Author Topic: EEVblog #1068 - Autonomous Uber Incident Update  (Read 177335 times)

0 Members and 1 Guest are viewing this topic.

Offline metrologist

  • Super Contributor
  • ***
  • Posts: 2199
  • Country: 00
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #100 on: April 06, 2018, 09:13:42 pm »
that's the idea, nobody needs to own a vehicle. it comes and gets yous...
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #101 on: April 06, 2018, 09:16:55 pm »
But I want to own a vehicle, and I want to drive it myself, and so do millions of other people. For just moving people around autonomously a mix of buses, trains and taxis makes more sense, and once you reach that point the gains of eliminating the human driver are less.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #102 on: April 06, 2018, 09:53:51 pm »
If manually driven cars ever disappear completely it won't happen in our lifetime. Autonomous cars will slowly become more and more common, it's not like they will replace human drivers over night.
 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 26892
  • Country: nl
    • NCT Developments
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #103 on: April 06, 2018, 10:59:33 pm »
that's the idea, nobody needs to own a vehicle. it comes and gets yous...
I doubt that. Think about the holiday season where millions of people go on a holiday at the same time. There will have to be enough cars for that and those cars will sit mostly idle for the rest of the year. Someone will want to get a return on investment on the cars, maintenance, parking space, etc. And who is going to clean the cars? These things will make (basically) renting a car more expensive than just owning one.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 8637
  • Country: gb
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #104 on: April 06, 2018, 11:05:58 pm »
that's the idea, nobody needs to own a vehicle. it comes and gets yous...
Perfect for rush hour, eh?
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #105 on: April 07, 2018, 12:08:26 am »
Buses and light rail are perfect for rush hour, large numbers of people going from one place to another at roughly the same time. It's a far more efficient method of transportation than personal cars for that situation.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 8637
  • Country: gb
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #106 on: April 07, 2018, 12:15:55 am »
Buses and light rail are perfect for rush hour, large numbers of people going from one place to another at roughly the same time. It's a far more efficient method of transportation than personal cars for that situation.
In a dense place like Hong Kong public transport can be fantastic. Most western city residential areas are too spread out for frequent public transport runs to reach close to each home. Its hard to see how public transport will ever be satisfactory for the high percentage of homes not close to a good public transport stop.
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #107 on: April 07, 2018, 12:20:05 am »
Well the way it works here in my corner of the US is there are park & ride lots in most suburbs with direct bus routes to and from the major cities. You walk, drive or bike to the local park & ride, bus to work and then bus back to the park & ride in the evening. When I worked downtown for a while I did that and it worked pretty well, the majority of my travel was on the bus taking up a fraction of the highway capacity of all those people in individual cars.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 8637
  • Country: gb
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #108 on: April 07, 2018, 12:28:23 am »
Well the way it works here in my corner of the US is there are park & ride lots in most suburbs with direct bus routes to and from the major cities. You walk, drive or bike to the local park & ride, bus to work and then bus back to the park & ride in the evening. When I worked downtown for a while I did that and it worked pretty well, the majority of my travel was on the bus taking up a fraction of the highway capacity of all those people in individual cars.
Park&ride schemes can help with city centre congestion, but do little to reduce driven miles, car ownership, etc. They also require considerable space for the car park at the park&ride terminals.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #109 on: April 07, 2018, 04:51:24 am »
There would have to be enough cars to be able to handle peak load, but that would still be much less than what we have today. Autonomous cars also help a lot with congestion for various technical reasons. Having mainly autonomous taxi services could in theory help with car pooling as well.
 

Online G7PSK

  • Super Contributor
  • ***
  • Posts: 3860
  • Country: gb
  • It is hot until proved not.
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #110 on: April 07, 2018, 08:22:24 am »
How well will autonomous cars work in the little country lanes we have where I live, lots of bends some sharp others not but most blind due to hedges and banks. Legally these roads are open to the maximum speed limit for single carriageways in the UK 60mph 97kph you would have to be an absolute idiot to do anywhere near that (there are a few) but on the long curving bend it is possible but they are blind and you cannot see more than about 25 meters sometimes less ahead and radar wont work due to embankments etc. Around these corners you can come accross horses, cyclists, pedestrians, farm machinery or even trucks that cannot be seen until the last moment, and as this is a holiday area we do get people from towns who do not realise that just because it is possible to drive fast on the bend that it is not safe to do so, it keeps the local recovery teams busy in the holiday seasons. But how will self driving cars handle such things.
 

Offline Towger

  • Super Contributor
  • ***
  • Posts: 1645
  • Country: ie
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #111 on: April 07, 2018, 08:48:21 am »
How well will autonomous cars work in the little country lanes we have where I live, lots of bends some sharp others not but most blind due to hedges and banks.

