Faculty Chair, PAVE (Princeton Autonomous Vehicle Engineering Smart Driving Cars: 2013 Year in Review By Alain L. Kornhauser Professor, Operations Research & Financial Engineering Director, Program in Transportation Faculty Chair, PAVE (Princeton Autonomous Vehicle Engineering Princeton University Board Chair, Advanced Transit Association (ATRA) Presented at

Outline Automated Vehicles: 1939 -> Beyond 2018 Defining “Automated Vehicles” (aka “SmartDrivingCars”) The 1st 75 years: (1939 -> 2014) 1939 -> DARPA Challenges DARPA Challenges -> Today The next 75 years: (2014 -> 2089) Today -> 2018 (next 5 years) Beyond 2018 (next 70 years) Questions/Discussion

Scope of Surface Transportation Vehicle + Running Surface (aka roadway, guideway, railway) Questions/Discussion

Scope of “Automated Vehicles” Tesla Car Transporter Rio Tinto Automated Truck Rio Tinto Automated Train Tampa Airport 1st APM 1971 Automated Guided Vehicles Copenhagen Metro Elevator Mercedes Intelligent Drive Rivium 2006 -> Milton Keynes, UK Aichi, Japan, 2005 Expo CityMobil2 Heathrow PodCar

Preliminary Statement of Policy Concerning Automated Vehicles What is a SmartDrivingCar? Preliminary Statement of Policy Concerning Automated Vehicles Level 0 (No automation) The human is in complete and sole control of safety-critical functions (brake, throttle, steering) at all times. Level 1 (Function-specific automation) The human has complete authority, but cedes limited control of certain functions to the vehicle in certain normal driving or crash imminent situations. Example: electronic stability control  Level 2 (Combined function automation) Automation of at least two control functions designed to work in harmony (e.g., adaptive cruise control and lane centering) in certain driving situations. Enables hands-off-wheel and foot-off-pedal operation. Driver still responsible for monitoring and safe operation and expected to be available at all times to resume control of the vehicle. Example: adaptive cruise control in conjunction with lane centering Level 3 (Limited self-driving) Vehicle controls all safety functions under certain traffic and environmental conditions. Human can cede monitoring authority to vehicle, which must alert driver if conditions require transition to driver control. Driver expected to be available for occasional control. Example: Google car Level 4 (Full self-driving automation) Vehicle controls all safety functions and monitors conditions for the entire trip. The human provides destination or navigation input but is not expected to be available for control during the trip. Vehicle may operate while unoccupied. Responsibility for safe operation rests solely on the automated system & Trucks SmartDrivingCars

Automated Vehicles Operating in their own Exclusive Roadway In the late 60s… Some thought that: “The automation & computer technology that took us to the moon could now revolutionize mass transit and save our cities from the onslaught of the automobile” Donn Fichter “Individualized Automatic Transit and the City” 1964 Westinghouse Skybus Late 60’s- PRT APM

APM Automated People Movers Automated Vehicle, Exclusive Guideway, Long Headway (Large Vehicles) Now exist in essentially every Major Airport and a few Major Activity Centers

PRT Personal Rapid Transit Automated Vehicle, Exclusive Guideway, Short Headway (Small Vehicles) Starting in the early 70’s: U of Minnesota became the center of PRT research focused on delivering auto-like ubiquitous mobility throughout urban areas J. Edward Anderson William Garrard Alain Kornhauser Since Demand very diffuse (Spatially and Temporally): Many stations served by Many small vehicles (rather than a few large vehicles). Many stations Each off-line with interconnected mainlines To minimize intermediate stops and transfers Many small vehicles Require more sophisticated control systems, both longitudinal and lateral.

