SmartDrivingCar.com/6.17-NewMode-041218
17th edition of the 6th year of SmartDrivingCars
Thursday, April 12, 2018
The way we regulate self-driving cars is broken—here’s how to fix it
T. Lee, Apr 10,”…Federal car safety regulation has traditionally been based on a thick book of rules called the Federal Motor Vehicle Safety Standards (FMVSS). These regulations, developed over decades, establish detailed performance requirements for every safety-related part of a car: brakes, tires, headlights, mirrors, airbags, and a lot more….
Federal regulations don’t say much about how companies develop and test cars before bringing them to market. … But that approach doesn’t work for driverless cars. Companies can do some testing of driverless cars on a closed course, but it’s impossible to reproduce a full range of real-world situations in a private facility. So at some point, carmakers need to put self-driving cars on public roads for testing purposes—before a manufacturer is able to clearly demonstrate that they’re safe. In effect, this makes the public involuntary participants in a dangerous research project.
But updating the FMVSS is neither necessary nor sufficient for effective regulation of driverless cars…. Read more Hmmmm…What needs to be recognized is that Driverless cars (much more so than Safe- and Self-driving cars) are really a NEW MODE. They are in many ways closer to an elevator than a conventional car. Sure they run on conventional roads and not vertical shafts and they can run into each other and have to deal with conventional drivers and “pedestrians”. but they will not be owned nor operated by consumers, but fleet operators (think buildings) . They will serve demand upon request to everyone and anyone, be shared when appropriate and convenient and don’t even have a driver’s seat, let alone the controls of a conventional car. Driverless cars are enormously different than conventional cars.
Just as railroads and airplanes have their own safety legislation and regulatory administration tailored to their needs, so should Driverless cars. The best way to approach regulation of Driverless is to start fresh by declaring them as a new mode. Alain
Smart Driving Cars Podcast Episode 34
F. Fishkin, Apr 13, “Should a brand new regulatory agency be formed to oversee self driving and driverless vehicles? Princeton University’s Alain Kornhauser says yes in Episode 34 of the Smart Driving Cars podcast with co-host Fred Fishkin. Also…Uber’s CEO calls self driving vehicles are in the student driver phase….and Tesla feuds with the NTSB.”
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 alaink@princeton.edu! Alain
2nd Annual Princeton SmartDrivingCar Summit
Application NOW OPEN
Early Bird Rates End April 15
Become a Sponsor and Promote your Wares
Ambarella Self-Driving Car | Inside Live Test of EVA Embedded Vehicle Autonomy w/ Alberto Broggi
Apr 1, “Ambarella is a Founder led semiconductor company with eyes on the Autonomous Driving prize. I’m a sucker for smaller companies seeking to achieve extraordinary things — “EVA” Embedded Vehicle Autonomy is a complete Level 4 solution, & powered at only 3 watts, it may just shock the world. Every Analyst in attendance at the company’s (first ever) Analyst Day got to experience, first hand, the future today. Congratulations to Fermi Wang (王奉民), Alberto Broggi, VisLab srl and the folks at Ambarella.” Watch Video Hmmmm… Very impressive. Come meet Alberto and learn more about EVA at the 2nd Annual Princeton SmartDrivingCar Summit. Alain
Autopilots
P. Perrone, “Fueled by a $2M prize, inventor Paul Perrone and his small team set out on a road trip across country. They brought with them their quirky robot car Tommy to compete head to head against well-funded giants in the DARPA Grand Challenge, a historic race of robotic vehicles across the Mojave Desert. Tragedies and triumphs occur during the tournament which set the stage for self-driving cars.” Watch Video Hmmmm… Paul, Thank you for putting this together. I never had more fun! Alain
Tesla puts blame on driver in fatal autonomous car crash
T. Lee, Apr 12 “Tesla on Tuesday escalated its media battle with the family of Apple engineer Walter Huang. Huang died in Silicon Valley last month when his Model X vehicle crashed into a concrete lane divider at high speed. Tesla’s Autopilot driver assistance system was engaged at the time. Tesla made its clearest statement yet that Huang—not Tesla—bore responsibility for his death on a Mountain View freeway.
Huang’s family has hired an attorney to sue Tesla. In an on-camera interview with local television station ABC 7, Huang’s wife, Sevonne, said that prior to his death, Huang had complained to her that the car had a tendency to drive toward the exact traffic barrier that ultimately killed him.
But in a statement to ABC 7 on Tuesday evening, Tesla turned this argument around. “We are very sorry for the family’s loss,” Tesla wrote. “According to the family, Huang was well aware that Autopilot was not perfect and, specifically, he told them it was not reliable in that exact location, yet he nonetheless engaged Autopilot at that location.”
Tesla didn’t stop there. “The crash happened on a clear day with several hundred feet of visibility ahead, which means that the only way for this accident to have occurred is if Mr. Huang was not paying attention to the road, despite the car providing multiple warnings to do so.”….” Read more Hmmmm… There are other ways for the crash to happen that have him “paying attention” and AutoPilot may not be “at fault”. What about the Automated Emergency Braking (AEB) system??? “Several hundred feet of visibility ahead” with the car is going straight into the butt-end of a NJ Barrier and no application of the AEB. Really?? Oh, I know… AEB is another misnomer. It really is “made to not rear-end a car ahead but explicitly disregards and not do anything to avoid smashing full speed into brick walls (or parked firetrucks, or trucks that straddle lanes ahead, or but-ends of NJ Barriers, or …) without doing anything” system. Sorry, I was doing a literal interpretation of the system’s name. Alain
Tesla, Safety Agency Feud Over Fatal-Crash Probe
M. Spencer, Apr 12, “An unusual public feud between Tesla Inc. and federal accident investigators escalated Thursday over the examination of a fatal car crash, sidelining the auto maker from an official probe concerning its semiautonomous driving system….The NTSB said Thursday that it tossed Tesla from the probe, asserting the auto maker violated a formal agreement when it released detailed information this week about the crash before government investigators had vetted it.
Tesla late Wednesday, and again on Thursday, contended it dropped out, saying restrictions on disclosures could jeopardize public safety by blocking the timely release of relevant information to the public including about its semiautonomous driving system, called Autopilot….
Despite a sterling world-wide reputation for dissecting aviation disasters, the NTSB lacks extensive experience looking into the complexities of autonomous systems controlling passenger vehicles…”
Read more Hmmmm… This is VERY interesting. Is it really possible that a private entity is more transparent about the safety of its products than the public-sector regulator. In this case it may indeed be possible. I tend to agree with Tesla that NTSB lacks experience (and resources) here. I remain shocked with NTSB’s investigation of the Joshua Brown Florida Tesla crash. The NTSB didn’t report that it investigated the operation of Tesla’s Automated Emergency Braking (AEB) system, didn’t perform “driver’s view animations” of various pre-crash trajectories to try to understand what the truck driver may have seen and the role that the truck’s right-side roof pillar might have played in the crash and didn’t report that it studied various turn-running trajectories that the truck might have taken. In the end, it seems to me that the NTSB failed to release any data that might have helped Tesla or any other manufacture to learn anything from that crash other than to blame Joshua for not being alert when they didn’t even conduct any “drivers view animations”. Very disappointing. I hope that in these cases, the NTSB will look into the AEB’s performance as will Tesla, if they haven’t already done it. And, if they wish, do an “over-the-air fix” ASAP. Anyway, stay tuned. Alain
Automated Vehicles and the Insurance Industry: A Pathway to Safety: The Case for Collaboration
Winter 2018, “…This document reflects the research, deliberations, and current thinking of the Automated Vehicles Task Force (AVTF) of the Casualty Actuarial Society (CAS). Our overarching theme and belief is that actuaries, and the insurance and risk management industries, have a critical role to play in responsibly and cost-effectively bringing new technologies such as this to market. We also believe that it is imperative for the various parties and stakeholders — manufacturers, technologists, policymakers, attorneys, risk managers, insurers, and actuaries — to cooperate during the development and rollout of AV technology, and address collaboratively issues such as defining and collecting appropriate data, considering potential liability systems for an AV world, and establishing appropriate performance benchmarks by which to evaluate AV technologies….
More generally, the CAS AVTF has taken a somewhat different approach to AV research than other organizations. Our motivation is action-oriented: we are focused on what should be done to ensure the technology is brought to market as safely and efficiently as possible….We have purposely restricted our studies to areas in which casualty actuaries have expertise: the implementation and evaluation of insurance pricing models, the quantification of liability costs, and the analysis of risks….” Read more Hmmmm… This is well worth reading however, there may be some troubling conclusions. For example: “Based on the model we used and on certain assumptions which are meant to be illustrative, a vehicle that reduces losses by 50 percent will only receive an 8 percent discount after four years.“. I think that I understand ” uncertainty and risk”, but that customer is making a hefty contribution to the insurer’s bottom line. In a competitive environment that consumer is going to find a better deal from another insurer.