They won't.  Americans have very little experience if such roads.  There are roads near here in the Dublin Mountains with all the above,  plus deep ditches along the sides and a very steep gradient.  They are barely wide enough for two cars to pass with one of the cars stopped and pulled into the side.  In places not even that.  Yet you can meet a double decker bus coming along.  Will the car be able to reverse back a 100m to allow a larger vehicle pass?

The worring problem is drivers skill levels will drop the more they rely on the car.
« Last Edit: April 07, 2018, 08:51:19 am by Towger »
 

Online G7PSK

  • Super Contributor
  • ***
  • Posts: 3860
  • Country: gb
  • It is hot until proved not.
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #112 on: April 07, 2018, 09:50:08 am »
I already see that with the big BMW 4X4's with drivers from London who do not realise that their car can drive on the verge in order to pass.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #113 on: April 07, 2018, 01:17:46 pm »
As someone pointed out, at least Googles cars rely heavily on detailed maps, which means you could just mark a good meeting spot on the map, and where you can drive on the verge. To allow for the cars to reverse back you would only need to add a rule that doing so is acceptable on this type of road. Enabling driving on narrow British country roads are probably far down on the to-do list though.

Worst case is that the car stops and asks the occupants (or a remote operator) for assistance.

One thing the cars really can not handle today is snow though, so no autonomous vehicles in the north for a while, at least during winter.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #114 on: April 07, 2018, 05:20:09 pm »
If the car cannot handle all situations automatically then will it need to have driving controls? If so, then the person in that seat will be needing a driving license. What happens if they are a person who wouldn't be capable of getting a license? Wasn't that part of the pitch in favour of AV's in the first place? Give freedom to the blind, the elderly and the infirm.
I believe both models with and without manual controls are being developed. In a taxi there will likely be none, I doubt the taxi company would like the passenger to start driving the cars manually. For privately owned cars there would probably be manual controls.

For the most part you wouldn't need driving controls, just a hint what to do. E.g. the car would say the road is blocked and give the occupants of the car some options on how to solve the problem: reverse to previous intersection, wait for someone to clear the block, pull over and park, call for assistance, switch to manual controls etc.

If the person in the car cannot answer that or the car has no human passengers, it could be handled by an operator at some other location who could see what had happened based on the sensordata and then help choose an action that solves the problem. Then they could tell other cars to avoid that road so other cars didn't get into trouble.

Who will want to be telling your Uber or taxi what to do and how to do it. You may as well have a taxi driver. Where's the progress.

There are many advantages:
* Safer driving. Self driving cars will drive more comfortably and much safer than your average taxi driver.
* No BS. You know the robocars will pick the best route. Not make detours and not charge you extra.
* Safe from crime: You don't have to fear being robbed, or if your a girl, to be raped.
* No driver you have to pay a salary. That is no doubt the highest part of the cost of a taxi journey. Unfortunately that probably won't matter much for the passengers since the companies will keep charging the same rates, but it's a huge incentive for the taxi companies at least.
* 1 extra seat for passengers in every car.
* Probably many other advantages i haven't thought of.

Who will want a remote operator in control? They have no skin in the game like a human driver does. You want someone motivated to live another day.
That is a fair point. The manufacturers and operating companies have a huge economic incentive not to kill their customers though.

As soon as they start killing dogs you'll have PETA and the RSPCA and every unaffiliated animal lover outraged and campaigning for AV's to avoid anything cute and furry on the roads.
In February Waymo had driven over 8 million km (5 million miles) autonomously on public roads without killing neither humans or dogs. I think it's safe to say human drivers kill more animals (humans and others).

And if I could customise the firmware I'd be swerving to flatten Cane Toads. Like I used to do.

Someone will hack the firmware and have them seeing Pokemons.
Yeah, that is a scary thought. You can make hacking prohibitively expensive/difficult but it also means an added cost for the manufacturers/developers, and they might skimp on that. Still a few hacked robocars will be better than the countless number of human idiots that drive today.

And to get back to being serious I recall many years ago following a car at about 60kmh on a road that carried heavy trucks. The trucks had worn the road surface so it had two ruts where the wheels traveled and in the rain that day shallow longitudinal puddles had formed. The car in front of me in a split second was wrapped around a tree. Ever since that day I always avoid such standing water. There are a couple of places in my vicinity where in heavy downpours the road can partially flood and drivers move from the kerbside lane  to avoid hitting 20-30cm deep water across part of the lane. Can an AV reliably do that?
Technically it's not a problem for the car to steer around it, it controls the car better than a human and has much better situational awareness. Would the sensors detect it? I believe the LIDAR would see a puddle of water as a flat surface for example, so you would have to rely on the cameras to detect such things. However, since the cars today rely on detailed maps they could have such road features annotated so the car knew in advance it should avoid them.

