its a model X in that video, so if that means ap2.5 too, then like my previous post maybe the ap1 model S would not have had a problem ... due to the better mobileeye.
the comments in the video I posted included
Autonomous Vehicles
- Thread starter adam cool dude
- Start date
the comments in the video I posted included
A Model X could have AP1, AP2 or AP2.5 depending on when it was built. The Model X release date was October 2015, AP2 came out in October 2016.
Yes exactly, the car uses computer vision to look at the lane lines and drive between them, it also 'follows' the car in front.
It seemed to pick up the stronger white line on the left, the one on the right looks patchy at the start which could be an explanation.
It should also be noted that there are no rumble strips on that intersection which is really odd, you would normally have rumble strips in the line its self or embedded in the concrete on the inside of the line to warn you that you are leaving the lane. Another contributing factor is that the crash barrier was taken out a few days before. In the UK there would have been at least a full set of cones and likely a speed restriction before a hard concrete barrier if a repair wasn't put in place immediately. You can see from the video that the wall has been repaired.
Based on the above video I do agree with Tesla's statement that the driver had 5 seconds to react to the car leaving the lane before impact. Just like the Uber crash this one was avoidable if the driver was paying attention.
The difference between Uber's and Tesla's systems are quite clear though, one is intended to be a full level 5 the other is barley level 2 and should be treated as such.
Waymo's research on humans in this circumstances is 100% correct though, we inherently do not pay attention. Same can be said if we are not relying on the car to do the work also...
It might be interesting to consider a subsidiary thread within our Autonomous Vehicle thread that while includes AVs also goes towards integrating AVs on a grander scale, namely Smart Cities. Am I alone here in thinking that Smart Cities is an interesting topic to include in our AV thread? A Smart City is likely to include AVs, robots, smart buildings that can be heated with clean energy and be easily convertible from shops to housing or vice versa depending on demand, etc.
I distinguish between a Planned City and a Smart City. Planned cities have been around for years---heck, Louis XIV moved his court from Paris to Versailles---and they include Brasilia in Brazil, Abuja in Nigeria, etc. Smart cities are a more recent phenomenon, so far with mixed results: Songdo in South Korea, approximately 25 miles from Seoul, was promised to be an alternative to the overpopulated Seoul but so far has been overpriced ($ 35 billion+), over-hyped and under-populated---now a case study for urban planners.
http://www.scmp.com/week-asia/busin...reas-smart-city-songdo-not-quite-smart-enough
On the drawing boards is Neom, a smart city to be built from the ground up in Saudi Arabia at an estimated cost of $ 500 billion. It is planned to be larger than Dubai and have more robots than humans.
I have previously described on these threads the partnership between the Google subsidiary Side Walk Labs and the city of Toronto planning to develop initially 12 acres of dis-used land on Toronto's Eastern Waterfront, attempting to reinvent urban space.
https://sidewalktoronto.ca/#vision
The goal in Toronto is to improve quality of life, energy use, transportation and affordable housing including buildings that can be repurposed from commercial to residential or back again, etc.
I also note a smart city concept for Weymouth, Mass., near Boston involving urban planners and GE:
https://www.wraltechwire.com/2018/0...from-the-ground-up-with-a-raleigh-firms-help/
So in this context, autonomous vehicles to be used for personal transport as well as delivery and logistics form an important part of any of these proposed smart cities.
Does anyone know what the UK plans are for smart (rather than planned) cities? Or will planners simply attempt to upgrade existing cities?
Comments?
I distinguish between a Planned City and a Smart City. Planned cities have been around for years---heck, Louis XIV moved his court from Paris to Versailles---and they include Brasilia in Brazil, Abuja in Nigeria, etc. Smart cities are a more recent phenomenon, so far with mixed results: Songdo in South Korea, approximately 25 miles from Seoul, was promised to be an alternative to the overpopulated Seoul but so far has been overpriced ($ 35 billion+), over-hyped and under-populated---now a case study for urban planners.
http://www.scmp.com/week-asia/busin...reas-smart-city-songdo-not-quite-smart-enough
On the drawing boards is Neom, a smart city to be built from the ground up in Saudi Arabia at an estimated cost of $ 500 billion. It is planned to be larger than Dubai and have more robots than humans.
