Autonomous Vehicles

The second VDI Autonomous Truck conference in Dusseldorf took place last week and much of the discussion was about the pace of developments towards Level 4 and Level 5 automation and the future role of the many existing and new players.

https://www.vdi-wissensforum.de/en/event/autonomous-trucks/

Many of the market players at the conference believe that by 2030, we will see widespread use of Level 5 autonomous trucks (ie, no driver, no steering wheel, no cab). Meanwhile Level 4 autonomous trucks are expected to be widespread in the 2020-2025 time period. They expect for Level 4 AV trucks to be applied first with closed loop systems (ports), then highway-only haulage and finally complex, multi-use urban applications. Some participants believe we might bypass Level 3 AV trucks altogether (ie, platooning with a driver in every truck in the platoon), given the pace of Level 4 autonomous developments and the higher benefits. Most participants believed start-ups will emerge quicker with solutions.

For example, a Swedish start-up, Einride, plans to deliver its L5 driverless delivery pod to commercial customers later this year, with broad rollout by 2020.

Interesting times in autonomous trucking too.
 
sesevans said:
One very clear observation. In November 2017, Alphabet's Waymo achieved the distinction of requiring just one disengagement on a California Street. That is a new low for the year which in my view shows considerable improvement.
Click to expand...
On the reading you have done you have not discovered the break-out of what fraction of the waymo test miles have been on streets then ?

The Tesla AP and cutting off link with NTSB maybe have raised an issue with respect to licensing of thw waymo driverless test too - law of unintended consequences
https://www.caranddriver.com/news/t...h-investigations-involving-autonomous-systems

unlike standadised black boxes in aeroplanes, seems the NTSB can do little without manufacturer help on incident debug - do we trust them ?

Tesla do not seem to have refined their narrative about the crash, but if they are being sued I guess that explains it,
if their data suggesting driver had not touched the steering wheel is false, per earlier link, then all bets are off.
 
jpaul said:
On the reading you have done you have not discovered the break-out of what fraction of the waymo test miles have been on streets then ?

The Tesla AP and cutting off link with NTSB maybe have raised an issue with respect to licensing of thw waymo driverless test too - law of unintended consequences
https://www.caranddriver.com/news/t...h-investigations-involving-autonomous-systems

unlike standadised black boxes in aeroplanes, seems the NTSB can do little without manufacturer help on incident debug - do we trust them ?

Tesla do not seem to have refined their narrative about the crash, but if they are being sued I guess that explains it,
if their data suggesting driver had not touched the steering wheel is false, per earlier link, then all bets are off.
Click to expand...

A few really interesting observations in your note.

It is true we do not have the complete picture of what "fraction of Waymo test miles have been on Streets", but California DMV does and they are currently reviewing Waymo's application to test drive AVs on their Streets and highways without any human safety driver in the vehicle. Waymo filed the application last Thursday and an initial indication of whether there is sufficient info in their filing for a thorough DMV review must be given by this weekend (ie 10 days). If California says there is sufficient info, then there is an indefinite period for their review and approval/denial. We should follow this process closely as it will, in effect, tell you what you need to know about the level of confidence the California DMV has in Waymo's testing.

I note your interesting comments about the Tesla/NTSB spat and have read your link closely. In some respects, although Tesla markets their product as AutoPilot, I believe the DMV will see a clear distinction between Tesla's L2 testing of its AP is and what Waymo's L4/L5 testing has been. Also keep in mind that Tesla's L2 vehicles drive everywhere while Waymo's L4/5 vehicles "drive" in geo-fenced, carefully planned and structured settings. In their application last week to California, it is important to keep in mind two things: Waymo has maintained a secret AV test track in Castle, California, information about which was only released to the public last year. Also Waymo's application initially covers Mountain View and will slowly expand to a couple of nearby towns once they too are fully geo-fenced, initially with human safety drivers in control of the mapping.

