2000 2500

59.12° N, 17.63° E 59.12° N,017.63° E 500 1000 1500 2000 2500
59.14° N, 17.65° E
92
Velocity [km/h]
Velocity [km/h]
LC
85 90 CC
80 88
86
75 LC 84
CC 82
70
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
LC acc
1 1
Fueling level [−]
LC brake
CC acc
Level [−]
0.5 CC brake 0.5
LC
CC
0 0
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
Gear number [−]
Gear number [−]

Fuel use [dm3]
Fuel use [dm3]

LC gear LC gear 0.6
CC gear 1 CC gear
12 12 0.4
0.5
LC fuel use LC fuel use 0.2
11 CC fuel use CC fuel use
0 0
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
Position [m] Position [m]

Outline
Extended analysis
Gear-shifts
Hybrid vehicles
Concluding Remarks

Extended analysis
Non-linear fuel map
Fuel map showing specific fuel consumption, sfc [g/kWh], of a combustion
engine. The disadvantageous speed 1500 rpm requires a higher sfc than
1300 or 1700 rpm.
sfc [g/kWh]
90
84
78
72
66
normalized torque [%]
60
54
48
42
36
30
24
18
1200 1300 1400 1500 1600 1700 1800
speed [rpm]
What are the consequences of a non-linear fuel map?

Extended analysis - Non-linear fuel map
Optimal solution for small gradient
20
15
Altitude [m]
10
0
0 500 1000 1500 2000 2500 3000 3500 4000
Velocity
v0
0 500 1000 1500 2000 2500 3000 3500 4000

Position [m]
Corresponding to the dashed line on previous slide for a small uphill.

For a mildly non-linear fuel map the solution would be constant speed.
Note the switching-like behavior.

Recall problem formulations
J = min M (P1) J = min M + βT (P2)

T =T0
Quasi-static analysis first

Neglect transient time and cost, i.e. it is assumed to be possible to jump
from one speed to another.

Recall problem formulations
J = min M (P1) J = min M + βT (P2)

T =T0
Quasi-static analysis first

Neglect transient time and cost, i.e. it is assumed to be possible to jump
from one speed to another.
Quasi-static analysis for (P1)

The optimal solution is to jump between two specific characteristic
velocities in a proportion so that the time constraint is met.

Quasi-static analysis for (P2)
The criterion for (P2) is
Z s
1
J= C0 Te sfc + β ds
0 v
With
1
F = C0 Te (N)sfc(N) + β
v (N)
where both Te (N) and sfc(N) depends on engine speed, the requirement
for optimality is
dF
=0
dN
This gives a relation between β and N.

Quasi-static analysis - N̂-β-plot
2000
1800
1600 NB
[rpm]
1400
N
A
1200
1000
800
0 2 4 6 8 10 12 14
β
Stationary solutions, N̂, where the global minimum, N̂ ∗ , is marked with fat
lines, and the boundary velocities, NA and NB .

Quasi-static analysis
The characteristic velocities for (P2) in the plot
NA NB
are exactly the same as the two switching velocities for (P1)!
Calculations give
d 2 F d N̂ 1 r 2π
= 2
dN 2 dβ v i60
d 2F d N̂
Hence dN 2
and dβ must always have the same sign. This means that if
d 2F
the slope of N̂(β) in the N̂-β-plot is positive, ddβ
N̂
> 0, then dN 2
> 0 and
there is a local minimum of the objective function.
The requirement for a significant non-linearity is
dF d 2F d 3F
= 0, = 0, =0
dN dN 2 dN 3
Full dynamic simulation

0
altitude [m]
−20
−40
0 1000 2000 3000 4000 5000 6000 7000 8000
50
25
0
0 1000 2000 3000 4000 5000 6000 7000 8000
high
v [km/h]
v0
low
0 1000 2000 3000 4000 5000 6000 7000 8000
distance [m]
Solid - optimal control. Dashed - constant speed control. ∆mf = −1% on

a 8000 m road stretch.

Same full dynamic solution plotted in the fuel map
sfc [g/kWh]
fuel map
90 constant speed
optimal control
84
78
72
66
60
54
48
42 high
36
30
24 low
18
1200 1300 1400 1500 1600 1700 1800
speed [rpm]
Plotting the optimal solution from the previous slide in the engine map
(inspired by Pfiffner and Guzzella).
The optimal trajectory can move around (not for long) to find favorable
spots.
Conclusions
Avoids unfavorable areas in switching-like bahavior
Concepts of characteristic velocities and significant non-linearity
defined.
Tuning β for velocity control not necessarily possible
Thus important to check in the design process and to monitor.
M. Ivarsson, J. Åslund, and L. Nielsen.

Look Ahead Control – Consequences of a Non-Linear Fuel Map on Truck Fuel
Consumption.
Journal of Automobile Engineering, 223:1223–1238, oct 2009.

Extended analysis
Gear shifting
Modeling of gear shifts in algorithm development?

Extended analysis
Gear shifting
Modeling of gear shifts in algorithm development?
Question: When to shift gears in an uphill?
Gear down before the uphill?

Wait as long as possible?
Somewhere in between?

Extended analysis
Model 1: Instant gear-shift with no significant loss

Reasonable idea since a gear shift is quick compared to the drive
mission.
Result: Wait as long as possible.

