8th edition of the 6th year of SmartDrivingCars
Broadening Understanding of the Interplay Between Public Transit, Shared Mobility, and Personal Automobiles
S. Feigon, Jan 2018, Pre-publication draft of TCRP Research Report 195 “Urban mobility is rapidly evolving in the United States, particularly since the introduction o appbased transportation network companies (TNCs) such as Uber and Lyft. A these services become more widespread, man have begun to question what effect they are having on the cities where they operate, including on public transit ridership, singleoccupancy vehicle trips and traffic congestion. In the face of widespread declines in public transit ridership after a decade or more of growth nationally, these questions have become especially pressing. Speculation has grown around whether TNCs are leading to real changes in how people use public transit and private automobiles, or if these fluctuations are caused by other factors.
Key findings from this research include:...3. There is no clear relationship between the level of peakhour TNC use and longerterm changes in the study regions’ public transit usage. …” Read more Hmmmm… Very interesting report. Worth reading. My opinion: The evolution of TNCs in the past “5 years” is an approximate blueprint of what can be expected in the first “5 years” of autonomousTaxis, give-or-take. That’s why this is worth reading. Alain
Hop in for episode 25 of the Smart Driving Cars Podcast with Princeton’s Alain Kornhauser, co-host Fred Fishkin and guest Nir Erez, CEO and co-founder Moovit. How will Moovit’s app play a role in the self driving future? Plus…California and Arizona move towards allowing truly Driverless trials without attendants, Nissan’s self-driving taxis in Japan, and more from NVIDIA, Uber, Lyft, Tesla and Google.
Hmmmm…. Now you can just say “Alexa, play the Smart Driving Cars podcast!” . Ditto with Siri, and GooglePlay. Alain
Real information every week. Lively discussions with the people who are shaping the future of SmartDrivingCars. Want to become a sustaining sponsor and help us grow the SmartDrivingCars newsletter and podcast? Contact Alain Kornhauser at firstname.lastname@example.org! Alain
P. Moorhead, Feb 8, “Every year, NVIDIA holds a series of GPU Technology Conferences, commonly referred to as GTC and gets kicked off with a big event in San Jose… If you or your company is involved or interested in AI and ML, GTC is the place to be whether you are a developer, a data scientist or a business person. This year, NVIDIA says GTC features more than 500 hours of scientists, researchers, and developers from leading institutions sharing their work — not just NVIDIA presenters. It also includes leading AI thinkers from …” Read more Hmmmm… I’ll be there with 4 of my students and recommend it highly. If you haven’t registered, use the VIP registration code NVDASHAPIRO for a 25% discount (Thank you Danni). Alain
S. Hanley, Feb 17, “… Later, Reddit user mikhpat added these details. “The driver of the Tesla is my dad’s friend. He said that he was behind a pickup truck with AP engaged. The pickup truck suddenly swerved into the right lane because of the firetruck parked ahead…” Read more Hmmmm… As I originally speculated. Is the fault here AutoPilot or the Automated Emergency Braking (AEB) system (or both). As I’ve pointed out, AEB are DESIGNED to NOT activate if: stationary object detected ahead and brake pedal not “touched” and crash is NOT imminent (def: crash remains avoidable). Once crash is imminent (unavoidable), then brakes are applied to reduce the impact velocity. This is a NON-GOOD design because the AEB explicitly will not function to avoid this crash, irrespective of AutoPliot being engaged or disengaged. The AEB MUST be redesigned to function in such situations: Following moving vehicle, moving vehicle moves out of lane followed by an immediate sensing of a stationary object in lane ahead, apply brakes gradually, say as per intelligent cruise control, in order to never reach the situation in which a crash is imminent. If a collision is imminent at the initial sensing of the stationary object, begin applying brakes at maximum braking force until either, vehicle stops, crashes into stationary object or stationary object ahead disappears (it was a false alarm) or begins moving so that a crash is no longer imminent. Or something like that.
Maybe this is already the design of Tesla’s AEB and, in the Tesla 3 situation above, the lead car moved out of the lane too late to leave enough room for the Tesla to stop. That is possible; however, I doubt it. While the Tesla driver may be faulted for not paying attention because AutoPilot was on, the driver ahead was paying attention and likely moved over early and may have even slowed down before moving over, in order to not hit the firetruck. So I’ll suspect that there was plenty of stopping distance available for a working AEB to stop short of the firetruck.
