Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Highway Code: Watching TV in self-driving cars to be allowed…

This site may earn commission on affiliate links.
If you've ever tried getting out of London on the A13 during evening rush hour, you'll understand why self-driving cars are cloud cuckoo land. They'll wait forever for gaps that aren't there. It takes nerves of steel, and no little courage to make any progress involving lane changing. It's usually experienced give and take that avoids complete gridlock.
 
  • Informative
Reactions: Durzel
If you've ever tried getting out of London on the A13 during evening rush hour, you'll understand why self-driving cars are cloud cuckoo land. They'll wait forever for gaps that aren't there. It takes nerves of steel, and no little courage to make any progress involving lane changing. It's usually experienced give and take that avoids complete gridlock.


on motorways I can see it. Perhaps cars alone will never do it - we may need infrastructure to meet half way. Eg segregated 'robot' lanes which are only allowed if you have XYZ compatible communication systems to talk to other vehicles. Benefits being you can drive a lot faster/closer to others, pre-negotiating merges/junctions with other cars on the same sections of road.

Crappy town centres with potholes and lines all over the place and random people pushing in - may never be meaningfully automised
 
There was an interesting article written about 3 years ago that basically said getting it working on motorways first was the way to go. Less complexity, safer roads statistically, maybe the higher speeds count against it, but arguably higher benefit to drivers. I think the pitch was essentially get level 4 working on motorways so you could work in your car, have a nap, watch TV, and then just take over at the end.


It's easy to to forget there are 3 levels or "self driving" where the car is responsible. The merc version seems to be level 3, the above level 4, and level 5 will be a pipe dream for a long time. I believe Tesla are still firmly in level 2 and haven't yet submitted any level 3 driving data to a regulator. The pilot in the US might be a fairly feature rich version but being 95% good at 95% of driving won't cut it, whereas being 100% good at 80% of driving might have a place.
 
There was an interesting article written about 3 years ago that basically said getting it working on motorways first was the way to go. Less complexity, safer roads statistically, maybe the higher speeds count against it, but arguably higher benefit to drivers. I think the pitch was essentially get level 4 working on motorways so you could work in your car, have a nap, watch TV, and then just take over at the end.


It's easy to to forget there are 3 levels or "self driving" where the car is responsible. The merc version seems to be level 3, the above level 4, and level 5 will be a pipe dream for a long time. I believe Tesla are still firmly in level 2 and haven't yet submitted any level 3 driving data to a regulator. The pilot in the US might be a fairly feature rich version but being 95% good at 95% of driving won't cut it, whereas being 100% good at 80% of driving might have a place.

honestly I'd pay for FSD for that - its the equivalent of a train journey and I don't mind driving for 15 mins either end. I don't need that edge case particularly solved
 
If you've ever tried getting out of London on the A13 during evening rush hour, you'll understand why self-driving cars are cloud cuckoo land. They'll wait forever for gaps that aren't there. It takes nerves of steel, and no little courage to make any progress involving lane changing. It's usually experienced give and take that avoids complete gridlock.
I think the logic is that once all (or most) cars are self-driving they'll somehow communicate with eachother and "agree terms" on merging and gaps and such, so confidence or nerves won't be a factor. Obviously it's a chicken and egg problem while the vast majority of cars on the road aren't self-driving in the slightest.

I think it could easily be another 10, maybe even 15 years, or more before things progress to a point where you can actually start to consider this stuff as viable though.
 
This would be fairly straight forward for Tesla to implement given the limited circumstances you could use it.

1. Consumer buys Tesla insurance.
2. <37 mph on a motorway, turn off the nag, enable TV content, enable big red take-over now klaxon.

Tesla’s are already registered cars but they might need approval by the Secretary of State for Transport under AEVA (The Automated and Electric Vehicles Act 2018). The question might be - are people (voters) ready? Are Brits comfortable with self-driving cars? | YouGov
 
After seeing a guy posting a video of Smart Summon over here on a car that has had the UNECE restrictions turned off, I wish I had a button where I could choose to accept liability rather than a monolithic organisation that takes years to do anything deciding for me. I understand that it's necessary for homologation, and that the UK is actually one of the more vociferous opponents of relaxing rules, but even so... it's annoying.

I'm mostly bitter that I bought FSD in 2020 when I got my car and essentially nothing has been developed on it in that time. My car stops for traffic lights on AP, and NoA has worked well when I've used it (but EAP has that anyway), but everything else is useless.
 
This is exactly what I'd expect to roll out to owners of cars with Enhanced Autopilot. Lazy driving in traffic jams, maybe an occasional lane change, and a prod to take back control if it gets stuck.

The only bit that worries me are HGVs with large pieces of metal overhanging the back of the trailer, as the usual tattered hi-viz vest that's wrapped around the pointy bit isn't seen by the camera or the radar.