Extended analysis - Gear shifting
Detailed physical modeling of Automated Manual Transmission
10
gear [−]
9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Cramp,sl
u [mg/stroke]
−Cramp u C
f,shift ramp
uf,zt uf,zt
f
0 t
adjust
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
ωe,post
ωe,end
ωe [1/s]
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

time [s]
M. Ivarsson, J. Åslund, and L. Nielsen.

Impacts of AMT Gear-Shifting on Fuel Optimal Look Ahead Control.
Provisionally accepted for SAE 2010.

Conclusions
Instantaneous gear shift model too simple
It is shown that the reduced propulsion of a gear-shifting process, and
the retardation of the truck while shifting gears, must be considered.
Compared to AMT without Look Ahead
Large gain when gear shifts can be avoided
If gear shifting is inevitable, then no large difference
The occasion of shifting gear must be chosen to ensure an adequate
engine speed in order to get a sufficient engine power after the gear
shift, since the truck is decelerated during gear shift.
Some characteristics balances the optimal solutions
Gearing down earlier gives higher engine speed, and thus power.
Gearing up early gives better engine efficiency.
Keeping as constant velocity as possible is beneficial.

Relating to “An Efficient Algorithm”
Must cope with
Velocity
Jgs (sn−1 , vm , g)
vm
J(sl−1 , vk , g ′ ) J(sl , vk , g ′ )
vk ′
g 6= g
vm +∆v J ′ (g ′ )
vk−1
J(sl−1 , vk−1 , g ′ ) J(sl , vk−1 , g ′ )
Position
sn−1 sn sl−1 sn−1 +∆s sl
Energy formulation, i.e. v 2 as state rather than v .

A functional framework (∆s, ∆v , ∆t, ∆m).
Interplay criterion, discretization, interpolation.
Extended analysis
Hybrid vehicles - an appetizer
The algorithm framework is used to combine hybridization and
velocity management.
The key idea is to introduce more equivalence factors in a way that
enables efficient computations, but also so that the equivalence
factors have a physical interpretation (β, γ, η).
From the hybridization perspective it can be seen as an extension of
the equivalence factor idea in the well known strategy ECMS.
From the perspective of look-ahead control, the extension is that
energy can be stored not only in kinetic energy, but also electrically.
From evaluation perspective optimal driving instead of drive cycle.
E. Hellström, J. Åslund, and L. Nielsen.

Management of Kinetic and Electric Energy in Heavy Trucks.
Provisionally accepted for SAE 2010.

Outline
Extended analysis
Gear-shifts
Hybrid vehicles
Concluding Remarks

Concluding Remarks
Better understanding of algorithm
Advantageuos to use energy formulation, v 2 , rather than speed
Residual cost gets natural formulation
Better interpolation
Stable discretization relative criterion
Better understanding of relevant characteristics
Developed beyond pure numerical optimization
Bang-singular-bang with singular being constant speed
Understanding of characteristics due to non-linear fuel map and
gear-shifting
Significant reduction of the fuel consumption is demonstrated
The behavior is perceived as natural and comfortable
Still a rich research field, and do not forget developments in vehicle
situation awareness, vehicle-to-vehicle communication, road-side
information and so on.

Look-Ahead Fuel-optimal Control of Heavy Trucks
Professor Lars Nielsen

lars@isy.liu.se
Vehicular Systems
Department of Electrical Engineering
E-COSM, December 1, 2009

Keynote Address 3 - Lars Nielsen PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Keynote Address 3 - Lars Nielsen PDF

Uploaded by

Copyright:

Available Formats

Look-Ahead Fuel-optimal Control of Heavy Trucks

Professor Lars Nielsen

E-COSM, December 1, 2009

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 1 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 2 / 48

Life cycle cost

Oil 1.5% Interest

An average class 8 (i.e. heavy truck for long haulage)

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 3 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 4 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 4 / 48

Introduction and Motivation

Outline and Initial Observations

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 5 / 48

Trains - Still important (energy, residual cost)

Road map for this talk

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 8 / 48

A computationally efficient approach is to consider a truncated

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 9 / 48

Drive mission s ∈ [0, S], S > H

The criterion is defined for s ∈ [0, S]: J = min M(S)

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 10 / 48

Minimize the Propulsive Energy

Minimum for Constant Speed

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 11 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 12 / 48

Some challenges trying to reduce fuel consumption

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 12 / 48

Introduction and Motivation

Outline and Initial Observations

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 13 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 14 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 14 / 48

Approaches to improve computational tractability

engine torque T (Nm)

Equivalence factor γ between fuel and work

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 15 / 48

A change in kinetic energy ∆e at the end of the horizon is

where C can be chosen to zero (optimization is relative).

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 16 / 48

Time is no longer necessary as a state, instead we have to find β.

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 17 / 48

A vehicle travels at speed v̂ on level road for ∆s. Then, typically

Continued use of equivalence factor (time vs fuel)

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 18 / 48

We consider (P2) for s ∈ [0, H], i.e.,

where R̃ = −γe(H) and β = 2γPair (v̂ ).

Introduction and Motivation

Outline and Initial Observations

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 20 / 48

Feed the cruise GPS

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 22 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 23 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 24 / 48

Average over four runs back and forth

Recall that 0.5% is worth pursuing.

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 25 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 26 / 48

Gear number [−]

Fuel use [dm3]

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 27 / 48

Lars Nielsen (Linköping University) Look-Ahead Control E-COSM, December 1, 2009 28 / 48

59.14° N,017.65° E 500 1000 1500 2000 2500