Transcript SmartDrivingCars - Operations Research and Financial Engineering

SmartDrivingCars
Disrupting 15% of World GDP
by
Alain L. Kornhauser, PhD
Professor, ORFE
(Operations Research & Financial Engineering)
Director, CARTS
(Center for Automated Road Transportation Safety)
Faculty Chair, PAVE
(Princeton Autonomous Vehicle Engineering)
Princeton University
Presented at
Schlumberger meeting
May 31, 2016
Princeton, NJ
SmartDrivingCars
• Today: Mobility world-wide: $10T/yr. (15% world GDP)
• Today’s Challenge (Self-Driving) … It is all about SAFETY!
SmartDrivingCars
• Today: Mobility world-wide: $10T/yr. (15% world GDP)
• Today’s Challenge (Self-Driving) … It is all about SAFETY!
• Ultimately (Driverless) …It is all about Enhanced Mobility
•
•
•
•
•
•
“On Demand” for All
Substantially Cheaper
Substantially Safer
Substantially More Energy Efficient
Substantially Lower GHG & Other Pollutants
Substantially Less/Elimination of Road Congestion
2005
Link to Presentation Not Easy Old House
2007
2005
2007
Link to GPS Tracks
In Process of Automating Prospect_13
Will Run Deep Learning Image Processing & Control
Deep Learning for Self-driving Car
http://deepdriving.cs.princeton.edu/paper.pdf
Chenyi Chen
Deep Convolutional Neural Network (CNN)
Convolutional layers
……
Parallel computing
Input image: {210x280x3}, int8 (1 byte)
Fully connected layers
Deep feature: {4096x1}, floats (4 bytes)
Output: {14x1}, floats (4 bytes)
Figure: modified of Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton
http://orfe.princeton.edu/~alaink/NJ_aTaxiOrf467F12/TRIP%20FILES/
http://orfe.princeton.edu/~alaink/NJ_aTaxiOrf467F12/Mufti%20MSE%201st%20DraftAlainReview112512.docx
All Trips
Home
County
ATL
BER
BUC
BUR
CAM
CAP
CUM
ESS
GLO
HUD
HUN
MER
MID
MON
MOR
NOR
NYC
OCE
PAS
PHL
ROC
SAL
SOM
SOU
SUS
UNI
WAR
WES
Trips
#
936,585
3,075,434
250,006
1,525,713
1,746,906
333,690
532,897
2,663,517
980,302
2,153,677
437,598
1,248,183
2,753,142
2,144,477
1,677,161
12,534
215,915
1,964,014
1,704,184
46,468
81,740
225,725
1,099,927
34,493
508,674
1,824,093
371,169
TripMiles AverageTM
Miles
Miles
27,723,931
40,006,145
9,725,080
37,274,682
27,523,679
11,026,874
18,766,986
29,307,439
23,790,798
18,580,585
13,044,440
22,410,297
47,579,551
50,862,651
33,746,360
900,434
4,131,764
63,174,466
22,641,201
1,367,405
2,163,311
8,239,593
21,799,647
2,468,016
16,572,792
21,860,031
13,012,489
29.6
13.0
38.9
24.4
15.8
33.0
35.2
11.0
24.3
8.6
29.8
18.0
17.3
23.7
20.1
71.8
19.1
32.2
13.3
29.4
26.5
36.5
19.8
71.6
32.6
12.0
35.1
16,304
477,950
29.3
Total 32,862,668
590,178,597
19.3
NJ_PersonTrip file
• 9,054,849 records
– One for each person in NJ_Resident
file
• Specifying 32,862,668 Daily
Person Trips
– Each characterized by a precise
• {oLat, oLon, oTime, dLat, dLon, Est_dTime}
http://orfe.princeton.edu/~alaink/NJ_aTaxiOrf467F
12/ORF467F12aTaxiFinalReport.pdf
http://orfe.princeton.edu/~alaink/NJ_aTaxiOrf467F15/Orf467F15
FinalReport/ORF467F15FinalReport_aTaxisForNJA1stLookAtWhatItWouldTake.pdf
http://orfe.princeton.edu/~alaink/SmartDrivingCars/Papers/Zhu,Shirley_Final_Thesis2016.pdf
Results
http://orfe.princeton.edu/~alaink/SmartDrivingCars/Papers/ZachariahTRB_AutonomousTaxis.pdf
http://orfe.princeton.edu/~alaink/NJ_aTaxiOrf467F14/Brownell_Kornhauser_2014TRB_Paper_FINAL.pdf
US_PersonTrip file will have..
• 308,745,538 records
– One for each person in US_Resident file
• Specifying 1,009,332,835 Daily Person
Trips
– Each characterized by a precise
• {oLat, oLon, oTime, dLat, dLon, Est_dTime}
• Will Perform Nationwide aTaxi AVO analysis
• Results
Bottom-Line of Level 4 Driverless
⇒
⇒
⇒
new
=
now
AVO
$now
$new =
8 x AVO
$now
$new =
8
$Capital = ZERO!
new =
X
now
AVO
DeepDriving and autonomousTaxi Research Team & Recent Products
Dr. Chenyi Chen*16 (resume) (Extracting Cognition Out Of Images For The Purpose
of Autonomous Driving, PhD Dissertation 2016)
Christopher Hay’17 (resume) (Risk-Based Collision Avoidance in Semi-Autonomous
Vehicles, Junior Independent Research)
Artur Filipowicz’17 (resume) (Driving School: Incorporating Learning Strategies, &
Video Games for Training & Testing Autonomous Driving Junior Independent Research )
Alexander Garzon'16 (resume) (Convolutional Neural Networks Applied to traffic
Sign Detection in Grand Theft Auto V, BSE Thesis 2016)
Andrew Grant'16 (resume) (Time to Contact for Autonomous Vehicles: Analysis and
Estimation from Image Sequences, BSE Thesis 2016)
Shirley Zhu'16 (resume) (Make America Great Again: autonomousTaxi Fleet
Management Strategies, BSE Thesis 2016)
Suraj Bhat'16 (resume) (Quantifying the Potential for Dynamic Ride-Sharing of New York
City’s Taxicabs, BSE Thesis 2016)
Alexander Schindele’16 (resume) (A Ride-Sharing Analysis with Hitchhiking
Modifications Applied to Taxi Trips in New York, BSE Thesis 2016)
Discussion!
Thank You
[email protected]
www.SmartDrivingCar.com