optimizing hybrid vehicles via route prediction jon froehlich & john krumm hci intern talk july...
TRANSCRIPT
Optimizing Hybrid Vehicles via Route
Prediction
jon froehlich & john krumm
HCI Intern TalkJuly 26th, 2007
1/1/1
996
5/1/1
996
9/1/1
996
1/1/1
997
5/1/1
997
9/1/1
997
1/1/1
998
5/1/1
998
9/1/1
998
1/1/1
999
5/1/1
999
9/1/1
999
1/1/2
000
5/1/2
000
9/1/2
000
1/1/2
001
5/1/2
001
9/1/2
001
1/1/2
002
5/1/2
002
9/1/2
002
1/1/2
003
5/1/2
003
9/1/2
003
1/1/2
004
5/1/2
004
9/1/2
004
1/1/2
005
5/1/2
005
9/1/2
005
1/1/2
006
5/1/2
006
9/1/2
006
1/1/2
007
5/1/2
0070
0.5
1
1.5
2
2.5
3
3.5
4
Date
USD
/Gal
lon
(inlc
ludi
ng ta
x)
$3.40
U.S. Department of Energy, 2007http://www.eia.doe.gov/emeu/international/gas1.html
u.s. gas prices
1/1/1
996
5/1/1
996
9/1/1
996
1/1/1
997
5/1/1
997
9/1/1
997
1/1/1
998
5/1/1
998
9/1/1
998
1/1/1
999
5/1/1
999
9/1/1
999
1/1/2
000
5/1/2
000
9/1/2
000
1/1/2
001
5/1/2
001
9/1/2
001
1/1/2
002
5/1/2
002
9/1/2
002
1/1/2
003
5/1/2
003
9/1/2
003
1/1/2
004
5/1/2
004
9/1/2
004
1/1/2
005
5/1/2
005
9/1/2
005
1/1/2
006
5/1/2
006
9/1/2
006
1/1/2
007
5/1/2
0070
1
2
3
4
5
6
7
8
9
Belgium FranceGermanyItalyNetherlandsUKUS
Date
USD
/Gal
lon
(inlc
ludi
ng ta
x)
global gas prices
U.S. Department of Energy, 2007http://www.eia.doe.gov/emeu/international/gas1.html
Italy
South Korea
United Kingdom
Canada
Germany
Japan
India
Russia
European Union
China
United States
0.0% 5.0% 10.0% 15.0% 20.0% 25.0%
1.60%
1.70%
2.00%
2.10%
2.90%
4.50%
4.60%
5.40%
13.70%
15.20%
21.20%
CO2 emissions (2003)
2003 United Nations Statistics Divisionhttp://www.un.org/
1960 1965 1970 1975 1980 1985 1990 1995 2000 20050
5
10
15
20
25
30
35
40
45
Transportation
Industrial
Residential and commercial
Electric utilities
BTU
s (Q
uadr
illio
n)
u.s. oil demand by sector
U.S. Department of Transportation, 2005http://www.bts.gov/publications/national_transportation_statistics/html/table_04_03.html
International Energy Agencyhttp://www.iea.org/
why hybrids?
U.S. Department of Energy, 2001http://www.fueleconomy.gov/feg/tech/TechSnapPrius1_5_01b.pdf
how hybrids work
series
CombustionEngine
parallel
CombustionEngine
at low speeds
http://www.toyota.com/vehicles/2007/prius
highway cruising
http://www.toyota.com/vehicles/2007/prius
heavy acceleration
http://www.toyota.com/vehicles/2007/prius
engine idling
http://www.toyota.com/vehicles/2007/prius
regenerative braking
http://www.toyota.com/vehicles/2007/prius
what if we could predict a driver’s route?
road grade road curvature traffic conditions
HEV Charge/Discharge Control System Based onNavigation Information
Convergence Transportation Electronics Association 2004Nissan Motor Company
road grade traffic conditions
Predestination:Inferring Destinations from Partial Trajectories Ubiquitous Computing 2006
John Krumm and Eric Horvitz
Trip starts, uniform destination probability
4 squares south, half of region eliminated
More squares in trip, ¾ of region eliminated
msmls dataset
As of today 251 subjects 2,131,440 data points
UbiComp 2006 189 subjects 1,351,669 points 73,903 miles 9414 trips
predestinationfor routes?
gps massagingroute segmentationroute comparisonroute periodicity (recurrence)route prediction
gps massaging
Breaking the sound barrier (and all other sorts of physical feats) on the way to work
GPS can be noisy
route recurrence
And they often have a temporal pattern: 8 of the 9 trips along this route occurred between 8:00 & 9:30AM
Yes, people do drive the exact same routes over and over again
route profiles
37:42.040:47.044:34.047:35.050:36.018:36.031:08.034:17.038:20.00
20
40
60
80
100
120
140
160
180
Elevation Gain
Elev
ation
(fee
t)
evaluation
stage 1Mine the MSMLS datasetCheck our route prediction algorithmCheck our deceleration prediction algorithms
stage 2Create (or purchase) hybrid simulatorDynamically shift power train policiesCompare with baselineEvaluate based on better electrical motor utilization, state of charge, and fuel economy
contributions
Predicting speed, acceleration (and deceleration), and elevation along route
Automatic route prediction
Studying effect of route prediction and dynamic power train policies through simulation
questions?comments?
other ideas?