Autonomous Vehicles

Road rules for self driving cars

"
Today, before a developer can put its robo-cars on the road, it must gather piles of data: which streets have which speed limits, where the school zones are, how many lanes any stretch of road has, and so on. The standard method of collecting this data is to drive all the streets in question, and use the cars’ sensors to spot and log every road sign and lane marking. That’s neither efficient nor infallible. It can take half a dozen trips up and down any street to get everything you need. Road signs go missing and lane lines fade. And even then, the process relies on the computer’s ability to translate symbols designed for human eyes into its mother tongue of zeros and ones.

INRIX-AV-Road-Rules_city-nodes.jpg

Built off tech that helps people find parking, Inrix's new program is the kind of thing everyone in the self-driving world can get down with.
INRIX
Inrix, which you may know for its rankings of cities by the horrorshowiness of their traffic, believes the Road Rules system can make this process easier by getting cities to provide the data. This solution seems pretty agreeable—any city that welcomes self-driving cars wants those cars to follow the rules—but the problem is that municipal governments, departments of transportation, and other governmental bodies rarely have all the necessary information in one place. There’s never been much of a need for a list of every road sign or school zone, and so there’s never been a good way to make that list.


Until now."

https://www.wired.com/story/inrix-road-rules-self-driving-cars/
 
Waymo's AVs are "driving" 25,000 miles in autonomous mode every day and have now driven 8 million miles on public roads.

Today they announced that their fleet size of 600 Chrysler Pacifica minivans are operating in 25 US cities.

That means that on average, each vehicle is driving 41 miles per day in autonomous mode. Also noteworthy is that Waymo "drove" 1 million miles in a month---7 million was reached in June and 8 million was reached in July, and has doubled its autonomous miles in 8 months. Waymo had driven 5 billion simulated miles equivalent to a 25,000 virtual vehicles driving 24 hours per day every day.

Waymo plans to launch a commercial ride hailing service in Phoenix later this year. In addition, Waymo is working on solutions for logistics (ie trucking), making public transport more accessible and eventually sale of self driving vehicles to individuals.

https://techcrunch.com/2018/07/20/waymos-autonomous-vehicles-are-driving-25000-miles-every-day/
 
Have they given any break out of what type of driving they are doing, this is still good weather driving (well it is summertime too) and also, mostly on urban/freeway as opposed to more accident prone inner city ? (per my earlier comments), how are the additional miles beneficial ?
if we compare the number of miles these cars need versus a human driver becoming competent ;)

I was surpsised, from your quoute that they 'learn' about individual roads from driving them,as opposed toi pre-loaded mapping data with sign/speed limit information.,
presumably this data gets shared between cars, but how much storage is needed per car; maybe the dynamically grab stuff from the cloud.
 
jpaul said:
Have they given any break out of what type of driving they are doing, this is still good weather driving (well it is summertime too) and also, mostly on urban/freeway as opposed to more accident prone inner city ? (per my earlier comments), how are the additional miles beneficial ?
if we compare the number of miles these cars need versus a human driver becoming competent ;)

I was surpsised, from your quoute that they 'learn' about individual roads from driving them,as opposed toi pre-loaded mapping data with sign/speed limit information.,
presumably this data gets shared between cars, but how much storage is needed per car; maybe the dynamically grab stuff from the cloud.
Click to expand...

Yes they have given details of the diverse driving conditions in the 25 cities they are testing in. Perhaps you might find this link to the official Waymo blog site useful to answer your other questions? Check out the FAQs at the end of this link too. Questions?

https://waymo.com/ontheroad/
 
ok so little clarity by waymo; but ok,
they will have been well tested with the usa population before we(luddites) have them in UK, 2021 when AV are allowed on some motorway sections, afaik.
 
jpaul said:
ok so little clarity by waymo; but ok,
they will have been well tested with the usa population before we(luddites) have them in UK, 2021 when AV are allowed on some motorway sections, afaik.
Click to expand...