PRT Personal Raid Transit Some early test- track success…

Was built and operational for many years DFW AirTrans PRT Was built and operational for many years

Remains a critical mobility system today & Planning an expansion Morgantown 1975 Video1 Video2 Remains a critical mobility system today & Planning an expansion

We can Build and Operate them Safely Today… We can Build and Operate them Safely Morgantown 1975 Remains a critical mobility system today & planning an expansion

Masdar & Heathrow are Operational; Suncheon in testing Masdar, Abu Dhabi Heathrow PodCar Suncheon, Korea

Far-term Opportunities for Driverless, Short Headway, Exclusive Guideway Transit Each Year: My students Design a NJ-wide PRT network Objective: to effectively serve essentially all NJ travel demand (all 30x106 daily non-walk trips) Locate Stations so that “every” demand point is within “5 minute walk” of a station; “efficiently” interconnect all stations; maintain existing NJ Transit Rail and express bus operations ) Typically: ~10,000 stations ~10,000 miles of guideway

County Stations Miles Atlantic 191 526 Middlesex 444 679 Bergen 1,117 878 Monmouth 335 565 Burlington 597 488 Morris 858 694 Camden 482 355 Ocean 540 1,166 Cape May 976 497 Passaic 1185 1,360 Cumberland 437 1,009 Salem 285 772 Essex 595 295 Somerset 568 433 Gloucester 412 435 Sussex 409 764 Hudson 467 122 Union 577 254 Hunterdon 405 483 Warren 484 Mercer 413 403 Total 11,295 12,261 Marginal

Cost of Far-term Driverless Guideway Transit Each Year: students Design a NJ-wide PRT network Objective: to effectively serve essentially all NJ travel demand (all 30x106 daily non-walk trips) Locate Stations so that “every” demand point is within “5 minute walk” of a station; “efficiently” interconnect all stations; maintain existing NJ Transit Rail and express bus operations ) Typically: ~10,000 stations: @ $2.5M/station………….. ~$ 25B ~10,000 miles of guideway: @ $10M/mile… ~$ 100B ~750,000 PRT vehicles: @ $100K/vehicle … ~$ 75B Optimistic Capital Cost: …………………………. ~$ 200B NJ’s Personal Expenditures for Mobility: …… ~$ 25B/year

But, Deployment has been Excruciatingly Slow … Many years ago…The Executive Director of APTA put his arm around me and said: “Alain… Personal Rapid Transit (PRT) is the ‘System of the Future’, And….. Always Will Be!!!

What he was saying was… Final Region-wide system would be really great, but… Any great Final System MUST evolve from some great (feasible) initial system and MUST be great (feasible) at every step along the way, otherwise… It will always be “a system of the future”. The dedicated grade-separated guideway infrastructure requirement of PRT may simply be too onerous and risky for it to be feasible let alone great at any point except at the ultimate nirvana which is unreachable. Yipes!

What about the Personal Car (Vehicles purchased by consumers for personal mobility)

Automated Cars: Pre-DARPA Challenges Focused on Creating New Roadways To be used Exclusively by Automated Vehicles GM Futurama @ 1939 World’s Fair Zworykin & Flory @ RCA-Sarnoff in Princeton, Late 50s* * VK Zworykin & L Flory “Electronic Control of Motor Vehicles on Highways” Proc. 37th Annual Mtg Highway Research Board, 1958 Cruise Control (1958) 1945 Invented by Ralph Teetor 1958 Chrysler Imperial 1st intorduction GM first offered on the 1959 model year Cadillac.

Automated Cars: Pre-DARPA Challenges Anti-lock Braking Systems (ABS) (1971) (wikipedia) Chrysler & Bendix Corporation, introduced a computerized, three-channel, four-sensor all-wheel[6] ABS called "Sure Brake" for its 1971 Imperial. Ford added an antilock braking system called "Sure-track" to the rear wheels of Lincoln Continentals as a standard option General Motors :"Trackmaster" rear-wheel only ABS . An option on rear-wheel drive Cadillac models and the Oldsmobile Toronado. “Wire Sniffing” Cars Robert E Fenton @ OSU, Early 70s* * “A Headway Safety Policy for Automated Highway Operations” R.E. Fenton 1979

Automated Cars: Pre-DARPA Challenges National Automated Highway System Research Program (NAHSRP) (1994-97) Goal to develop specifications for a fully automated highway system Benefits: #1. Roadway capacity (led to focus on platooning) #2. Safety #3…. Robust to Weather, Mobility, Energy, Land Use, Commercial efficiency, travel time Focused on automated roadways serving only fully automated vehicles. References The National Automated Highway System That Almost Was; TRB Special Report 253 “National Automated Highway System Research Program: A Review” Cheon, Sanghyun “An overview of Automated Highway Systems (AHS) and the Social and Institutional Challenges they Face” (Excellent critique, well worth reading) Video Summary of NAHSRP (compare the tone to what Mercedes did in Frankfurt. We’ve come a very long way!) Program Cancelled! No Feasible Initial Condition No reason to build Automated Highway If there are no Automated Cars & visa versa.