Simply put, one gets the feeling that Waymo's approach is careful, structured, more likely to be in a collaborative spirit with DMV than the seeming seat-of-the-pants, confrontational, disruptive approach taken with the DMV by Tesla. I do not know whether one approach is ultimately going to be more successful than the other but if I am correct in my assumption, I would be more inclined to think that Waymo's approach will prove more successful, esp in obtaining regulatory harmony. And if true, DMV might be more inclined to give Waymo a favourable runway vs its competitors.
 
from your #887 article - seemed a bit of a minor distinction between the waymo and tesla car platforms - tesla can put in the redundant systems it suggests are waymo's advantage, and invest in more computing power, but waymo are not designing the car platform and even though tesla may have production/scaling problems for their new model - they do have a production line, and can modify body to accept any sensors, unlike google with converted prius/other. .... maybe i missed the point.

I don't know how many miles tesla AP's have 'driven' but I would have though with all the ownership Tesla potentially has more data, for lane/obstacle identification 'database'

Saw another video of tesla AP failure courtesy of reddit
at 20mph the system did not see the vans lane change because of the rain ! if it cannot save me in this situation

liked this reddit comment to about airline systems

https://www.reddit.com/r/SelfDrivingCars/comments/8bojp3/can_teslas_even_see_lane_dividers/
Yeah, but human brains are not equipped for maintaining this level of awareness without stimulation that requires it. Regular cruise control still requires you to be mentally engaged to steer.
This is why the TSA software has to periodically put in false positive images so that the agents don't zone out.
Click to expand...
so need a hud display with a few obstacles inserted to check driver awareness
 
jpaul said:
"but waymo are not designing the car platform and even though tesla may have production/scaling problems for their new model - they do have a production line, and can modify body to accept any sensors, unlike google with converted prius/other."
Click to expand...

This was true historically for Waymo with Toyota and to a lesser extent with Fiat Chrysler. With Toyota Prius, Waymo adapted an existing vehicle while with Fiat Chrysler I believe Waymo engineers were embedded into FC's Ontario plant during the production stage of the Chrysler Pacifica SUV. I also assume that their partnership with Lyft would also involve elements of the Prius (adapting existing vehicles) and Chrysler Pacifica model.

However with the recently announced deal with Jaguar Land Rover for the I-Pace EV, Waymo and JLR are designing the EV together from the ground up at JLR's Austrian (sub-contractors) plant. Presumably this would gain the efficiencies enjoyed by Tesla that you refer to in your post.

As I posted in #885, I believe comparing Tesla and Waymo today is apples and oranges because Tesla in a L2 solution currently whereas Waymo is a L4/L5 solution. Your video of the Tesla accident with the armoured vehicles confirms this in my mind.

You may recall that Waymo has said on multiple occasions that it does not have plans currently or in future to manufacture self driving vehicles. It seeks to partner with existing OEMs and provide the "driver". The partnership with JLR seems a good case in point. By 2020, Waymo's CEO has said that they plan to provide 1 million ride sharing journeys per day with their I-Pace EV fleet. They are rolling our their commercial ride hailing taxi service in Arizona this year and their intention is to increase their service city by city as they geo-fence each new expansion territory. Waymo will own a fleet of 20,000 I-Pace EVs at a total cost in excess of $ 1 billion.

I assume that Waymo does not intend to spend billions of dollars buying vehicles but would prefer to insert their "driver" solution principally onto existing vehicles or at worst make the adjustment at factory level as painless as possible for the OEM partner. I would expect that a partnership with Lyft and Avis (also announced) might be optimised if the Waymo "driver" could be easily inserted just as the vehicle left the factory or at an after-sales service centre--ie, expanding their AutoNation partnership to provide such after sales service on existing vehicle fleets.

Time will tell.
 