My feeling is that it will be sooner for us "Luddites". Recall that a few months ago Waymo said it was in discussion with potential European partner(s) to bring the Waymo "driver" to Europe. They said that Waymo was not a well known brand outside the US so it would be more likely that they would partner with a well known brand in Europe. My thought is that Fiat or Jaguar, their US partners are the most likely potential partners. Then again, it would depend on whether they are to pursue a ride hailing service first or perhaps one of the other early options--eg logistics or public transport. Vehicle ownership was not mentioned as an early priority so I would assume it would appear first in fleet ownership.
 
@jpaul

You make a very good point about the fact that "they will be well tested with the US population" before they arrive in the UK.

It strikes me that Waymo is well on its way to creating a strong, safe and industry leading algorithm. If they achieve this, they will be able to achieve a strong differentiation from others competing in this space.

With its fleet of 600 Fiat Chrysler Pacifica minivans, Waymo has been able to add 3 million real world miles in just 5 months. Compare that to the 8 years it took Waymo to drive its first 3 million real world miles. Waymo has been able to increase is real world data gathering 3 times because 222,000 miles driven per month compares with 600,000 miles per month now.

Why is this so significant? Because Waymo has clearly established the industry leading algorithm and it has the ability to gather real world miles which will continue to improve its algorithm, creating a further differentiation gap and the potential scenario of winner take most. There is a direct link between safety and the quality of Waymo's algorithm.

You can probably compare Waymo's algorithm evolution to Google's search volume evolution. The huge search volume at Google allowed it to continue to improve its search product and establish itself as the clear market leader.
 
https://www.theverge.com/transporta.../tesla-waymo-self-driving-car-data-simulation
in contrast tesla data could be more exstensive, and potentially lucrative

Tesla had logged 780 million miles of data, with 100 million of those miles coming while Autopilot was “in at least partial control”
Later that summer, Musk said that Tesla was collecting “just over 3 million miles [of data] per day.” As of last July, though, the total number of fleet miles driven had jumped to 5 billion. As Tesla sells more cars, the amount of data that can be collected increases exponentially.
Click to expand...

The article reminded me, I have not seen one of these "Musk promised" safety briefings on the auto-pilot statistics yet ? too worried about other things ?
 
jpaul said:
in contrast tesla data could be more exstensive, and potentially lucrative



The article reminded me, I have not seen one of these "Musk promised" safety briefings on the auto-pilot statistics yet ? too worried about other things ?
Click to expand...

Yes, I remember seeing that article in The Verge comparing the approaches of Tesla and Waymo. In fact I may have posted it here on the thread then.

You might consider the following: collecting millions of miles of data on a car being driven with off and on human intervention and judging how a "semi-autonomous" (some might argue Level 2 Tesla vehicle) performs is lower quality data compared to Waymo's combo of billions of miles of virtual data from computer simulations that crunch real live autonomous data with millions of combinations and permutations and now 8 million miles of pure autonomous driving on public roads. Even the Verge article discusses the view of the Carnegie Mellon analyst who says that Tesla may be at a disadvantage due to its lack of LIDAR and the skills that Waymo has (leaning on Google's machine learning/AI algorithms and raw compute power in its data centres) in processing and applying and learning from the data.

You also might consider the other inherent advantages Google possesses with Google Cloud Platform, Google Tensor Processing Unit (I appreciate Tesla works with Nvidia), Google Maps, Waze and Google Assistant. Google is a juggernaut when it comes to machine learning and AI.

A few tidbits from Google's conference call last night with analysts, while not directly relevant to this discussion on the comparison of Tesla data collection v Waymo data collection, gives me a better appreciation of Google in house capabilities on what investments it has made and continues to make in compute power, AI and machine learning and how that is of benefit to many of their platforms, include the 7 Google platforms with more than 1 billion daily users (search, maps, YouTube, Google Play, Chrome, etc):

1. Google Assistant is available in 30 languages in 80 countries and connect to more than 5,000 devices (Internet of Things). This is all about delivering data to the user (in the case of the autonomous vehicle, the "driver" I assume) in the most useful way.