DARPA Challenges 2004 CMU “Red Team” 2005 StanfordRacing Team 2007 CMU “Tartan Team” DARPA Grand Challenge Created in response to a Congressional and DoD mandate: a field test intended to accelerate R&D in autonomous ground vehicles that will help save American lives on the battlefield. 

2005 2007 Slide 49 Link to Presentation Not Easy Old House 2005 2007

Prospect Eleven & 2005 Competition

the making of a monster

2005 Grand Challenge

Objective Constraints Guiding Principles Enrich the academic experience of the students Constraints Very little budget Simplicity Guiding Principles

a homemade set of gears to drive their pickup. I could see from the “Unlike the fancy “drive by wire” system employed by Stanford and VW, Princeton’s students built a homemade set of gears to drive their pickup. I could see from the electronics textbook they were using that they were learning as they went.” http://www.pcmag.com/slideshow_viewer/0,1205,l=&s=1489&a=161569&po=2,00.asp

Fall 2004

Fall 2005

It wasn’t so easy…

(a video presentation) Pimp My Ride (a video presentation)

Achievements in the 2005

Link to GPS Tracks

Participation in the 2007

Prospect12_TestRun

Today.. Continuing to work on Prospect 12 Vision remains our focus for depth mapping, object recognition and tracking Objective is to pass NJ Driver’s Test.

Fundamental Contribution of DARPA: 2007 CMU “Tartan Team” 2004 CMU “Red Team” 2005 StanfordRacing Team Change the mind set to: focus on the Individual Vehicle Make it “Autonomous”: Able to “Drive” without any outside help, behavior modification of conventional roadway users and, physical changes in existing roadways.

Automated Vehicles: Post-DARPA Challenges (2007 -> today) Driverless (Level 4); Non-Exclusive Roadway; Low Speed (< 25kph) CityMobil (2008 -> 2012) Low speed (< 25 kph), Lidar pedestrian detection Non-exclusive simple roadway just lane markings Includes pedestrians & low speed vehicles La Rochelle (CyberCars/Cybus/INRIA CityMobil2 (2012->…) Slated for 5 city deployments + Singapore + ???

Where Are We with Automated Cars/SmartDrivingCars?

Current State of Automated Cars Much of the Public Interest has been induced by the Self-Driving Car. http://www.youtube.com/watch?v=cdgQpa1pUUE It is not driverless… Not yet What did they do? Adopted the DARPA “mind Set” ( & Sebastian Thrun from Stanford DARPA Team + others) Bought “Standard Lexus” (w/ electronically controlled Brakes, Throttle & Steering (BTS)) “Straped on” Sensors: LIDAR’s “depth point cloud”, GPS, Radars & Cameras (on the expensive side) Communications: to access 3D digital map data Software & Computer: Converts Sensor Output into BTS Inputs Developed a self-driving vehicle that can operate in the existing environment. Stated Motivation: >90% of road traffic accidents involve human error. So… remove the human from the loop. Have Self-Driven over 500,000 miles on Existing Roads in Existing Traffic Conditions http://www.youtube.com/watch?v=cdgQpa1pUUE

SmartDrivingCars: Post-DARPA Challenges (2010-today) VisLab.it (U. of Parma, Italy) Had assisted Oshkosh in DARPA Challenges Stereo vision + radars (Video has no sound)

SmartDrivingCars: Post-DARPA Challenges (2010-today) University Efforts: CMU + GM Stanford + VW Va Tech + Google? Oxford + Nissan Parma + Fiat Princeton (PAVE) …