While Waymo develops their commercial activities, there are still many hurdles to overcome to make their service nationwide. Each US State can regulate its own roads even though the US Department of Transportation has ruled that States cannot ban driverless vehicles outright. This mish-mash runs the risk of a patchwork quilt of regulations limiting industry growth.

An example of this patchwork can be seen with New Hampshire:

http://www.ttnews.com/articles/new-hampshire-self-driving-vehicle-legislation-faces-speed-bumps
 
"6 surprising ways driverless cars will change our world"/Mach

Had not considered organ donations from people currently killed in car crashes expected to drop signicantly!

And business will come to you: "Imagine ordering up a gym that drives to you and parks in your driveway for an hour or two before moving on to the next customer. Gerdes thinks mobile gyms are a possibility, along with mobile clothing stores stocked with apparel and equipped with dressing rooms. “You could try on a bunch of things you had requested, see what size works for you, and then the rest of it simply leaves at the end of your session.”

https://www.nbcnews.com/mach/scienc...verless-cars-will-change-our-world-ncna867061
 
jpaul said:
from your #887 article - seemed a bit of a minor distinction between the waymo and tesla car platforms - tesla can put in the redundant systems it suggests are waymo's advantage, and invest in more computing power, but waymo are not designing the car platform and even though tesla may have production/scaling problems for their new model - they do have a production line, and can modify body to accept any sensors, unlike google with converted prius/other. .... maybe i missed the point.

I don't know how many miles tesla AP's have 'driven' but I would have though with all the ownership Tesla potentially has more data, for lane/obstacle identification 'database'

Saw another video of tesla AP failure courtesy of reddit
at 20mph the system did not see the vans lane change because of the rain ! if it cannot save me in this situation

liked this reddit comment to about airline systems

https://www.reddit.com/r/SelfDrivingCars/comments/8bojp3/can_teslas_even_see_lane_dividers/

so need a hud display with a few obstacles inserted to check driver awareness
Click to expand...

I saw this link in Verge today that you might find interesting in comparing approaches of Waymo and Tesla.

https://www.theverge.com/transporta.../tesla-waymo-self-driving-car-data-simulation

One of the comments however did note that the work being done by Mobileye was not discussed.

Helpful?
 
jpaul said:
from your #887 article - seemed a bit of a minor distinction between the waymo and tesla car platforms - tesla can put in the redundant systems it suggests are waymo's advantage, and invest in more computing power, but waymo are not designing the car platform and even though tesla may have production/scaling problems for their new model - they do have a production line, and can modify body to accept any sensors, unlike google with converted prius/other. .... maybe i missed the point.

I don't know how many miles tesla AP's have 'driven' but I would have though with all the ownership Tesla potentially has more data, for lane/obstacle identification 'database'

Saw another video of tesla AP failure courtesy of reddit
at 20mph the system did not see the vans lane change because of the rain ! if it cannot save me in this situation

liked this reddit comment to about airline systems

https://www.reddit.com/r/SelfDrivingCars/comments/8bojp3/can_teslas_even_see_lane_dividers/

so need a hud display with a few obstacles inserted to check driver awareness
Click to expand...

It's a false sense of security. The driver won't notice it's not going to stop until it's to late, there's a point of no return where it will take to long to react and take control back. So on a road with traffic it becomes safer to just drive the car yourself.
 
39760007120_cd61f99eda_o_d.jpg
 
sesevans said:
I saw this link in Verge today that you might find interesting in comparing approaches of Waymo and Tesla.

https://www.theverge.com/transporta.../tesla-waymo-self-driving-car-data-simulation

One of the comments however did note that the work being done by Mobileye was not discussed.

Helpful?
Click to expand...

This opinion piece in Forbes comments on the Verge article comparing the approaches of Waymo and Tesla but makes the point well: it is really about "the right" data strategy. This will ultimately separate winners from losers in the space.

