The Self-Driving Vehicles Thread

[Horse]

V2V comms would sort that. The most challenging period is the transition before all vehicles are equipped and you have to deal with hoomans.

With one finger typing on the phone, apologies CJ, I'm going to comment here

Some of what you ask, depending obviously on mapping (V1) or programming (V2), they could be aware of / count / negotiate to passing places better than most human drivers.

There are several videos on the page, one shows it meeting and dealing with an oncoming car. I don't know whether there was any 'negotiation' . However, the article talks about AI learning. How, I have no idea, but that knowledge is then available across the fleet and could continue throughout the vehicle's life.

One potential issue with (re narrow lanes) AVs is that they can be too accurate with steering, potentially focusing wear of the road surface. And the sensor systems and the vehicle's 'awareness' of its size and position will, again, probably be far better than most drivers. Self-park has been available for years. But they may be unable to determine whether there's an overgrown bush to brush past, or a solid wall to barge into.

However, I've seen (early last year) a demonstration of a video analytics system that - then - could identify 1000 different items. As a f'rinstance, heading into The Mall in London, it not only identified (placing a bounding box on screen) usual stuff such as pedestrians, cyclists, cars, bollards, traffic lights - but also (identified and boxed separately) a horse and rider.

Re emergency vehicles, many already have GPS location live to control rooms, wifi hotspots, livestream video capabilities. Only a small step to broadcast location and direction to the EV and 'arrange' a clear path through.

None of this will happen overnight, or anytime soon. That's an understatement

But - despite the hype, smoke and mirrors, missed deadlines - the progress in systems development has really been quite incredible.


PS Did anyone notice the car breaking a couple of laws?

[Cousin Jack]

V2V comms would sort that. The most challenging period is the transition before all vehicles are equipped and you have to deal with hoomans.

Quite. There are vehicles on the road today that date back to the 1920s, so 100 years of chaos during the 'transition'.

[Horse]

The most challenging period is the transition before all vehicles are equipped and you have to deal with hoomans.

Quite.

So they're doing the development in worst case conditions.

[slowsider]

V2V comms would sort that. The most challenging period is the transition before all vehicles are equipped and you have to deal with hoomans.

Quite. There are vehicles on the road today that date back to the 1920s, so 100 years of chaos during the 'transition'.

Not once the oil stops

[Cousin Jack]

Not once the oil stops

Some of the oldest ones run on coal.
And could probably manage on wood at a push.

[The Spin Doctor]

o deploy in a new location, an AV1.0 team starts by manually driving sensor-equipped vehicles down every street so they can paint a 3D picture of the environment, down to the centimeter. They process this data into a detailed map with additional context such as speed limits, lanes, and traffic light locations. The maps are then tested and verified before being deployed, as well as constantly updated as city streets change.

So they drive every street on a daily basis to check for new potholes that drivers have to swerve around?

And do they get first dibs on going round a new road layout?

[Horse]

So they drive every street on a daily basis to check for new potholes that drivers have to swerve around?

The same ones that people aren't very good at avoiding either?

https://www.driving.co.uk/news/pothole- ... ounds/amp/

SINGLE POTHOLE PAYOUT COST COUNCIL NEARLY £250,000 LAST YEAR, AS TOTAL CLAIMS HIT £8M
RAC attends 1,000 pothole-related breakdowns per month

I have no idea how good any of the AV systems are at detecting potholes. You would be better off asking developers directly if you want to know.

And do they get first dibs on going round a new road layout?

Doubt it. But there's 'digital twin' format which, combined with digital CAD/CAE design of new roads might reduce the requirement.

https://www.pbctoday.co.uk/news/bim-new ... ork/98666/

National Highways’ has set out plans to deliver a digital twin of the UK’s road network that can predict the time and location of potholes and other maintenance issues


“We should strive to replace drawings and static 3D models with dynamic and data-rich digital twins, pdf documents with databases, file exchange with cloud permissions exchange, passive materials with smart materials able to sense and heal themselves and automate all manual routine maintenance."