No one thinks robocars will be perfect, only that they will be a lot safer than human drivers. Even if it can't avoid every single type of danger as well as you would have, it can still prevent far more accidents that you wouldn't/couldn't have. So statistically it will be a much safer bet to ride with a robocar than a human driver (most likely yourself included).

What happens if a box or a plank or a PVC pipe falls from a vehicle ahead? If you can swerve, do you? What will an AV do? I once ran over a mattress that was poorly secured. I did it in a split second deliberately choosing to not swerve from my lane or brake hard. If the mattress was a bookcase I'd have responded differently. Not to say effectively but I wouldn't have just run straight over it by choice.
I'm almost certain no autonomous car today can tell the difference of a mattress and a bookcase so the car would have braked and/or swerved to avoid collision. But again, robocars will never be perfect but a ride with a robocar will still be the safer bet.

The situations that present little challenge to a human need to be programmed. But you cannot cover every imaginable situation.
It doesn't have to be perfect to be useful. As long as they fail safely (e.g. pull over and stop) and it happens very rarely it won't be a showstopper. Don't forget they have many advantages to human drivers. E.g. they don't fall asleep or get a stroke, etc, etc.

There is more than just bicyclists and pedestrians and other vehicles to be avoided. And as soon as the accidents start and people and property are harmed there will need to be legal redress, financial redress and blame apportioned. None of this seems to be anywhere but in the too hard basket.
Well, as I said, Waymo alone has driven over 8 million km autonomously on public roads without any such problems (billions of km in simulator). So doesn't look like there will be a lot of accidents, but time will tell. If one model of robocars did cause more accidents than humans then they shouldn't be allowed to drive on public roads obviously. The standards for robocars can hopefully be set much higher than that of human drivers, but what regulations there will be is ultimately a political decision.
« Last Edit: April 07, 2018, 05:37:42 pm by apis »
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #115 on: April 07, 2018, 05:39:59 pm »
People have always underestimated what can be automated, not long ago lots of people said computers would never be able to beat a human at Chess, (much less Go). If you suggested computers would soon be automatically translating languages (not to mention automatically adding translated subtitles to video) people would either yell or laugh at you. But lo and behold: computers are way better than the best human players at both Chess and Go by now. To make a really good translation you need a lot of contextual understanding which computer translators lack at the moment, but google translate and such programs are already good enough to be very useful.

Humans are just a biological machine (although a very complex one) that have been shaped and programmed by evolution, there isn't anything magical humans can do that a machine can't.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 8637
  • Country: gb
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #116 on: April 07, 2018, 05:47:07 pm »
People have always underestimated what can be automated
Nonsense. People have badly underestimated what can be automated in some areas, and massively overestimated what can be automated in other areas. People typically find it hard to believe that anything they do can be easily automated, yet find it easy to believe that the activities of others can.
 

Offline drussell

  • Super Contributor
  • ***
  • Posts: 1855
  • Country: ca
  • Hardcore Geek
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #117 on: April 07, 2018, 07:09:14 pm »
...
* Safe from crime: You don't have to fear being robbed, or if your a girl, to be raped.

Wait...  What?!

How is someone safer from attack in an autonomous car?  Is the car going to magically come to your defense somehow if you're being attacked?!

I would think that having another human around would make travel significantly safer than a lone person traveling in an autopod...
 
The following users thanked this post: nugglix

Offline TheAmmoniacal

  • Supporter
  • ****
  • Posts: 1188
  • Country: no
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #118 on: April 07, 2018, 08:06:14 pm »
...
* Safe from crime: You don't have to fear being robbed, or if your a girl, to be raped.

Wait...  What?!

How is someone safer from attack in an autonomous car?  Is the car going to magically come to your defense somehow if you're being attacked?!

I would think that having another human around would make travel significantly safer than a lone person traveling in an autopod...

I assume it refers to being attacked by the driver.
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #119 on: April 08, 2018, 12:11:37 am »
Park&ride schemes can help with city centre congestion, but do little to reduce driven miles, car ownership, etc. They also require considerable space for the car park at the park&ride terminals.

They help greatly around here where the freeway/motorway is the choke point along with the roads and limited parking downtown. Many thousands of cars stay in the suburbs and those people take a bus to downtown, keeping thousands of cars off the freeways and out of the crowded downtown core.

Autonomous cars that are not privately owned don't solve the issue anyway, all they do is eliminate the taxi driver, it's a fancy way of replacing another relatively low skill job with automation. Autonomous cars that are privately owned don't solve any problem except laziness and/or inattention. If they can take themselves somewhere without anyone on board then I imagine we'll see hundreds of them clogging the streets circling the block while their owners eat lunch or conduct business once they figure out that driving a few miles is cheaper than parking for a half hour.
 