2. Google capital expenditure this year is expected to be US $ 21 billion, double its spend from last year. Google has shown that this spend is directly of benefit in driving more and more users to its platforms and creating the AV "driver" will indirectly benefit from this corporate spend on data centres, overseas fibre cable, servers, networks, etc and eventually, when Waymo gets into commercial ride share mode, on LIDAR, sensors, cameras, etc. Waymo cash on hand is $ 102 billion (I note that Tesla low cash position and lack of profitability may be constraints in these areas in the near term).

3. To give you some idea of the scale capability of Google Platforms, Google announced that Google Translate processes 140 billion word translations every day. Again, not directly related to Waymo but an indication of Google's ability to process at scale.

4. Google is "weaponising" Google Maps, infusing it with Google Assistant and AR features such that their machine learning/AI capabilities have increased by 25 times their ability to use their generated algorithms to build maps for use in Google Maps. These algorithms have added 110 million buildings into Google Maps. So this again gives an idea of how Google's algorithms can create the kind of data that an AV will need to cope with its task (ie, little if any human intervention at the point of use). Any Company aspiring to provide an AV driver has to compete with Google's AI/machine learning capabilities which seem to be world class.

Thoughts?
 
Ford taking on Waymo and GM Cruise by setting up its own self driving standalone division

Ford has set up Ford Autonomous Vehicles LLC following their purchase of Argo AI.

It expects to attract third party investment, much like GM Cruise attracted investment from Softbank.

Expect discussion when Ford reports earnings tomorrow.

https://nordic.businessinsider.com/...ivision-to-rival-gms-cruise-and-waymo-2018-7/
 
Waymo announced partnerships in Phoenix Arizona with 5 companies including WalMart, Autonation and Avis. The WalMart service, which is to begin later this week, it to provide its Early Rider clients who order their shopping on WalMart.com the ability to use a Waymo self driving vehicle to collect them, pick up their pre-ordered shopping and return them to their destination.

The partnership with Autonation is to provide "loaner" self driving vehicles to customers having their personal vehicle repaired, instead of Autonation using its own fleet.

Looks interesting:

http://wsau.com/news/articles/2018/...s-to-ferry-walmart-shoppers-in-arizona-trial/
 
sesevans said:
You might consider the following: collecting millions of miles of data on a car being driven with off and on human intervention and judging how a "semi-autonomous" (some might argue Level 2 Tesla vehicle) performs is lower quality data compared to Waymo's combo of billions of miles of virtual data from computer simulations that crunch real live autonomous data with millions of combinations and permutations and now 8 million miles of pure autonomous driving on public roads
Click to expand...

Waymo (and Uber for tha matter) seem to think the currency of public acceptance is miles of autonomous covergage, it is very much a PR campaign/strategy.
The car learning, they discuss as the beneficiary of these miles is being manuallly converted into a complex s/w state machine of scenarios (the neural net aspect is just for image recognition) and the intergity of the system is down to this coding;
at what point will they have the confidence to hand a car over to the automotive press (or independent authenticity) and have them set up a few obscure scenarios to see if the system is robust. ? they talk of setting up 20000 scenarios at their test village, it is pattern matching and the car does nonetheless not learn like a human.

There is little data on the variety of software solutions Waymo are exploring, they have tight commercial deadlines so innovative disruptive solutions are not on their radar,
I think, like the microsoft OS say, there is scope for them to be gazumped from their position., I read about 'startup' Zoox, $800M funding, there are enormous rewards and all the companies keep their algorithmic cards too close to their chest.

Imagine the car review magasines in a few years time, where we have mortality/accident statistics from the different AV vehicles with their different s/w solutions, you pay your moiney and make your choice ?
 