I have previously described on these threads the partnership between the Google subsidiary Side Walk Labs and the city of Toronto planning to develop initially 12 acres of dis-used land on Toronto's Eastern Waterfront, attempting to reinvent urban space.
https://sidewalktoronto.ca/#vision
The goal in Toronto is to improve quality of life, energy use, transportation and affordable housing including buildings that can be repurposed from commercial to residential or back again, etc.
I also note a smart city concept for Weymouth, Mass., near Boston involving urban planners and GE:
https://www.wraltechwire.com/2018/0...from-the-ground-up-with-a-raleigh-firms-help/
So in this context, autonomous vehicles to be used for personal transport as well as delivery and logistics form an important part of any of these proposed smart cities.
Does anyone know what the UK plans are for smart (rather than planned) cities? Or will planners simply attempt to upgrade existing cities?
Comments?
Who needs maps and GPS for an AV?
I saw this interesting experiment from Deep Mind, the original British AI company purchased by Google a few years ago, in which it directs a self driving vehicle around cities using only Google Street View images! They say this is for instances when GPS or mobile phone service is down. Impressive!
Result: gets you from point A to point B without maps or GPS, only AI.
https://www.outerplaces.com/science/item/18240-artificial-intelligence-google-street-view
I saw this interesting experiment from Deep Mind, the original British AI company purchased by Google a few years ago, in which it directs a self driving vehicle around cities using only Google Street View images! They say this is for instances when GPS or mobile phone service is down. Impressive!
Result: gets you from point A to point B without maps or GPS, only AI.
https://www.outerplaces.com/science/item/18240-artificial-intelligence-google-street-view
Still much work to do to convince California residents of the safety of self driving vehicles, if this poll, taken four days after the fatal Uber crash in Arizona, is any guide:
https://www.vcstar.com/story/news/2...s-wary-self-diving-cars-poll-finds/496974002/
https://www.vcstar.com/story/news/2...s-wary-self-diving-cars-poll-finds/496974002/
Jaguar Land Rover is a relatively small automobile manufacturer, when compared to many of its established rivals. But according to this article is Forbes from conversations with JLR's US CEO, its relatively small size can make it much more nimble in changing direction and spotting market opportunities.
It took JLR two months to reach a partnership deal with Waymo to produce 20,000 Jaguar I-Pace EVs for Waymo's self driving service, something that would take far longer for a large OEM much longer to decide. And JLR has made impressive gains overall in the US market with record years in 2016 and 2017 (although off to a slower start in 2018). JLR North America has now introduced three new vehicles in the past year in the US: the F-Pace SVR, the Range Rover SV coupe and now the I-Pace SUV EV.
https://www.forbes.com/sites/dougne...bets-to-best-larger-competitors/#2cd31e377a00
It took JLR two months to reach a partnership deal with Waymo to produce 20,000 Jaguar I-Pace EVs for Waymo's self driving service, something that would take far longer for a large OEM much longer to decide. And JLR has made impressive gains overall in the US market with record years in 2016 and 2017 (although off to a slower start in 2018). JLR North America has now introduced three new vehicles in the past year in the US: the F-Pace SVR, the Range Rover SV coupe and now the I-Pace SUV EV.
https://www.forbes.com/sites/dougne...bets-to-best-larger-competitors/#2cd31e377a00
wrt to driver assist technology (for me a subset of AV technology) accidents, what kind of reports, if any, do Tesla produce for these ?
there are incidents/fatalities over several years so they must have had time to reach conclusions, same go for other manufacturers too, I guess.
I would expect manufacturers would want to differentiate themselves for marketting etc. but if the capabilities of these systems are subjective then what weight would a potential purchaser give to them ? - Yes you legally have to maintain control, but, would like to believe the system can be helpful.
(For driver attention/drowsiness warning I have seen mention of eye tracking, so incriminating data from these could be used for prosecutions I imagine where assist ha been wrongly used)
there are incidents/fatalities over several years so they must have had time to reach conclusions, same go for other manufacturers too, I guess.
I would expect manufacturers would want to differentiate themselves for marketting etc. but if the capabilities of these systems are subjective then what weight would a potential purchaser give to them ? - Yes you legally have to maintain control, but, would like to believe the system can be helpful.
(For driver attention/drowsiness warning I have seen mention of eye tracking, so incriminating data from these could be used for prosecutions I imagine where assist ha been wrongly used)
There is of course a big debate over the value of driver assist technologies and whether the human driver can play any role as we move to more advanced categories. This link to NHTSA provides an accurate definition of categories leading up to the ultimate self driving vehicle (category 5). Waymo's fleet in Chandler Arizona is a Level 4 fleet. Tesla is a Level 2 fleet.
Good luck in trying to find the section in the Official Tesla Website dedicated to Accident Reports. Tesla's approach is defined in this link I found from First Post and links also to a recent Guardian article--giving accident reports directly to the media without first consulting its customers. The other day NHTSA issued a critical comment about this practice in respect of the recent Tesla fatality in California.
https://www.firstpost.com/auto/tesl...without-consulting-its-customers-3727467.html
I have mentioned it before as it seems most credible. Note that it reflects the conclusions reached by Waymo in 2015: humans get distracted easily while driving and learn to grow overly trusting of driver assist technologies to the point where they assume the driver assist technologies are self driving technologies. They are not.
Eye tracking of a human in a driver assist vehicle is a way of checking if the human driver complies with the manufacturers' instructions. It will help determine fault. But it still misses the big picture: Get humans away from any driving input as quickly as the self driving technology permits.
Thanks (I had previously read the excel sheet with the 13 hands-on Waymo AV events), but as you say, for assist they are supposedly providing assist crash data to the media (probably like other manufacturers) whether they would be guilty if the system diverted the user into an fatal situation, or prevented them getting control.
Tesla seem to have jumped the gun on california - yes
I liked this comment on tesla site
big problem of mid lane hoggers in the uk.
Yes, if proof were needed that Tesla's Autopilot is at best a Level 2 solution.
With Waymo and others also beginning to focus their attention on delivery vehicles/logistics and with a number of companies, including Google, investing in blockchain technology, a picture is emerging of what the future will look like for logistics---a highly efficient one. According to some industry players, shippers and carriers will be more seamlessly connected and there will be fewer intermediaries (ie brokers) needed.
http://www.ttnews.com/articles/visions-future-logistics
http://www.ttnews.com/articles/visions-future-logistics
Bloomberg: Alphabet's Waymo taking a page out of Google Android's playbook
Author asks if Google intent on using its cash balance to buy and place millions of vehicles on the road to test its ride sharing chops against others or whether it will, like with Android, leverage lesser scale auto manufacturers (Fiat Chrysler, Jaguar Land Rover and perhaps Honda) to achieve what they were successfully able to do with Android 10 years ago when they launched with lesser scale players like HTC, Samsung (the fourth and sixth largest phone manufacturers in 2008) and through a superior Android experience was able to leverage their relationships to help their partners and themselves to the leading market share. Fiat Chrysler and Honda spend much less on R & D than their major competitors. Will a superior "driver" mean that the larger automakers will have to end up licensing their technology from Waymo?
https://www.bloomberg.com/gadfly/ar...takes-a-page-out-of-google-s-android-playbook
Author asks if Google intent on using its cash balance to buy and place millions of vehicles on the road to test its ride sharing chops against others or whether it will, like with Android, leverage lesser scale auto manufacturers (Fiat Chrysler, Jaguar Land Rover and perhaps Honda) to achieve what they were successfully able to do with Android 10 years ago when they launched with lesser scale players like HTC, Samsung (the fourth and sixth largest phone manufacturers in 2008) and through a superior Android experience was able to leverage their relationships to help their partners and themselves to the leading market share. Fiat Chrysler and Honda spend much less on R & D than their major competitors. Will a superior "driver" mean that the larger automakers will have to end up licensing their technology from Waymo?
https://www.bloomberg.com/gadfly/ar...takes-a-page-out-of-google-s-android-playbook
In this note in Ars Tecnica, the author argues for a fundamental rethink in how to regulate self driving vehicles. When self driving vehicles appear in great numbers in the next couple of years, they will be hired, not sold as they will be employed by fleet owners such as Waymo, GM, etc for use in ride hailing.