The following users thanked this post: nugglix

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 16642
  • Country: 00
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #120 on: April 08, 2018, 09:26:10 am »
I assume it refers to being attacked by the driver.
Yep. All those fake Mexican taxis that drive you to quiet place where the rest of the gang is waiting. They'll be out of business.

 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #121 on: April 08, 2018, 03:06:46 pm »
How is someone safer from attack in an autonomous car?  Is the car going to magically come to your defense somehow if you're being attacked?!

I would think that having another human around would make travel significantly safer than a lone person traveling in an autopod...
Taxi drivers are much more at risk of being robbed by the passengers of course, but the opposite also happens. And apparently it is much more common for women to be sexually harassed or raped than what most people believed (before the metoo thing), so a young lady who've had a few drinks might prefer to take a robocar taxi home. That humans won't have to do the risky job of driving taxis is another bonus.

There will also be a way to contact a human operator at the taxi company in each car, so you could use that to call for help if the car was attacked by people on the outside I suppose.
« Last Edit: April 08, 2018, 03:09:24 pm by apis »
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #122 on: April 08, 2018, 03:30:25 pm »
They help greatly around here where the freeway/motorway is the choke point along with the roads and limited parking downtown. Many thousands of cars stay in the suburbs and those people take a bus to downtown, keeping thousands of cars off the freeways and out of the crowded downtown core.
And you could have the same system with robocars, you just take your (or a taxi) robocar to the park&ride area and then you switch to robobus/train.

Autonomous cars that are not privately owned don't solve the issue anyway, all they do is eliminate the taxi driver, it's a fancy way of replacing another relatively low skill job with automation.
In a functioning free market the price of taxi travel would go down significantly compared to today, however there is a risk there will only be one or a few very large companies offering the service (as is so common these days), and then you no longer have a working free market so the prices will not 'magically' reach the optimum value. But that is a political problem, not a problem with the technology.

Autonomous cars that are privately owned don't solve any problem except laziness and/or inattention. If they can take themselves somewhere without anyone on board then I imagine we'll see hundreds of them clogging the streets circling the block while their owners eat lunch or conduct business once they figure out that driving a few miles is cheaper than parking for a half hour.
Operating a car today costs about $20/hour, which is much more than any hourly priced parking. If you have a robot car you could just let it drop you off at the restaurant, etc, then it checks online for the cheapest free parking lot nearby and drives there directly and parks. So it will probably mean less driving compare to humans driving around looking for a free parking spot. Robot cars also have the potential to park much closer together (if there is no one inside that has to get out) so you can utilise the available parking space more efficiently.
« Last Edit: April 09, 2018, 09:38:10 am by apis »
 

Offline Muttley Snickers

  • Supporter
  • ****
  • Posts: 2340
  • Country: au
  • Cursed: 679 times
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #123 on: April 09, 2018, 04:41:38 am »
Their department responsible for the R&D could probably learn a thing or two from B&R, these fellows had it all sorted out and working a treat fifty odd years ago from what I remember.   :D ;D

     
 
The following users thanked this post: apis

Offline gregariz

  • Frequent Contributor
  • **
  • Posts: 545
  • Country: us
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #124 on: April 09, 2018, 05:46:36 pm »
Unlike what the Sheriff said, it's looking more like a preventable accident. From Ubers video I thought the incident happened in a less built up area.
But those mobile phone videos show otherwise. Even though  digital cameras over expose at night when set to automatic, the mobile videos show many light sources and Ubers video shouldn't been so dark.   

And it probably doesn't matter either way, because it's the LIDAR that should have picked it up and it didn't, under almost ideal practical circumstances.

Yes, I am in Phoenix quite often. Phoenix has been the testing ground for a lot of self driving car testing. You can barely drive anywhere without passing them several times a day - I don't know how many are on the roads in phoenix but it's a lot.  There are a couple of reasons for it but I suspect the prime one is that Phoenix has an ideal road system - ie the whole city is made up of roads that go east-west and north south - however it is a big city so there can be plenty of traffic to contend with. There are relatively few winding roads anywhere in the city.

In my own driving experience I have seen them do dangerous things on 2 occasions. One was they cut me off at the lights as I was right turning and had right of way to move into any appropriate lane. The google self driving car decided it would cross traffic and enter into the centre lane - the one I was going to take as I had a left turn further ahead. The second case was on a left turn across traffic into a residential street. The green light was given and it sat there jerking for a bit. 2 cars behind it (including me) eventually just drove around it and entered the street. It finally made it into the street before pulling over.

I honestly can't say I have seen an Uber self driving car - unless they are using Google's but IIRC they are not Volvo's. I have used Uber here but as a matter of principal generally avoid them as I worry about the quality of driver testing and insurance.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf