7.46-Maybe-110919

Saturday, November 9, 2019

https://www.princetondiary.com/smartdrivingcar/7.46-Maybe-110919

Hailing a Driversless ride in a Waymo

E. Niedemeyer, Nov 1, "Congrats! This car is all yours, with no one up front,” the pop-up notification from the Waymo app reads. “This ride will be different. With no one else in the car, Waymo  will do all the driving. Enjoy this free ride on us!”  Moments later, an empty Chrysler Pacifica minivan appears and navigates its way to my location near a park in Chandler, the Phoenix  suburb where Waymo has been testing its autonomous vehicles since 2016.

Waymo, the Google  self-driving-project-turned-Alphabet unit, has given demos of its autonomous vehicles before. More than a dozen journalists experienced driverless rides in 2017 on a closed course at Waymo’s testing facility in Castle … Doesn’t count!… ; and Steve Mahan, who is legally blind, took a driverless ride in the company’s Firefly prototype on Austin’s city streets way back in 2015 … Counts, but just as a ‘one < 10 mile ride’, (We’ll assume that the road conditions weren’t staged.)…  .

But this driverless ride is different — and not just because it involved an unprotected left-hand turn, busy city streets or that the Waymo One app was used to hail the ride. It marks the beginning of a driverless ride-hailing service that is now being used by members of its early rider program and eventually the public…"  Read more  Hmmmm…  Maybe!?   If it is, it is reasonably well hidden that my student wasn’t able to find one that was being used in this way to try to verify that this is not just another staged event.  If this is really going to be an economically sustainable mobility service it is going to have to be able to deliver "millions" of rides without a driver, without an attendant and without staging. It would be nice to be able to just find one.  FYI: In Chandler on a typical day there are approximately a million person trips taken using a vehicle.  Across the nation there are about a billion.  See also Tim Lee’s reporting of this trip.  Alain

  Smart Driving Cars Podcast Episode 131

F. Fishkin, Nov. 9, "Hoping for a driverless ride from Waymo? So are Princeton’s Alain Kornhauser and co-host Fred Fishkin. In this edition…Waymo, more trouble at Uber, the latest at Lyft and Tesla…and more! Tune in to Smart Driving Cars and subscribe."  Just say "Alexa, play the Smart Driving Cars podcast!".  Ditto with Siri, and GooglePlay …  Alain

  Thousands of Tesla Owners Share Their Close Calls With Autopilot

T. Randall, Nov 5, "We asked 5,000 Model 3 owners about Tesla’s software for automated driving on highways and parking lots. More than 90% said the feature makes them safer…. 

Six drivers claimed that Autopilot actually contributed to a collision, while nine people in the Bloomberg survey went so far as to credit the system with saving their lives. Hundreds of owners recalled dangerous behaviors, such as phantom braking, veering or failing to stop for a road hazard. But even those who reported shortcomings gave Autopilot high overall ratings….

Autopilot Puts Owners in Danger—And Saves Them From It
Explore owners’ close calls with Autopilot engaged" Read more  Hmmmm…  Very interesting.  Substantive 1st hand accounts that can be taken seriously.  See also Tim Lee’s take on this. Alain

  VEHICLE AUTOMATION REPORT

E. Becic, Nov. 5, "Uber ATG developmental ADS installed on the crash-involved vehicle was designed to operate in a fully autonomous mode only on pre-mapped designated routes. Although the system was designed to be fully automated along a specific route, a human operator located inside the vehicle was tasked with overseeing the operation of the system and monitoring the environment.  Unless stated otherwise, the ADS discussed in this report refers only to Krypton platform, software version 2018.071.3p1 that was installed on the crash-involved vehicle….

1.6 Object Detection and Hazard Avoidance
When ADS is activated, it performs all driving tasks, including changing lanes, overtaking slow moving or stopped vehicles, making turns, or stopping at traffic lights and stop signs.
1.6.1. Object Detection and Classification, and Path Prediction
As the ADS navigates and controls the vehicle along a designated route, the system continually monitors the environment for any objects, whether moving or stationary, on or outside a roadway. The detected objects are incorporated into the virtual environment, and the system dynamically updates the vehicle’s motion plan to avoid potential conflicts. Object detection is conducted primarily by the lidar, radar and camera systems, each of which has different specialized functions. When an object is detected, it is tracked, its heading and velocity calculated, and classified by the perception system. Detected objects can be classified as vehicles, pedestrians and bicyclists; a detected object may also be classified as “other”, indicating an unknown object….