The data from the Tesla should tell us if my speculation is valid. It will tell us the “exact” time, location, & speed of the Tesla when the vehicle ahead had changed lanes, time & location when AutoPilot’s AEB sensed the stationary object, and the location, speed and brake force every 1 /10th (or 1/20th) of a second the instant AEB began to apply the brakes until the Tesla came to rest after the crash. If: the brakes came on at max level “instantaneously” with the sensing of the stationary object ahead, then I MUST “eat crow“. A huge plate of it. Else: NHTSA has do something! Maybe NHTSA needs to recall all AEBs and have them fixed, the way it recalled airbags. Yipes!!! (I hope I have to eat crow!).
Or, maybe the AEB never “saw” the stationary object. If that’s the case, then this AEB MUST be recalled and AutoPilot turned off until Tesla has an AEB that works with or without the activation of AutoPilot.
I speculated that it was the AEB that didn’t work in the Joshua Brown Florida crash. It is my understanding from the Tesla data log that the AEB never activated. The “excuse” given was that the “Mobileye” vision system could not detect a white trailer. But the tractor had black tires, etc. that passed in the lane ahead of the white trailer. Why were those detected? Were they detected, but was it something else in the AEB logic that failed to activate the AEB. For example, longitudinal speed of object essentially zero, it is still early, fuhgettaboutit!
The scenario may well have been: Truck begins crossing Brown’s lane. Truck cab enters Brown’s lane ahead with some cross-lane speed (which is not measurable by any of the Tesla sensors/algorithms), but essentially zero speed in the direction Brown is traveling, therefor it “looks” like a stationary object instantly appearing to Tesla’s AEB. At that time, Brown may have been too far away. A crash was not imminent … AEB doesn’t activate. Now the trailer enters Brown’s lane, crash is now imminent, but because the trailer doesn’t have a skirt, the radar “sees” under the truck and reports no object ahead, so AEB is not activated… Bang! … all in less than 2 seconds or so!? 🙁 Alain
A. Barr, Feb 16, “The first U.S. commercial ride-hailing service without human drivers has been approved.
Waymo, a unit of Alphabet Inc., got a permit in late January from the Arizona Department of Transportation to operate as a Transportation Network Company, according to Ryan Harding, a spokesman at the state agency.
The designation lets Waymo’s fleet of driverless Chrysler Pacifica minivans pickup and drop off paying riders in Arizona through a smartphone app or website …” Read more Hmmmm… More than a passing detail if the permit explicitly allows Waymo to do it without a driver or attendant in the vehicle. The ADoT TNC Compliance requirements include: “2. Inside/Interior Signage: Statute requires the TNC digital network or software application to display a picture of the TNC driver…”. How did/will Waymo comply???? Alain
Exclusive: Self-driving vehicles without any human backups could be on California roads as soon as April
A. Green, Feb 21, “Self-driving vehicles without any backup driver in them could be allowed on California roads as soon as April under California Department of Motor Vehicles rules up for approval Monday ― even sooner than the previously anticipated June launch.
The new rules have been with the Office of Administrative Law for review since January and are expected to be approved Monday, a DMV spokeswoman told the Business Times. If the regulations are approved, the DMV could issue a public notice soon after and start approving applications 30 days later. That means the state could issue permits as soon as April 2 for fully driverless vehicles. A remote operator would be required to be able to take over if anything goes wrong… ” Read more Hmmmm… This is for TESTING …. a huge advance for California, but does NOT permit this to be a business, …YET. Just another of the substantial hurdles that need to be negotiated in the process of going from concept to a viable business that delivers value to some normal folks. 🙂 Alain
F. Salmon, Feb 18, “…And so a tax naturally emerges…” Read more Hmmmm… Maybe, but not what is proposed here. 1st, Felix hasn’t read Sharon’s report, so his arguments are not sound. 2nd, High-quality Ride-sharing is the way to address congestion. Unfortunately classical Ride-sharing the Public Transit way is, for many, not high-quality. When it is, it is used. When it’s not, then it’s not. Which is where the TNCs come in. But TNCs have the opportunity to provide high-quality ride-sharing, two, three rides at a time… which would take one, two, .. three cars off the road at those time. So the focus needs to be to incentivize TNC ride-sharing and to de-centivize single occupant ride-sharing. So… if the suggestion is to tax TNCs in congested areas at congested times in congested directions, then tax Single-occupant TNCs ONLY!!!! Alain
M. Bergen, Feb 21, “On any given day, there could be a half dozen autonomous cars mapping the same street corner in Silicon Valley. These cars, each from a different company, are all doing the same thing: building high-definition street maps, which may eventually serve as an on-board navigation guide for driverless vehicles.
These companies converge where the law and weather are welcoming―or where they can get the most attention. For example, a flock of mapping vehicles congregates every year in the vicinity of the CES technology trade show, a hot spot for self-driving feats. “There probably have been 50 companies that mapped Las Vegas simply to do a CES drive,” said Chris McNally, an analyst with Evercore ISI. “It’s such a waste of resources.”…” Read more Hmmmm… I agree! However, I agree for different reasons… I’ll argue that you don’t even want them unless they are essentially free. That’s because they aren’t sufficient to enable any of the vehicle automation tasks in any of the driving scenarios that really matter. Yes, you might want them if you want to drive your Jeep down the river bed that Jeep has conveniently brain washed you to want to drive down in the commercial that enticed you to buy the Jeep in the first place, but not many other places.
What is important if I (or the AI) want to stay in my lane and not crash into things is for me/AI to know where I/AI am relative to lane markings, stationary things that might be located on a map and moving things that have no hope of being on any map. In order for the “HD” aspects of HDmaps to be of any use is for me/AI to know where I/AI am to a level of precision that is as good or better that the HDmap precision. I/AI can then compute the difference in order to stay in the lane and not crash into things. Unfortunately, I only have vague independent knowledge about “where am I?”. Even Differential GPS isn’t good enough in enough places at enough times for my AI to know where it is to sufficient precision to use the HD aspects of HD maps. SLAM is useful only in so far that it has access to sensors that can measure relative distance to objects that it can see and recognize that are HD Located in the HDmaps. SLAM is of no help with objects (pedestrians, bikes, other cars) that are not located in HD maps. For those object I/AI need a mechanism/sensor that can “see” those objects, recognize them and determine a relative position (and velocity, acceleration, …) to a level of precision that will keep me and them safe. This is a necessary condition. To be sufficient, I/AI also needs to be able to stay in my lane and not crash into stationary things. But since lane markings are stationary and stationary things are easier to “see”, recognize and determine a relative position than things that are moving, then whatever I’ve created that doesn’t use HD maps to satisfy the necessary condition is now sufficient in satisfying the safety condition. Therefore, HD maps are not useful and a “waste of resources“. Alain
Press release, Feb 21, “Moovit today announced it has closed a $50 million Series D round led by Intel Capital. All Moovit’s earlier investors participated, including Sequoia Capital, BMW iVentures, NGP, Ashton Kutcher’s Sound Ventures, BRM, Gemini, Vaizra, Vintage, and newcomer Hanaco.
Moovit’s free app provides comprehensive transit information to more than 120 million users in more than 2,000 cities in 80 countries. The company has amassed the world’s largest repository of transit data and generates more than one billion movement data points a day. ..” Read more Hmmmm… It is all about scale of user crowd sourcing. Given their scope they need real scale. I wonder what they define as a “user”. Throughout the US there are about 15M transit users (use transit more than occasionally). If 10% are Moovit “users”, that means that roughly 99% of Moovit’s users are non-US or US-occasional-users-of-transit. Alain
R. Lohmann, Jan Feb 2018, “… In public discussion and industry debate on autonomous cars, the concepts of automated, shared, and electric are often cheerfully confused. The automated car doesn’t reduce the number of cars on the road – ride-sharing does. It won’t make the city more sustainable – only electricity from a sustainable power source will. And yes, automation will improve traffic safety immediately for people who were previously driving themselves – however, autonomous cars will need to be at least ten times safer to match the current safety levels of public transit (in developed countries)….” Read more Hmmmm… Yes!! Please read on. Alain
Feb, 2018, “…SAFETY: Reduce Transportation-Related Fatalities and Serious Injuries Across the
Safety has consistently been DOT’s top strategic and organizational goal. To improve transportation safety, DOT seeks to…
Strategic Objective 1: Systemic Safety Approach: Mitigate risks and encourage… ensure that automation brings significant safety benefits…” Read more Hmmmm… At least automation is (far down) on the “Approach” list and “Connected” didn’t make it at all. 🙂 Unfortunately, “Mitigate” remains front and center, rather than sharing the podium with “Avoid/Eliminate”, especially as it might pertain to NHTSA who has done an excellent job “mitigating” crashes, but by embracing automation it could do an equally excellent job “avoiding/eliminating” crashes. I didn’t see this important 2nd dimension being part of the 2018-22 Strategic Plan. Maybe it will gain respect in the 2023-27 Strategic Plan. Alain
Press release, Feb 23, “Nissan Motor Co., Ltd. and DeNA Co., Ltd. will begin a field test of Easy Ride, the robo-vehicle mobility service being developed by both companies, on March 5.