I actually had a surprisingly peaceful drive up the M1 today. 220 miles on AP and only one phantom brake when a van towing a trailer veered towards my car, which I thought was forgivable.
 
honestly I'd pay for FSD for that - its the equivalent of a train journey and I don't mind driving for 15 mins either end. I don't need that edge case particularly solved
I agree. I don't do the miles I used to, but we drove to Cumbria recently, only 3 hours of motorways although did include a really slow stretch around Manchester, and if I could have just let the car do the motorways it would have been fantastic. I live down a country lane and it will be a long time before a self driving car manages it, a fair few drivers struggle plus you're constantly dodging potholes.

The only advantage to the "last mile" being self driving, is the trip back from the pub. Luckily mine is only a short walk.
 
I'm mostly bitter that I bought FSD in 2020 when I got my car and essentially nothing has been developed on it in that time. My car stops for traffic lights on AP, and NoA has worked well when I've used it (but EAP has that anyway), but everything else is useless.

It's very frustrating.

Even in the USA, where the progress is, it still drives rather like a learner driver. The car isn't learning as quickly as a human, despite Tesla's AI claims.

It will happen, but often with technology there is an initial breakthrough which solves many problems but becomes over hyped until reality sets in, then it follows a more gradual improvement.

I think Elon Musk imagined this working but his imagination was limited to the world he knew, not the global complexity of driving in every situation. If you imagine lovely wide quiet grid roads or highways, the problem is simpler. When you imagine a busy school run with narrow streets, speed bumps, pot holes, some rural roads, other impatient drivers, random parked cars and lots of traffic congestion, it's a very different complexity.

I'll stick my neck out and say that outside of A roads and motorways, it's still at least 10 years away.
 
4 miles of single track hills and blind bends, here. Perhaps I need to trade my Tesla for a horse and cart with a mattress in the back.
I think you can be done for drunk in charge of a horse, I certainly think you can on a cycle.

The point being suggest though is prioritisation. Do you try and delivery it all and go all out for level 5, or if you're accepting that a sensible stepping stone is to nail level 4 for certain roads, and then move onto more edge cases then a logical progression would be motorways, then dual carriageways, then major trunk roads - all of which have higher standards (ie exit and entry is more tightly controlled, roads are minimum width, no stopping, etc). You'll know as well as I do that driving a country lane often results in nods, waves, flashes of headlights and the odd bit of reversing and accepting that your right of way might be trumped by the other vehicles type - it's unlikely you'd ever force a tractor with trailer to reverse more than a few meters even if it was your right of way, and these situations occur. Maybe our roads are very different to the US and so they don't have our peculiarities, which just means it;'s going to be even longer before we get level 5 here even if they do crack it in the US because we as a nation becomes the edge case, not some of the roads we drive upon.
 
It's very frustrating.

Even in the USA, where the progress is, it still drives rather like a learner driver. The car isn't learning as quickly as a human, despite Tesla's AI claims.

It will happen, but often with technology there is an initial breakthrough which solves many problems but becomes over hyped until reality sets in, then it follows a more gradual improvement.

I think Elon Musk imagined this working but his imagination was limited to the world he knew, not the global complexity of driving in every situation. If you imagine lovely wide quiet grid roads or highways, the problem is simpler. When you imagine a busy school run with narrow streets, speed bumps, pot holes, some rural roads, other impatient drivers, random parked cars and lots of traffic congestion, it's a very different complexity.

I'll stick my neck out and say that outside of A roads and motorways, it's still at least 10 years away.
Your car doesn't learn, that's not what AI means in this context. I don't believe the version of AutoPilot we have in the UK has had any significant new training since 2019, and as per what we've been told at various Tesla AI events all their focus shifted to starting again on a new model to eventually replace but also have the ability to do city streets.

Elon has said we will see something new soon as the AP team move the new model to cover motorway driving, but broken clocks are more accurate than his timelines.

Personally I maybe agree with your timeline, but I don't really see it as an issue. Already FSD/EAP provide a motorway driving experience I value, and it will get better with the new models at some point. My benefit to have the car drive me to the shops is low, the benefit of having the car drive me all the way up the M6 M74 to Glasgow is high. I paid £4600 for FSD in 2019, I'm happy with that purchase, I'm getting a new Tesla in June (hopefully) and will pay £3600 for EAP.
 
  • Like
Reactions: Durzel
Your car doesn't learn, that's not what AI means in this context. I don't believe the version of AutoPilot we have in the UK has had any significant new training since 2019, and as per what we've been told at various Tesla AI events all their focus shifted to starting again on a new model to eventually replace but also have the ability to do city streets.

I never said the car did learn, which is actually part of the problem as the future is federated learning. The learning is done back at Tesla, the model is constantly retrained from ever increasing datasets.

If there is no learning anywhere, it's not machine learning.

Plus I was not talking about autopilot, I was responding to FSD.