Part of the design process for new roads is a set of audits. It shouldn't be impossible to incorporate making that, and other design data, immediately available as part of the twin.

Massive datasets are already available and open. For example, there's no reason why route planning (whether human, GPS or AV, shouldn't be informed by (as just one example) scheduled road works:

https://info.westberks.gov.uk/roadworks

(Actually a national map, just zoom out)

It's not perfect, sometimes sites will be booked for longer than necessary to allow for activities than can only be done in dry weather.

Another example is live access to electronic sign information:

http://adamish.com/lab/roads/

A further example of digital twins is to try and reduce cable and other utility strikes.

Picked at random:



This is next one more 'glossy' than detailed for someone digging holes, but there's no reason why (if the information is there) that everything above road shouldn't be peeled away to reveal the services, or the road surface 'marked' with outcomes from surveys which show not just where potholes *are* - but, from predictions based on deterioration, where they *will be*.




I really don't know much about this stuff. I learn oddments from trade magazines, exhibitions, presentations and being inquisitive.

[The Spin Doctor]

The same ones that people aren't very good at avoiding either?

Have you driven recently? Some of the roads in London look like they have been carpet-bombed. It's not a case of failing to avoid potholes, it's simply trying to pick a course that's likely to hit the least worst. I've got a dinged wheel rim on the XJ from hitting a pothole on an unlit road at night.

The A2 was in such a state a couple of years back (the road leading to the Channel Tunnel, remember) that locals were deliberately avoiding it, it was causing so many blown tyres and damaged wheels.

Don't make light of a real infrastructure problem.

[Horse]

.

Don't make light of a real infrastructure problem.

Ah, that'll be why I sent you info the other day on two different proposals for automated surface defect detection?


The point that I obviously failed to make is that you're expecting something from AVs that human drivers apparently can't achieve.

I would guess that, unfortunately, few road users actually bother to report road defects, so delaying repairs. Automation of detection (identifying deterioration before failure) could mean that roads are repaired before potholes appear.

But people will then probably complain about the road works And also complain about any increase in taxation if more is spent on roads maintenance (I doubt they'll want reduced spending elsewhere).


However, not that I particularly have the ear of anyone who could affect roads maintenance, what's your proposal for sorting it?

[slowsider]

"Don't make light of a real infrastructure problem." says the man comparing potholes to carpet-bombing

[The Spin Doctor]

The point that I obviously failed to make is that you're expecting something from AVs that human drivers apparently can't achieve.

And the point I obviously failed to get over is that you are blaming 'failing drivers' for damaging their own cars. No-one does that willingly, and potholes are often unavoidable - it's a case of picking the least worse.

We've had two blowouts, and four tyres damaged that needed replacement from potholes since I've been back in London.

[The Spin Doctor]

"Don't make light of a real infrastructure problem." says the man comparing potholes to carpet-bombing

I doubt you would believe the state of some roads in London.

[slowsider]

"Don't make light of a real infrastructure problem." says the man comparing potholes to carpet-bombing

I doubt you would believe the state of some roads in London.

[Horse]

Your posts:

potholes that drivers have to swerve around?

potholes are often unavoidable

So you want the AV to be aware of, then negotiate, humanly unavoidable potholes?

You're reinventing the 'trolley' problem.

Should AVs be equivalent to average drivers, or better? If better, how much better? How should this equivalence or superiority be assessed?

We've had two blowouts, and four tyres damaged that needed replacement from potholes since I've been back in London.

Talk us through your thinking on approaching those potholes. You're telling me I'm blaming drivers, so explain. At some point, you will have decided stop / go, then chosen a speed. If you can't get it right, how do you expect an AV to achieve it?




Like many other "Ah, but!" challenges, there seems to be an assumption that no-one else has thought of this.