Expected Safety Implications of “Google Cars” Using AAA Liability Rates per Fatality and Injury DOT HS 810 767 Pre-Crash Scenario Typology for Crash Avoidance Research

Expected Safety Implications of “Google Cars” Using AAA Liability Rates per Fatality and Injury

Expected Safety Implications of “Google Cars” Using AAA Liability Rates per Fatality and Injury

Expected Safety Implications of “Google Cars” $475/yr are “Pass-through” Dollars Assume: Emergence of “Price Leaders” $450/yr Discount for Me $25/yr for Flo or the Gecko or ??? To enhance “Combined Ratio” Could Discount Finance: ??

What are the Current Challenges?

Google has “shown”: “We can build it!” Don Draper (Mad Man) Has Done a Great Job Convincing Us that We Love to Drive! (Or, do we?? Are we any good at it?) Safety Hasn’t Been Great at Selling Cars What is salient the “Value Proposition”? What will Consumers Buy?

The Automobile’s 1st 125 Years (1886-2011) Benz patent 1886 1st Automobile Benz Circa 2011 Delivered: Enormous Personal Freedom & Mobility But…Safe Operation Requires Continuous Vigilance

We Love the Freedom & Mobility But…Continuous Vigilance is an unrealistic requirement for drivers

Txtng while driving is out of control…

TravelTainment Industry Wants Everyone’sAttention

In In 717 out of 723 accidents ((99%) http://orfe.princeton.edu/~alaink/SmartDrivingCars/NHTSA_Hendricks2001_UnsafeDrivingActs.pdf In In 717 out of 723 accidents ((99%) “In 717 out of 723 crashes (99%), a driver behavioral error caused or contributed to the crash”

Early Estimate of Motor Vehicle Traffic Fatalities in 2012

But… Not Likely to be Effective Response is Laudable Kirkland, WA But… Not Likely to be Effective

Up to today: The Primary Purview of Crash Mitigation (air bags, seat belts, crash worthiness, …) Up to today: The Primary Purview of Good News: Effective in reducing Deaths and Accident Severity Bad News: Ineffective in reducing Expected Accident Liability & Ineffective in reducing Insurance Rates

Preliminary Statement of Policy Concerning Automated Vehicles Extending its vehicle safety standards from Crash Mitigation to Crash Avoidance with Aim at Full Self-Driving Automation Level 0 (No automation) The human is in complete and sole control of safety-critical functions (brake, throttle, steering) at all times. Level 1 (Function-specific automation) The human has complete authority, but cedes limited control of certain functions to the vehicle in certain normal driving or crash imminent situations. Example: electronic stability control  Level 2 (Combined function automation) Automation of at least two control functions designed to work in harmony (e.g., adaptive cruise control and lane centering) in certain driving situations. Enables hands-off-wheel and foot-off-pedal operation. Driver still responsible for monitoring and safe operation and expected to be available at all times to resume control of the vehicle. Example: adaptive cruise control in conjunction with lane centering Level 3 (Limited self-driving) Vehicle controls all safety functions under certain traffic and environmental conditions. Human can cede monitoring authority to vehicle, which must alert driver if conditions require transition to driver control. Driver expected to be available for occasional control. Example: Google car Level 4 (Full self-driving automation) Vehicle controls all safety functions and monitors conditions for the entire trip. The human provides destination or navigation input but is not expected to be available for control during the trip. Vehicle may operate while unoccupied. Responsibility for safe operation rests solely on the automated system

Preliminary Statement of Policy Concerning Automated Vehicles What the Levels Deliver: Levels 1 -> 3: Increased Safety, Comfort & Convenience

“Level 2- Combined Automation wit Constant Vigilance ” Where Are We Now? Available in ShowRooms for Consumers “Level 2- Combined Automation wit Constant Vigilance ”

What’s in the Showroom Today? A number of car companies have been offering active Forward Collision Avoidance Systems for several years Volvo Acura Mercedes Subaru … http://www.iihs.org/iihs/sr/statusreport/article/47/5/1