Comments?
https://arstechnica.com/cars/2018/0...f-driving-cars-is-broken-heres-how-to-fix-it/
Comments?
https://arstechnica.com/cars/2018/0...f-driving-cars-is-broken-heres-how-to-fix-it/
Uber CEO now saying that they see a future in autonomous vehicles and want to play a part.
They are co-operating with the Federal authorities and have not resumed testing since the fatal crash. They have no timeline yet for resuming testing.
https://uk.reuters.com/article/us-u...vehicles-have-a-future-says-ceo-idUKKBN1HI2KI
They are co-operating with the Federal authorities and have not resumed testing since the fatal crash. They have no timeline yet for resuming testing.
https://uk.reuters.com/article/us-u...vehicles-have-a-future-says-ceo-idUKKBN1HI2KI
suppose this is to be expected,
for silicon valley autpilot crash
also an 'amusing' conclusion by tesla
(I think the people who can afford to own tesla's will have different accidengt statistics to the rest of the population so am not convicned the conclusion the 3.7x reduced fatalities would apply for general population) - where do they get off.
for silicon valley autpilot crash
also an 'amusing' conclusion by tesla
(I think the people who can afford to own tesla's will have different accidengt statistics to the rest of the population so am not convicned the conclusion the 3.7x reduced fatalities would apply for general population) - where do they get off.
I assume the lawyers see the AV as an extension of their existing lucrative business and figure early on, they will test their chances against Tesla. They must figure that Tesla would rather settle than go through a potentially lengthy trial. Should be interesting to see what Tesla's approach will be: stand and fight or settle.
Anyone taking bets on which way Tesla will jump?
It is for this very reason that I predict that soon Alphabet will spin out Waymo as a separate legal entity to mitigate risk from lawsuits. Would not be surprised to see other AV related companies do the same.
Early indications seem to be that Tesla will stand and fight---blaming the victim. See their comments Tuesday and this note from Ars:
https://arstechnica.com/cars/2018/0...as-well-aware-that-autopilot-was-not-perfect/
Have Tesla previously settled similar cases (maybe their offer was refused by relations in this case), otherwise, this is THE test-case ?,
I wonder how the defense is funded, any crowd-funding ?
(building on my earlier post) If they find that the AP1 version of the hardware would have avoided the problem, versus the Ap2.5 version that should be damning,
saving money with their own solution versus the Mobileye one, it does seem contentious which drivers think is better.
If they have prior documentation of similar cases where AP was failing, and had not told the users, that seems delinquent; it should be like grounding aircraft, they tell users to disable system (or enforce) until problem is identified ?
I wonder how the defense is funded, any crowd-funding ?
(building on my earlier post) If they find that the AP1 version of the hardware would have avoided the problem, versus the Ap2.5 version that should be damning,
saving money with their own solution versus the Mobileye one, it does seem contentious which drivers think is better.
If they have prior documentation of similar cases where AP was failing, and had not told the users, that seems delinquent; it should be like grounding aircraft, they tell users to disable system (or enforce) until problem is identified ?
@jpaul @dowie
From the NTSB today: "....uncoordinated releases of incomplete information do not further transportation safety or serve the public interest."
The NTSB was unhappy with Tesla releases before it finished its investigation.
Tesla said it wanted transparency.
Hmmmm....
https://www.theverge.com/2018/4/12/17229518/tesla-ntsb-autopilot-crash-investigation-removed