A quick Google finds:

How will driverless cars avoid potholes?
https://www.sciencefocus.com/future-tec ... holes/amp/

Detection of potholes in autonomous vehicle
https://www.researchgate.net/publicatio ... us_vehicle

PotNet: Pothole detection for autonomous vehicle system using convolutional neural network
https://ietresearch.onlinelibrary.wiley ... ell2.12062

A configurable pothole for testing autonomous cars – Part 1
https://qxf2.com/blog/a-configurable-po ... rs-part-1/

UK joins project for autonomous vehicle platform that can find potholes
https://www.autocar.co.uk/car-news/auto ... d-potholes

[Cousin Jack]

Potholes are very easy to detect. The problem is getting them repaired, properly, not by putting a bit of cold tarmac in them and smacking it down with a spade. The local council do this bit regularly, it rarely lasts more than a week and is a total waste of time and resources. It does mean they can tick it off as having been 'repaired'.

[The Spin Doctor]

For f*cks sake...

The question I was asking was how often the street level data gets upgraded, using the example of potholes "that drivers have to swerve around?"

I said absolute f*ck all about the ability or otherwise of the autonomous vehicle to dodge those potholes - you somehow made that jump when you made the ridiculous statement "The same ones that people aren't very good at avoiding either?"

Please stop putting words in my mouth. It's not the first time you've done it, and frankly it's tiresome.

[Horse]

For f*cks sake...

The question I was asking was how often the street level data gets upgraded, using the example of potholes "that drivers have to swerve around?"

I said absolute f*ck all about the ability or otherwise of the autonomous vehicle to dodge those potholes - you somehow made that jump when you made the ridiculous statement "The same ones that people aren't very good at avoiding either?"

Please stop putting words in my mouth. It's not the first time you've done it, and frankly it's tiresome.

I've tried to answer your (and others') questions, politely and often at length. Sometimes putting the anti- side too. But you didn’t ask "how often the street level data gets upgraded", you phrased it in what read to me as a deliberately provocative question:

So they drive every street on a daily basis to check for new potholes that drivers have to swerve around?

And, actually, questioning the abilities (or lack of) AVs to be aware of and avoid potholes would have been a fair challenge, as those links show.

Currently AFAIK there are no AVs on UK roads without a safety driver, so I would presume that person would (or could) choose to take control, at which point the action would be logged for later review.

But you introduced the comparison with human drivers. So I made the comparison that humans aren't always great at identifying and avoiding them either, which you agreed with by giving your personal experiences.

However, it's my fault - when earlier you accused me of 'blaming'.


I've done a lot of answering of questions, perhaps these are worthy of consideration and answer:

- Should AVs be equivalent to average drivers, or better?

- If 'better', how much better, and how should this equivalence or superiority be assessed?

And those are directly relevant to your decisions to drive into or around potholes. Or anything else.

[Horse]

Potholes are very easy to detect. The problem is getting them repaired, properly, not by putting a bit of cold tarmac in them and smacking it down with a spade.

Known in the trade as 'stamp and dash', because it's often done without closing the lane (until just a few years ago, even on motorways).

The local council do this bit regularly, it rarely lasts more than a week and is a total waste of time and resources. It does mean they can tick it off as having been 'repaired'.

Sadly, once the integrity of the top layer is lost, it's possible that the only long-term solution is remove and replace. That's why tar and chippings, horrible as it is, extends the life of the surface.

'My' road was resurfaced properly a few years ago. One section, the location of a persistent pothole, had to be taken down much deeper and built back.

Half way house pothole 'fixes' exist that are better than cold fill.

https://www.devon.gov.uk/news/dragon-pa ... ads/?amp=1

https://www.constructionenquirer.com/20 ... es-faster/

[The Spin Doctor]

Stuff

Sorry Malc, can't be @rsed, other than to say it's stating the bleedin' obvious to say that AVs will be compared with human drivers - they are doing the same effing job!

[Horse]

Ok, fair enough. Just let us know which questions you want answers to and which you don't