However, if the perception system changes the classification of a detected object, the tracking history of that object is no longer considered when generating new trajectories. …

When the system detects an emergency situation, it initiates action suppression. This is a one-second period … one second is a very long time in an emergency situation…  during which the ADS suppresses planned braking while the (1) system verifies the nature of the detected hazard and calculates an alternative path, or (2) vehicle operator takes control of the vehicle. ATG stated that it implemented action suppression process due to the concerns of the developmental ADS identifying false alarm —detection of a hazardous situation when none exists—causing the vehicle to engage in unnecessary extreme maneuvers.   

Read more   Hmmmm… … false alarms, aka mistakes, are the problem here.  Only with hindsight is a mistake identified.  So waiting any amount of time expecting to correct a mistake has no value.  If a ‘better whatever’ would have not made the mistake, why wasn’t the ‘better whatever’ used in the first place?

Mistakes (false alarms) tend to be transient (short lasting) rather than persistent.  Plus, these systems re-evaluate everything 10 to 30 times a second.  Doing nothing while waiting for the system to stop making mistakes might be a valid design; however, the brakes take time to activate.  By the time a brake force actually begins to be applied, the system may well have stopped making mistakes thus suspending the brake action, maybe even before any discomfort is felt by those inside the car. All good.

If the mistakes are so persistent the brakes are not released and discomfort is felt in the car for no reason, then the system is simply not good enough to be used on public roadways.  Unfortunately it gets worse…

… if the collision cannot be avoided with the application of the maximum allowed braking, the system is designed to provide an auditory warning to the vehicle operator while simultaneously initiating gradual vehicle slowdown. In such circumstance, ADS would not apply the maximum braking to only mitigate the collision.  … Unbelievable!!  Did Uber really say …  crash faster??? Unbelievable!!

… Table 1. Selected parameters recorded by the ADS.  … Please read.  This is really scary. Classification of objects is a very important objective of this algorithm.  Probably because future actions of object is assumed to be well correlated to object type.  Given an object,  its future behavior is better anticipated by an historical pattern assigned to that object rather than what has been observed about its behavior since it was first sighted.  That is why, if the classification changes of an object that is being tracked, its previous position and velocity data are disregarded.  I doubt that this approach has any merit, even if one could reliably classify objects in real time.  Seems like Uber tried to get fancy here but its capabilities were simply not up to the task. It couldn’t reliably classify objects and even if it did, historic behaviors are unlikely to be better predictors of future behavior than simple extrapolations of recent observations. 

Using recent tracking data has to be better than . … ADS predicts the object’s path as static… if the objects classification changes.  That is the big mistake!  In fact, only two classifications are needed: OK to hit versus Do NOT hit.  Track each (position, velocity, …) and use the tracking data to predict each object’s its future trajectory. 

But there is more…
1.9 Interaction Between Volvo ADAS and ATG ADS
When the SUV was operated in a manual mode—controlled by a vehicle operator—all the Volvo ADAS components were active and operated as designed. When the SUV was operated in the autonomous mode—controlled by the ADS—all Volvo ADAS components were automatically disengaged. …

1.10 Crash History of the ATG Fleet of ADS-Equipped Vehicles
ATG shared records of fleet crash history with NTSB investigators. The records showed that between September 2016 and March 2018 (excluding the current crash), there were 37 crashes and incidents involving ATG test vehicles which at the time operated in autonomous mode. Most of these crashes involved another vehicle striking the ATG test vehicle—33 such incidents; 25 of them were rear-end crashes and in 8 crashes ATG test vehicle was side swiped by another vehicle.

In only two incidents, the ATG test vehicles were the striking vehicles. In one incident, the ATG vehicle struck a bent bicycle lane bollard that partially occupied the ATG test vehicle’s lane of travel. In another incident, the vehicle operator took control of the vehicle to avoid a rapidly approaching oncoming vehicle that entered the ATG vehicle’s lane of travel; the vehicle operator steered away and struck a parked car. In the remaining two incidents, an ATG vehicle was damaged by a passing pedestrian while the vehicle was stopped. …To be fair, this is actually a really good crash history.   My interpretation is that the crash probability of the Uber ATG vehicles with trained operators behind the wheel is almost 10 times safer that of the driving public’s overall crash probability.  (I don’t buy the argument that these cars cause the driving public to hit them because of the way the automated system drives.)  Alain

Report on Uber crash questions testing of self-driving cars

T. Krisher, Nov. 7. "  A National Transportation Safety Board report on the March 18, 2018 crash in Tempe, Arizona, that killed Elaine Herzberg, 49, found that the Uber self-driving system couldn’t determine if she was pedestrian, vehicle or bicycle. It also could not predict that she was jaywalking in the path of the moving SUV.

“The system design did not include a consideration for jaywalking pedestrians,” the agency said in its report, released ahead of a Nov. 19 board meeting to determine the cause of the Tempe crash.That, and the fact that Uber disconnected braking systems and relied on the human safety driver to stop the SUV in an emergency, shows that the Uber system wasn’t ready to be tested on public roads, experts say. Some say that stronger standards or more government regulation are needed to set standards before testing is allowed.

“These have to be much better than that before they can go out there (on public roads),” said Alain Kornhauser, chair of autonomous vehicle engineering at Princeton University. “If you can’t do better than that, stay on your test tracks. Don’t come out in public.”

Consumer Reports said the report showed “outrageous safety lapses” by Uber and called for stronger rules governing autonomous vehicle testing.  “We hope Uber has cleaned up its act, but without mandatory standards for self-driving cars, there will always be companies out there that skimp on safety,” Ethan Douglas, senior policy analyst for the magazine and website, said in a statement. “We need smart, strong safety rules in place for self-driving cars to reach their life-saving potential.”…"  Read more  Hmmmm… See above.  See also Tim Lee’s take on this.  Alain

  Lyft is adding Chrysler Pacificas to its AV fleet and opening a new dedicated self-driving test facility

D. Etherington, Nov. 8,  "Lyft  has another year of building out its autonomous driving program under its belt, and the ride-hailing company has been expanding its testing steadily throughout 2019. The company says that it’s now driving four times more miles on a quarterly basis than it was just six months ago, and has roughly 400 people worldwide dedicated to autonomous vehicle technology development.

Going into next year, it’s also expanding the program by adding a new type of self-driving test car to its fleet: Chrysler’s  Pacifica hybrid minivan, which is also the platform of choice for Waymo’s  current generation of self-driving car. The Pacifica makes a lot of sense as a ridesharing vehicle, as it’s a perfect passenger car with easy access via the big sliding door and plenty of creature comforts inside…"  Read more  Hmmmm… Sounds great! Alain

  Uber faces costly choices after expert finds it uses Waymo self-driving tech

P. Dave, Nov 7, " Uber (UBER.N) said it “will likely” have to strike a licensing deal with Waymo or opt for costly changes to its autonomous driving software, after an expert found the ride-hailing giant still used technology from the Alphabet Inc (GOOGL.O) unit.  While it was unclear by when the company needed to decide on its next move in the blockbuster trade secrets dispute, Uber, in a quarterly securities filing on Tuesday, said that a detour in its software development “could limit or delay our production of autonomous vehicle technologies.”…

Waymo told Reuters in a statement that the independent software expert’s findings “further confirm Waymo’s allegations that Uber misappropriated our software intellectual property. We will continue to take the necessary steps to ensure our confidential information is not being used by Uber.”…" Read more  Hmmmm…  Really??? That’s all Uber had to do to get Waymo to give it a licensing deal on its technology?   Doesn’t seem that Waymo would want to do that unless the licensing deal was at least as valuable as the service that it enables.  The price for a goose that lays golden eggs is the price of the goose plus the net present value of the eggs that it produces ad infinitum.  Ironically, such pricing wouldn’t help Uber reach profitability nor its ability to compete with Waymo One.   See also Tim Lee’s reporting of this. Alain

  Alphabet’s self-driving car project Waymo is shuttering its Austin operations

W. Feuer, Nov.8, "…The Alphabet subsidiary is removing employees at the facility and says it will offer them relocation to the company’s production center in Detroit, or to Phoenix, Arizona, where Waymo has a fleet of hundreds of self-driving cars. Those who do not wish to relocate will be offered a transition pay package. The move will affect fewer than 10 employees, but an undisclosed number of contractors also worked in the office. The relocations will not affect other Waymo employees around the country…

“Waymo is growing our investment and teams in both the Detroit and Phoenix areas, and we want to bring our operations teams together in these locations to best support our riders and our ride-hailing service,” a Waymo spokesperson said in a statement sent Friday to CNBC. “As a result we’ve decided to relocate all Austin positions to Detroit and Phoenix…."   Read more  Hmmmm…  I guess that they weren’t welcomed, even thought they’ve tried to play nice for more than a few years.  Waymo should send the team to Princeton where we are working to create a welcoming environment for these mobility machines to affordably serve everyone, especially the mobility marginalized (the physically, mentally and/or financially challenged, the too young or too old to drive and those who prefer not to drive their own car), throughout Mercer County and eventually all of New Jersey.   See the excellent NBC  video that is embedded. Alain  

Uber lost $1.1 billion last quarter, says rides are profitable (sort of)

T. Lee, Aug 28, "Uber lost another $1.1 billion in the third quarter of 2019, the company announced on Monday. This wasn’t a surprise: Uber lost about the same amount in the first quarter of 2019 and lost even more last quarter.

Yet the company argues that things aren’t as bad as that headline figure suggests. To show why, Uber broke its earnings down by business area, distinguishing its core "rides" app from Uber Eats, Uber Freight, and other operations.

Uber says that, if you exclude certain non-operating expenses—mainly interest, depreciation, and stock-based compensation—the "rides" app actually earned a substantial $631 million profit. That’s enough to cover the company’s core operating expenses, the company said. But Uber’s profitability was dragged down by losses in its other businesses—mainly a $316 million loss from Uber Eats.

Of course, interest, depreciation, and stock-based compensation are real costs. So the fact that Uber looks less unprofitable excluding them isn’t going to be particularly reassuring to Uber investors."  Read more  Hmmmm…Also read the comments.  Uber stock price is not pretty. Neither is Lyft’s.  Alain

Uber stock hits new low as post-IPO lockup expires

H. Somerfield, Nov. 6, "Uber Technologies Inc. shares hit an all-time low Wednesday as the "lockup" period following its May initial public offering ended, delivering a blow to a company that has struggled to satisfy investors.

The expiration of the period sent a flood of shares onto the market, pushing the stock as low as $25.58, down 43% from its IPO price. While stock volatility and losses aren’t unusual following lockup expirations, the loss of $2 billion from its market capitalization in 24 hours added injury to Uber after a lackluster earnings report this week… " Read more  Hmmmm…Yup!  Alain

Just How Far Ahead Is Tesla In The Self-Driving Race?

Staff, Nov. 8, "… Tesla’s total autonomous miles logged has grown exponentially from 0.1 billion in May 2016 to an estimated  1.88 billion as of October 2019…."  Read more  Hmmmm… Unfair to compare Tesla’s Self-driving mileage with that of Waymo & GM/Cruise.  The functionality of Tesla’s system is very much inferior to that of Waymo &  GM/Cruise.   None the less, the AutoPilot miles are impressive. Alain

  Top 10 Myths About Autonomous Vehicles

A. Roy, Nov. 2, "Outlaw racer Alex Roy & roboticist Bryan Salesky sit with Sam Abuelsamid — Navigant Research Analyst, Forbes writer and one of the world’s leading experts on the automotive & technology sectors — to discuss and debunk the Top 10 Myths about Autonomous Vehicles…." Listen Hmmmm… Very interesting.  Alain

  Watch Tesla Autopilot handle a construction zone with traffic cones

F. Lambert, Nov. 7, "A new video showing what Tesla Autopilot sees when navigating a construction zone gives an impressive look at Tesla’s evolving computer vision system.

Earlier this week, we reported on Tesla updating its driving visualization to detect and render traffic cones.  Tesla has been working on integrating construction zone navigation in Autopilot for a long time…"  Read more  Hmmmm…See video.  Alain

  ‘Smart pods’ blaze a trail for autonomous public transport

K. Monks, Nov. 1, "The NEXT Future Transportation module is a far cry from the sleek visions of self-driving cars designed by Tesla or Mercedes.
With an average cruising speed of 20 kilometers per hour, the electric pods are unlikely to set pulses racing.
But perhaps the most crucial distinction is that this self-driving vehicle is passenger ready… "  Read more  Hmmmm…   Good images of many of today’s drivered-driverless shuttles (each has an attendant on board).  Alain

 

Half-Baked

  Why autonomous vehicles will rely on edge computing and not the cloud

G. Anadiotis, Nov 4, "…Autonomy and cloud don’t go well together
To understand why, let’s consider the notion of autonomy. Autonomy is defined as ‘independence or freedom, as of the will or one’s actions’. Can you be autonomous, when relying on someone else’s computer? Not really…."  Read more  Hmmmm…  That’s why Edge doesn’t cut it either.  Autonomy, by definition,  needs its own on-board dedicated computing; else it’s constrained to locations that have Edge, making it not so autonomous.  Alain