Preliminary Results are Disappointing

But what has been sold so far doesn’t really work…

Tested: Autonomous Emergency Braking (AEB) Systems

Tested: Autonomous Emergency Braking (AEB) Systems Model Score: City Rating: Inter-Urban Mercedes-Benz E-Class 3.0 points Good (video) 2.7 points Good Volvo V40 2.1 points 2.2 points Mitsubishi Outlander 1.9 points Adequate Volvo XC60 Adequate (video) - - - Fiat 500L 1.8 points Ford Focus 1.7 points Volkswagen Golf Honda Civic 0.44 points Marginal (video)

However, It seems that some manufacturers are finally …. Getting Serious about Collision Avoidance And Serious about using it sell cars And Serious about having it work

Click images to view videos S-Class WW Launch May ‘13 MB @ Frankfurt Auto Show Sept ‘13 Intelligent Drive (active steering  ) Volvo Truck Emergency braking MB Demo Sept ‘13

Enhancements to Driver Asst. Package History and Development of DISTRONIC: Price reduction, intelligent packaging and availability cross-carline MY 2000 MY 2006 - Current MY 2014 - Future DISTRONIC PLUS Autonomous Braking Intervention In “Driver Assistance Package” with Blind Spot Assist/Lane Keeping Assist $2,950 Available Cross-Carline on almost every model Enhancements to Driver Asst. Package Steering Assist BAS with Cross-Traffic Assist PRE-SAFE Brake with Pedestrian Detection PRE-SAFE PLUS – protection during rear collisions In “Driver Assistance Package” with Blind Spot Assist/Lane Keeping Assist +$2,800 Only available on S/E Introduce DISTRONIC Adaptive Cruise Control In package for +$3,700 Only available on S/CL

Near Term Opportunities DOT HS 810 767 Pre-Crash Scenario Typology for Crash Avoidance Research $475 are Pass-through Dollars

Could “Pass-through” Finance Mercedes Intelligent Drive?? (Available in ‘14 S-Class) May well Deliver 2/3rds the safety of Google Car $475 Pass-through becomes: $320 $20/yr for Flo or the Gecko $300/yr to Mercedes Therefore: I get Intelligent Drive for Free (thank you Google) !! Plus: Prob. of my car killing me is reduced factor: 2/3*.81= 0.54 (half) Prob. of my car injuring me is reduced factor: 2/3*.65= 0.44 I “Save” my expected “deductible self-insurance”: $247/yr I have more Comfort I have more Convenience I have “Anxiety” relief What a great Business Case! I’m a Buyer!

Expected Implications… With Mercedes as the Market Leader offering “Level 2-” SmartDriving Technology at an incremental price tag that can be absorbed by a “Price Leading” Insurance Company, should lead to: Other automakers will be enticed to follow (race seems to have already started) Viral adoption by the car buying public “Moore’s Law type of price/performance improvement Market-driven Transition to “Level 2” and “Level 3” at same or even lower price structure Adoption and enhancement rates that are comparable to that enjoyed by airbags may be likely

Where might We End Up?

Preliminary Statement of Policy Concerning Automated Vehicles What the Levels Deliver: Levels 1 -> 3: Increased Safety, Comfort & Convenience Primarily a Consumer Play Level 4 (Driverless Opportunity) : Mobility, Efficiency, Equity Revolutionizes “Mass Transit” by Greatly Extending the Trips that can be served @ “zero” cost of Labor. (That was always the biggest “value” of PRT; zero labor cost for even zero-occupant trips) Primarily a Fleet Play

Initial Deployment… Needs to be Driverless Doesn’t Need To Be… With Excellent Pedestrian Recognition Doesn’t Need To Be… Fast Everywhere Let’s start Slow and Narrow: Like CityMobile2… Say 10-15 mph Along a Corridor, or Throughout one of Florida’s Retirement Communities Slide 98

What About……

What About…… Driverless electric shuttle to be trialled in Singapore (video of Luxembourg Demonstration)  Slide 84

What About……

Far-term Opportunities for Driverless Transit Using “All” Existing Roads”

Slide 98 Slide 89

NJ_PersonTrip file 9,054,849 records   All Trips Home County Trips TripMiles AverageTM # Miles ATL 936,585 27,723,931 29.6 BER 3,075,434 40,006,145 13.0 BUC 250,006 9,725,080 38.9 BUR 1,525,713 37,274,682 24.4 CAM 1,746,906 27,523,679 15.8 CAP 333,690 11,026,874 33.0 CUM 532,897 18,766,986 35.2 ESS 2,663,517 29,307,439 11.0 GLO 980,302 23,790,798 24.3 HUD 2,153,677 18,580,585 8.6 HUN 437,598 13,044,440 29.8 MER 1,248,183 22,410,297 18.0 MID 2,753,142 47,579,551 17.3 MON 2,144,477 50,862,651 23.7 MOR 1,677,161 33,746,360 20.1 NOR 12,534 900,434 71.8 NYC 215,915 4,131,764 19.1 OCE 1,964,014 63,174,466 32.2 PAS 1,704,184 22,641,201 13.3 PHL 46,468 1,367,405 29.4 ROC 81,740 2,163,311 26.5 SAL 225,725 8,239,593 36.5 SOM 1,099,927 21,799,647 19.8 SOU 34,493 2,468,016 71.6 SUS 508,674 16,572,792 32.6 UNI 1,824,093 21,860,031 12.0 WAR 371,169 13,012,489 35.1 WES 16,304 477,950 29.3 Total 32,862,668 590,178,597 19.3  9,054,849 records One for each person in NJ_Resident file Specifying 32,862,668 Daily Person Trips Each characterized by a precise Origination, Destination and Departure Time

“Pixelated” New Jersey (“1/2 mile square; 0.25mi2) aTaxi “PRT” Concept During peak demand aTaxis operate between aTaxiStands Autonomous vehicles wait for walk-up customers Located in “center” of each pixel (< ~ ¼ mile walk) Departure is Delayed to facilitate ride-sharing Focus of Analysis: what is the ride-share potential? Ridesharing delivers: Congestion relief Energy savings Reduced costs/passenger Environmental sustainability

Warren County Population: 108,692 Intra-pixel Trips

NJ Transit Train Station “Consumer-shed”

“Pixelated” New Jersey (“1/2 mile square; 0.25mi2) aTaxi Concept – (PRT) Model Personal Rapid Transit Model aTaxi Concept – SPT Model Smart Para Transit Transit Model Ref: http://orfe.princeton.edu/~alaink/Theses/2013/Brownell,%20Chris%20Final%20Thesis.pdf

New Jersey Summary Data Item Value Area (mi2) 8,061 # of Pixels Generating at Least One O_Trip 21,643 Area of Pixels (mi2) 5,411 % of Open Space 32.9% # of Pixels Generating 95% of O_Trips 9,519 # of Pixels Generating 50% of O_Trips 1,310 # of Intra-Pixel Trips 447,102 # of O_Walk Trips 1,943,803 # of All O_Trips 32,862,668 Avg. All O_TripLength (miles) 19.6 # of O_aTaxi Trips 30,471,763 Avg. O_aTaxiTripLength (miles) 20.7 Median O_aTaxiTripLength (miles) 12.5 95% O_aTaxiTripLength (miles) 38.0

State-wide automatedTaxi (aTaxi) Serves essentially all NJ travel demand (32M trips/day) Shared ridership potential: Slide 98

State-wide automatedTaxi (aTaxi) Serves essentially all NJ travel demand (32M trips/day) Shared ridership potential:

State-wide automatedTaxi (aTaxi) Fleet size (Instantaneous Repositioning)

State-wide automatedTaxi (aTaxi) Abel to serve essentially all NJ travel demand (32M trips/day) Shared ridership allows Peak hour; peak direction: Av. vehicle occupancies to can reach ~ 3 p/v and eliminate much of the congestion Essentially all congestion disappears with appropriate implications on the environment Required fleet-size under 2M aTaxis (about half) (3.71 registered automobiles in NJ (2009)

Discussion! Thank You alaink@princeton.edu www.SmartDrivingCar.com