Waymo are an interesting propersition, they are likely to be first to market but I just don’t see them being able to make hyper profit levels from the sector in the long term. There are so many companies working on this now, a lot will not make it but the long term prospect of competition is currently looking high.

I suspect they may actually get pushed out of the market once mainstream car manufacturers catchup. There is really nothing stopping them doing it themselves. Without making their own cars they will slowly run out of possible partners and be left with relatively low volume companies that don’t have the resources to do it themselves.

Another possible scenario is the ‘driver’ turns into a subscription. At this point I think a regulatory body would step in and regulate the cars and ‘drivers’ in the name of competition. They could force the standardisation of the sensor suite and force the drivers to become swappable.

Either way the 15 year outlook is very interesting.
 
b0rn2sk8 said:
Waymo are an interesting propersition, they are likely to be first to market but I just don’t see them being able to make hyper profit levels from the sector in the long term. There are so many companies working on this now, a lot will not make it but the long term prospect of competition is currently looking high.

I suspect they may actually get pushed out of the market once mainstream car manufacturers catchup. There is really nothing stopping them doing it themselves. Without making their own cars they will slowly run out of possible partners and be left with relatively low volume companies that don’t have the resources to do it themselves.

Another possible scenario is the ‘driver’ turns into a subscription. At this point I think a regulatory body would step in and regulate the cars and ‘drivers’ in the name of competition. They could force the standardisation of the sensor suite and force the drivers to become swappable.

Either way the 15 year outlook is very interesting.
Click to expand...

You make a number of interesting assumptions about Waymo and their route to commercialisation/profit.

1. First to market. Keep in mind there are some reasons they are likely first to market. They have been working on their autonomous "driver" for nearly a decade. They were the first if not one of the first to recognise the need to jump to Level 4 autonomy rather than dabble in Level 2 or 3, which is what many "legacy" companies are doing. A "disrupter" can start with a clean sheet. The "mainstream manufacturers" like GM, Ford, Daimler, etc are, in my opinion, encumbered by the transition they need to make to bring their existing customers from Level 1 or 2 to higher levels of autonomy. They do not want to cannabilise their existing business. Further, Waymo, at least in the US, is becoming a well known brand and one that is razor-focused on passenger safety in their autonomous vehicles.

2. You imply a couple of routes: direct manufacturing (which Waymo has said it will not do), owning large fleets of vehicles and sharing them with passengers (eg, ride hailing in Phoenix starting this year and then with their fleet expanding within two years to more than 80,000 vehicles manufactured by Fiat Chrysler and Jaguar Land Rover) or licensing their "technology" once it is proven to Waymo's satisfaction. I believe that in a perfect world, Waymo would maximise its profit margins through licensing as they would not have their money tied up in fleets of owned vehicles. Yet the 80,000 vehicles could comfortably operate a ride hailing service in many US cities.

3. In an earlier post, I compared the idea of "winner take most" for AVs. I believe there are similarities to how Google approached search and mobile/Android. There were other search engines in the market when Google introduced their search product. There were other software platforms on mobile when Google introduced Android. They created the best mouse-trap for search and the open source solution with Android that has expanded explosively, with more than 5,000 different Android devices in the market today (with few manufactured by Google---except usually as a reference design). Waymo has already announced partnerships with Fiat Chrysler and JLR and the examples of how its Phoenix partners (WalMart, Avis, Autonation, etc--see earlier post) are able to use the "Waymo driver" to assist their customers is just the tip of the iceberg. Once Waymo has scaled its "passenger base", it will be able to earn additional revenues (probably the most significant contributer) from entertaining the passenger who will no longer need to focus on the road.

4. You have probably seen that some research analysts have begun assigning a monetary value to Waymo within Google. I have seen estimates of $ 100 billion in value already, before the first commercial "driver" hits the road.