There are a lot of assumptions and challenges in computing “expected loss”. It is these assumptions and challenges that need to be addressed so that insurance can offer discounts that are more inline earlier with reduced loss so that the consuming public will be more incentivized to adopt these technologies that reduce loss, save lives, avert injuries and reduce suffering. Helping consumers more rapidly acquire these safety technologies is in everyone’s best interests. Alain
China just made it easier for self-driving tests to take place on any road in the country
E. Huang, Apr 12, “…China’s Ministry of Transport, along with other authorities, has said that from next month (link in Chinese) all provincial and city governments will have the authority to allow road-testing of autonomous vehicles in their areas. ….The notification says that companies applying for public road tests should have already tested the cars in closed roads and areas, and that a driver with at least three years of experience should always be in the driving seat and ready to take over if necessary. Firms have to record and store driving data for at least three years, and there’s also a time limit on the road test period of 18 months….” Read more Hmmmm… Since this regulation is just for Self-driving (vehicles that require “adult supervision” in the form of a safety driver) and not Driverless (not even a driver’s seat in the car) China is still way behind California and Arizona. But as we look in the rear view mirror, China is gaining fast. Alain
3 Million Uber Drivers Are About to Get a New Boss
A. Madrigal, Apr 10, “Every day, the world’s 3 million Uber drivers spend 8.5 million hours logged into the ride-hailing company’s app. That’s roughly 1,000 years of Uber driving packed into any given 24 hours.
Because of this tremendous scale, Uber is the most important test case for the gig economy, the new economic arrangement where contract workers are arranged into a cohesive labor force by software. There are many companies that share Uber’s controversial approach to doling out work, but none has amassed 3 million people who use the service to try to make money. Never before has an app’s design been so important to so many people.
The Uber app is the drivers’ workplace, as much as the city where they’re driving is. Each decision about its interface structures drivers’ interactions with Uber the company as well as Uber the transportation marketplace. And Uber is now putting the finishing touches on a from-scratch rebuild of the driver app….” Read more Hmmmm… This is a REALLY INTERESTING article. The Uber driver app may be Uber’s most valuable asset/IP because of its ability to manage a very large number of gig workers (herd an enormous number of cats). This is a non-trivial accomplishment. Granted that Driverless cars cap the value of this asset in the mobility space, it also has substantial value in other markets such as home assistance/care, … Alain
Lawyers for Uber Victim’s Family Say Issue Is ‘Resolved’
A. Prang, Mar 29, “A law firm hired by the family of an Arizona woman who died after being struck by a self-driving Uber Technologies Inc. vehicle said Thursday the issue “has been resolved.”…” Read more Hmmmm… Now Uber must divulge exactly what went wrong with their system so that this tragedy isn’t repeated. Alain
Manufacturers make progress on voluntary commitment to include automatic emergency braking on all new vehicles
Press release, Dec 21, 2017 “Four of 20 automakers report that automatic emergency braking (AEB) is standard on more than half of their 2017 model year vehicles, the National Highway Traffic Safety Administration (NHTSA) and the Insurance Institute for Highway Safety (IIHS) announced in the first update of manufacturer progress toward equipping every new vehicle with the crash avoidance technology. Even without making it standard, another five automakers report that more than 30 percent of vehicles they produced in 2017 were equipped with AEB.
…”AEB systems halve rear-end crashes, while FCW alone reduces them by nearly a third, IIHS research indicates.” Read more Hmmmm… This is about rear-end crashes, which is great!! However, it is NOT about crashes with fixed objects (“brick walls”) There is no mention about that, where the crash prevention rate may well be ZERO!!! (don’t/are supposed to work) Why… because you may be able to pass under the fixed object ahead that is detected. Say it is an overpass, or a sign or a branch. Can’t be applying brakes in situations where the detected object has a longitudinal velocity close to zero. So, disregard all of those. However, brick walls, parked firetrucks, ends of NJ Barriers, trucks stopped broadside in front of you each have zero longitudinal velocity. Woops, just created a false negative. Yipes!!!
NHTSA should put out a warning that “AEBs are only designed to keep you from crashing with moving vehicles ahead of you. They will NOT help you from crashing into fixed objects”. Alain
Aurora CEO Chris Urmson says self-driving tech too important not to succeed
A. Aupperlee, Apr 10, “A pioneer of self-driving cars who led ground-breaking work at Carnegie Mellon University and Google on Tuesday urged Pennsylvania’s top transportation official to take a cautious approach to regulating the evolving industry. Chris Urmson, CEO and co-founder of Aurora Innovation, warned that overly strict rules on self-driving cars could stifle innovation.