Easy Ride is envisioned as a mobility service for anyone who wants to travel freely to their destination of choice in a robo-vehicle. During the field test, in the Minatomirai district of Yokohama, in Japan’s Kanagawa Prefecture, the participants will be able to travel in vehicles equipped with autonomous driving technology along a set route. The route spans about 4.5 kilometers between Nissan’s global headquarters and the Yokohama World Porters shopping center. For efficient fleet operation and customers’ peace of mind, Nissan and DeNA have set up a remote monitoring center that uses the two companies’ advanced technologies….”
Read more Hmmmm… Sounds really great; UNFORTUNATELY, the operative word is: envisioned . The video looks like Driverless; the Press Release reads like Driverless; the remote monitoring center sounds like Driverless, but envisioned says: still HYPE not REALITY! 🙁 Alain
J. Boyd, Feb 16, “…we have developed an AI-based object-recognition technology that can instantly detect objects up to about 100 meters away…with an 81 percent accuracy….”
To achieve this, the Mitsubishi system uses two technology processes consecutively. A computational visual-cognition model first mimics how humans focus on relevant regions and extract object information from the background even when the objects are distant from the viewer. The extracted object data is then fed to Mitsubishi’s compact deep learning AI technology dubbed Maisart. The AI has been taught to classify objects into distinct categories: trucks; cars; and other objects such as lane markings. The detected results are then superimposed onto video that appears on a monitor for the driver to view. …” Read more Hmmmm… Do you really want to spend time putting cars and trucks into different categories??? Those scarce resources should be used to improve on the “81% accuracy”???? Within 100 meters you basically want to know with the fewest false positives or negatives: Is there something in my lane that I shouldn’t hit and how fast am I gaining on it. The rest is largely irrelevant. (I’m sure I’m forgetting something, but you get my point). 🙂 Alain
Some other thoughts that deserve your attention
Half-baked stuff that probably doesn’t deserve your time
L. Stangel, Feb 21, “Hackers broke into Tesla’s Amazon Web Services account and used the automaker’s cloud computing capacity to secretly mine cryptocurrency, security researchers say. Tesla confirmed the hack on Tuesday, telling reporters the company had closed the vulnerability within hours of being notified by security researchers…” Read more Hmmmm… Whatever… That’s like using your fake plates with Treasury Department printers to make counterfeit “Benjamins“. If the miners have to stoop that low, its there anything that is legitimate about BitCoins/Craptocurrency? Alain
J. Hughes, Feb 22, “Americans are still hesitant to let their cars do the driving, according to a Gallup poll performed…This is all very interesting, but what is unclear is exactly how the survey defines a self-driving car. …But despite the ambiguous definition…” Read more Hmmmm… What???? If you ask for opinions about something that is so ambiguous, how can you conclude anything about anything except: “This survey is Flawed”. We all know: if it is “Garbage In”; it will be “Garbage Out”. This whole subject matter is ambiguous even to the people who deal with it every day.
To whatever number of of “U.S. adults” Gallup tried to “gauge attitudes” on something Gallup didn’t precisely define is flawed from the very beginning. They can’t even do a credible job on tangible things like, are you going to vote for Hillary? Alain
L. Estrada, Feb 17, “…But its tiny Smart division has other ideas. Last month, transportation reporter Sean O’Kane got a “ride” in Smart Vision EQ Fortwo concept ― Daimler’s idea of what a self-driving urban car should be. …But back to the present and back to Smart’s present. The brand went all-electric in the US and Canada for 2018…Merging onto a busy San Diego freeway sounded like a death wish at first, but the electric motor propelled a Smart with two adults easily to 65 mph, and even up to 70 or so to pass some slow-moving Toyota Yarises.