The authors point about both Tesla and Waymo: "Even before it leaves the factory, an autonomous vehicle has dealt with more road situations and problems than the best human drivers will have and will be able to solve them expertly and efficiently."

https://www.forbes.com/sites/enriqu...a-strategy-makes-the-difference/#3db77598256a
 
LOL: yes looks as though the Muskovites (do they have a name) pay for their Tesla's and accept to upload
Tesla declined to comment on what data is being collected from which sensors, or the quality of that data. It could be all the video from the car, from just some of the cameras at certain moments (like crashes), or data from the ultrasonic sensors without video. And, Rajkumar says, and it’s also unclear whether it’s the full frame-rate video or something with less fidelity.
Click to expand...

so seems like a facebook scenario/harvest, but even more luctrative
seems
That's what OP is talking about. Tesla vehicles can connect to a home wifi, but even home Internet plans are coming with caps these days. If the car is uploading 1-2GB per day, that's a significant 5% per month of a 1TB cap, even higher on a 300GB cap which many ISPs have.
AFAIK, the LTE connection is free for the life of the vehicle, but someone with Model S should confirm that.
Click to expand...

so they know where you have been, could be watching your driving style ...maybe even does a bit of face recognition for the FBI on the side.
 
jpaul said:
LOL: yes looks as though the Muskovites (do they have a name) pay for their Tesla's and accept to upload


so seems like a facebook scenario/harvest, but even more luctrative
seems


so they know where you have been, could be watching your driving style ...maybe even does a bit of face recognition for the FBI on the side.
Click to expand...

Yes, data, data, and more data. And how it is being harvested and will the driver do the same trade-off we all make on the internet generally.....a "free" service for you if you provide your data. But I think it is early to really get to exercised about this yet as driverless vehicles are still only really in trial stage and not commercially employed yet.
Will the West's fascination for limiting your transfer of data limit the ability to compete with Chinese AV companies that are under no similar constraint?

http://www.scmp.com/magazines/post-...hinas-self-driving-vehicles-track-take-global

Are we shooting ourselves in the feet while China establishes themselves as global leaders?

Meanwhile interesting piece in Wired about what the author believes the AV companies should do, apart from offering a commercial ride sharing service, to help "sell" the service to a sceptical audience.

1. A free service for young people (read drinkers) between the hours of 10 pm and 2 am---to keep children safe.

2. A short "last mile" service to help people get from the train station to home.

https://www.wired.com/story/uber-self-driving-crash-strategy/
 
There is a race for tech supremacy between the US and China and recent tariff talk also sanctions involving steps to prevent China stealing intellectual property has heightened our awareness of this issue.

In this environment, it is interesting to note that there are still situations where such cross border investments in technology still appear to make sense. See this one in cargo transport, suggesting a future for Google/Waymo in western markets:

"
Japan’s SoftBank Group Corp. and Alphabet Inc.’s late-stage venture investment fund are among investors plowing close to $2 billion into a Chinese truck-hailing company, according to people familiar with the investments.

Manbang Group, which runs a mobile app platform that matches truck drivers with shippers looking to transport cargo, is raising more money than it originally set out to fetch, according to these people. The Wall Street Journal previously reported the company was planning to raise between $500 million and $1 billion.

Manbang’s previous backers have included internet giant Tencent Holdings Ltd. and a private-equity firm co-founded by Chinese billionaire Jack Ma.

CapitalG describes itself as a growth-equity investment fund whose goal is to "make return-driven investments in leading companies around the world,” according to its website. A person familiar with the fund’s strategy said it typically invests for profit and not strategically for Google parent Alphabet.

Most of Google’s services have been unavailable in China since 2010, when the U.S. technology company withdrew from the market due to censorship concerns.

Since then, the company has been trying to make small steps to re-enter the market. It launched an artificial intelligence lab in China last December, and Google Chief Executive Sundar Pichai has visited the country at least twice in the past six months to speak at Chinese forums."
 
You must log in or register to reply here.
Back
Top Bottom