“Despite a lot of the headlines, this is very early,” Urmson said during a discussion with PennDOT Secretary Leslie Richards. “We want to be careful that the regulation is not overly prescriptive on how you solve the problem because it will lead to unintended consequences, and it will cause us to almost certainly to end up in some kind of dead end.” ..” Read more Hmmmm… Nice article. Alain
Three problems with Facebook’s plan to kill hate speech using AI
W. Knight, Apr 12, “Mark Zuckerberg told the US Congress this week that Facebook will increasingly rely on artificial intelligence to catch hate speech spread on the platform. “I am optimistic that over a five-to-10-year period we will have AI tools that can get into some of the linguistic nuances of different types of content to be more accurate,” …But it won’t be easy, for three reasons…” Read more Hmmmm… Not easy… That’s an understatement. Building AI to drive a car is trivial in comparison. One can assemble a wealth of pristine training data and the environment that driverless cars operate is made by humans, so it is really well structured and simple. Speech, by definition is vague with a vast grey area. Thus there is no pristine training data so AI will at best be “maybe”. (Please don’t suggest that Mark’s AI is going to define its own “hate speech”. That is left for Real Intelligence (RI) to define.). It is really very disingenuous for Mark to suggest that AI will clean up Facebook. While the platform does provide some good, it is too easily exploited by those who wish to inflict harm. What’s even worse is that it is in this weird dynamic that as it tries to deliver even more good, it becomes easier to exploit to inflict even more harm. Chat Roulette was a really good idea that was immediately commandeered for undesirable purposes that killed it (I hope. I refuse to even find out). Such may be Facebook’s fate and AI won’t save it. (although if anyone can, Yan LeCun can.) Alain
Safety Performance Management (Safety PM)
April 2018, “Here you will find information on the Highway Safety Improvement Program’s (HSIP) safety performance management measures and State safety performance targets. The Federal Highway Administration (FHWA) published the Safety PM Final Rule in the Federal Register on March 15, 2016, with an effective date of April 14, 2016.” …” Read more Hmmmm… Look at bottom of page for interactive map that allows you to click on any state and retrieve its safety data. Even more reason why US DoT should be focused on Safe-driving Cars (Trucks & Buses). Alain
TomTom Bets It’s Found a Way to Deal With Slow Death of Sat Nav
W. Vergauwen, Apr 8, “Self-driving cars may not be on the road anytime soon, but that isn’t stopping Dutch navigation device maker TomTom NV from investing in high-definition maps needed when autonomous driving becomes the norm. … Goddijn, who has been leading the company since 2001, may not have much choice. TomTom, which has years of experience creating static maps — the kind embedded in traditional car navigation screens — is contending with the terminal decline of what used to be its core business — personal navigation devices. Also, it’s vying with companies such as HERE and Alphabet Inc.’s Google in the race to dominate the “dynamic,” or HD maps, market….” Read more Hmmmm… No mention of Carmera? Be careful, look over your shoulder. Sure am glad ALK didn’t get into the “device” business. Alain
Half-baked stuff that probably doesn’t deserve your time
As e-biking grows, U.S. cities consider easing rules on where the machines may be used
L. Lazo, Apr 7, “A surge in the use of electric bicycles throughout the United States is prompting cities to revise regulations that restrict their use, including bans against riding them on sidewalks and trails. …” Read more Hmmmm… Sorry… eBikes & pedestrians don’t mix. Create eBike/skateboard/Segways/solWheel/TotalJerk lanes but sidewalks are for pedestrians!!! Alain
Microsoft Garage Project Road Runner trains self-driving cars
JC Torres, Apr 10, “…This is part of the deep reinforcement learning method that Sharma and his teammates advocate. In other words, Project Road Runner gives AI a completely safe environment to learn how to drive safely by crashing repeatedly…” Read more Hmmmm…My opinion: deep reinforcement learning is absolutely the wrong way to approach this AI problem. Happy that the article states: “…As a skunkworks project, Road Runner is unlikely to become a commercialized product…” Msft. don’t jump in. Alain
C’mon Man! (These folks didn’t get/read the memo)
Uber CEO Says Self-Driving Cars Are ‘Student Drivers’
C. Lombardo, Apr 12, “Uber Technologies Inc. Chief Executive Dara Khosrowshahi said the ride-hailing firm is “absolutely committed” to its self-driving-car program …, but said the technology needs to be refined. “Ultimately, self-driving cars will be safer than humans,” Mr. Khosrowshahi said on NBC’s “Today” show Thursday. “But right now self-driving cars are learning. They’re student drivers.”…Mr. Khosrowshahi said the company is adding a 911 button to its ride-hailing application that will directly connect riders to emergency operators and share their locations…” Read more Hmmmm… Did Dara just insult “student drivers”? Listen to interview. Also, doesn’t every smartPhone already have a 911 button (and every Uber rider has a smartPhone). Doesn’t every 911 app also hit our GPS receiver and transmits your lat/long? If not then, it and Uber should simply log into Facebook, because it knows. We’re not being tossed much of a bone here. We need from Uber real transparency on what went wrong here and we can’t expect the NTSB to throw us much of a bone, either. C’mon Dara! Alain
Calendar of Upcoming Events:
2nd Annual Princeton SmartDrivingCar Summit
May 16 & 17, 2018
Princeton University
Princeton, NJ
Application NOW OPEN
Early Bird Rates End April 15
Become a Sponsor and Promote your Wares
On the More Technical Side
https://orfe.princeton.edu/~alaink/SmartDrivingCars/Papers/
Capsule Networks (CapsNets) – Tutorial