The kicker in all of this, however, is a range of just 58 miles,…” Read more Hmmmm… That’s what you call “Range Anxiety”!! What demographic will say: that’s great!!??? Alain
Assessment of RideSharing, Empty Vehicle Management Needs and ‘Last-Mile’ Ridership Implications on the Existing Rail Transit, Amtrak and Airline Networks Associated with Having autonomousTaxis Efficiently Serve the Billion or so PersonTrips Taken Throughout the US on a Typical Day… Final Project Description
A. Kornhauser, Jan 13, “… What if no one owned a personal car or truck any more? What operational characteristics would a fleet of autonomousTaxis (aTaxis), operating nation-wide, need to have to deliver a comparable level-of-Service (LoS), in conjunction with existing Rail Transit, AmTrak and Airline networks (with appropriately enhanced LoS between existing stations/airports)? How many of what size would be needed? How would they need to be managed? What would be the fundamental economics in order to adequately serve the Billion or so person trips that take place on a typical day across the US? Because details matter, we synthesized each of the 310 or so, million people in the US. For each we synthesized their mobility needs throughout a typical day to accomplish their activities such as get to and from work/school/play/shopping/entertainment/… Preliminary results include…
- In order to deliver a Level-of-Service (LoS) comparable to that offered by today’s conventional automobile in its service of the roughly 1 Billion trips that take place on a typical day across the USA would requite a fleet of approximately 35 million autonomousTaxis (aTaxis).
- In serving those trips throughout the day, those aTaxis would travel almost 50% fewer vehicle miles than today’s road vehicles if:
- people traveling from about the same place at about the same time to about the same place agreeing to ride together, much as they do today in elevators, (shared-ride), accounts for more than 50% of the reduced vehicle miles.
- The remainder comes from offering a reliable and attrative LoS to/from the existing fixed rail transit systems and, surprisingly, to and from existing AmTrak stations but assuming that the assistance of extremely improved AmTrak frequencies if service.
- It is amazing how, across the country, so many segments of the AmTrak network could be of service to so many 100-400 mile trips that take place on a typical day. If these trip makers had a reliable, convenient and affordable way to get from their origin to the nearest AmTrak station AND to their destination from that nearest AmTrak station, then the ridership potential on numerous segments of the AmTrak system beyond the NorthEast Corridor (NEC) would justify a LoS that is even better than what exists today on the NEC.
- If this preliminary result holds up under closer scrutiny (there isn’t an error someplace), this opportunity may be this study’s most significant finding. There is little literature on “long auto trips” yet, because they are “long” they log a significant amount of daily VMT on existing highways. Many of these trips today essentially parallel the AmTrak network. By providing convenient “first 1 – 20+ mile / last 1 – 20+ mile” accessibility to AmTrak’s existing stations AND by having AmTrak provide a high-quality LoS, the a significant percentage of these travelers would become AmTrak customers.
Very interesting… aTaxis Save AmTrak!! 🙂 More later. Alain
Read more Hmmmm… Most interesting! We hope to have a draft of the final report for all of USA out soon. Alain
Calendar of Upcoming Events:
March 26-29, 2018
San Jose, CA
VIP registration code NVDASHAPIRO for a 25% discount
Self Racing Cars
March 24-25, 2018
Episode 24 of the Smart Driving Cars Podcast with Princeton University’s Alain Kornhauser and tech journalist Fred Fishkin. Today: Softbank making insurance move that could spur smart driving technology? A gas tax hike for infrastructure funding. Will Amazon surprise again with self driving technology of its own? Waymo’s ride hailing app…and NVIDIA continues to soar.
Waymo and Uber settle, Amazon reportedly preparing to launch delivery service to compete with FedEx and UPS, the problem with automatic emergency braking systems and keeping watch over the testing reports for self driving cars. Join Princeton University’s Alain Kornhauser and Fred Fishkin for episode 23 of the Smart Driving Cars Podcast..
Episode 18 of the Smart Driving Cars Podcast with Princeton University’s Alain Kornhauser, co-host Fred Fishkin and guest research engineer Steven Shladover of UC Berkeley. Topics: General Motors, Waymo, the Transportation Research Board, CES, nVIDIA and how #MeToo may impact ride sharing technology in the future.
Episode 13 of the Smart Driving Cars Podcast with host Fred Fishkin and Princeton University Professor Alain Kornhauser. This edition In this edition Fred and Alain are joined by Bernard Soriano, the Deputy Director of the California Department of Motor Vehicles. On the agenda: Waymo’s CEO says real driverless testing is coming soon.; Waymo’s autonomous fleet now has traveled four million miles; Lyft gets the green light from California to test self driving on public roads
Episode 11 of the Smart Driving Cars Podcast with host Fred Fishkin and Princeton University Professor Alain Kornhauser. Fred and Alain are joined by leading expert and Internet pioneer Brad Templeton. Waymo makes some history, Thee tech needed to make it work..cameras…lidar or both? Navya bringing new robotic vehicles to Paris. And an accident…as a self driving shuttle is launched in Las Vegas.