5. A $ 102 billion cash arsenal at Google, other assets including Google Maps, Waze, machine learning and artificial intelligence Google Cloud and Android Auto infotainment should provide a number of barriers to entry to any but the most well capitalised competitor.
 
sesevans said:
1.
Further, Waymo, at least in the US, is becoming a well known brand and one that is razor-focused on passenger safety in their autonomous vehicles.
Click to expand...
Is there a start date yet ? this will be the big bang, and make or break (google itself ?), until then it is a marketting strategy (what is the budget).
Hopefully they can deliver on passenger safety (that is the message the public expected to hear though, similar to facebook saying they are bringing people closer together - the British are more cynical)

The partnership aspect also gives waymo a shared liability, in so much if there are problems they will not be held exclusively culpable. (unlike tesla AP problems).

Lack of figurehead in Waymo does not help either, in Telsa, he's marmite but Musk exists (like Jobs or Gates) but google lacks this.


sesevans said:
I have seen estimates of $ 100 billion in value already, before the first commercial "driver" hits the road.
Click to expand...
GM apparently has 'hard' assets of 200Bil, so for the point engineering problems of AV vehicles they can ramp up a competitor eg. Zoox. quickly.
 
I don't disagree with your arguments but being first to market doesn't mean you will be the biggest or the best long term. For example, Symbian, Palm, Windows Phone and Blackberry.

I could list 10+ extremely well funded companies that are doing the same thing as Waymo and ultimately there isn't anything stopping them making a product good enough to work on the road.

All of them say the biggest bottleneck is regulatory and not actually the software its self. It will be a few years before they are actually allowed on the roads without significant restrictions. That might as well be a lifetime in the software world. I also expect that other efforts are actually more advanced than is known because their efforts are just not widely followed or scrutinised anything like the extent that Tesla and Waymo are.

As the regulators don't actually allow AV's on the roads in any meaningful capacity there really isn't any reason to have product close to market and it's much more efficient to just have it bubbling away in the background until the time is right. Waymo are ultimately going to get to a point where they have a product that is ready to go but nowhere to actually sell it outside of a few niche markets because the regulators don't allow them to. Meanwhile everyone else is playing the quiet long game and there is nothing fundamental stopping them progress (like patents etc.).

I don't see anything in #5 being a significant barrier, in fact almost non of them really have any tie in's with AV's. The actual barriers to entry will greatly reduce over time as hardware gets better, cheaper and expertise in the field gets diversified as people move around the industry (which is extremely common).

Having $100 billion in the bank is great but they will not be able to burn through it to try and build a dominant market position without clear signs of profit. Google's shareholders expect profit, they are well beyond the silicon valley startup bubble these days. Just look at Facebook's share price, dropped 20% ($76 billion wiped of their value) in one day after they issued a profit warning below investors expectations.

AV's are not like mobile phones or computers where it is actually beneficial to the consumer to just have a couple of platforms as it creates a better user experience. The auto industry can get away with being very fragmented with lots of players in the game because there is no need 3rd party integration. In the unlikely event that Waymo are the only company to succeed then regulators like the EU will simply not allow Waymo to gain a monopoly they will force some kind of pseudo competition or simply put a price cap on it and essentially regulate the profit to a 'normal' margin. You also shouldn't underestimate a well represented unionised work force, just look how well the tube drivers do and their job is far easier to automate and the technology had been available for a decade....

Ultimately economics will win this war, the cheapest and not the best product will be the driving factor in the market. Even if only 5 of those companies actually make it, there is enough competition to drive profits to a fairly low margin. Companies like GM with their cruise software will be significantly more vertically integrated than Waymo which will greatly increase their competitiveness. Waymo need to license or partner with an automaker who also want their own piece of the pie. Waymo will ultimately have a place in the market, I just don't see them becoming the dominant player long term (in say 10 to 15 years), just in the same way I don't ever see Tesla out selling VW in EV's long term but both will have their place in the market.
 
You must log in or register to reply here.
Back
Top Bottom