passive guides in roads

[YoungStranger]

Road based guides.

"Before the Romans came to Rye, or out to Severn Strode
The rolling English Drunkard made the rolling English Road"

I am currently commuting through the English countryside using various A class and B class road. It is very pleasant but I would not trust any Autonomous system on them... The roads are generally in good condition, but there are lots of blind bends, dips, farm gates before bends etc that I am sure would fool even the best system.

I wonder whether the road surface could be treated economically to create foolproof guideways. For example painiting lines at centre of lane infused with iron filings, to create a magnetic force strong enough to be detected, or using special paint. It would have to be hardwearing, perhaps baked into surface

There maybe a downside, as today when I went into a bend and a dip and encountered at articulated truck coming the othe way and in my lane. I scraped by almost hitting a telephone pole.

What do people think? Should be used in conjunction with other sensors of course

 

[RDoc]

Well, exercising great restraint, I'm not going to comment on the apparent British collective suicide pact when driving on country B roads.

However, from what I've seen in the US, just well painted lines are quite sufficient. I suspect that in conjunction with the high resolution GPS mapping Tesla is supposed to be building, that will be more than enough. The problem is that the roads' lines tend to get worn, particularly in areas undergoing construction.

 

[Pluto]

I wonder whether the road surface could be treated economically to create foolproof guideways. For example painiting lines at centre of lane infused with iron filings, to create a magnetic force strong enough to be detected, or using special paint. It would have to be hardwearing, perhaps baked into surface

Even that wouldn't really be foolproof as another car could be set up to create an electromagnetic field to guide an autonomous car off the lane (ie. if someone wanted to harm someone else or move anywhere in traffic). I think well painted lines really are the best option as there's no way to mess with those without changing the road. Something that could be done as an improvement however would be using special paint that reflects light at other wavelengths (ie. ultraviolet light) so that painted lines could be further validated by cameras even when faded (basically providing redundant data to assist in lane tracking).

The problem is that the roads' lines tend to get worn, particularly in areas undergoing construction.

In some cases lanes aren't even just worn, but changed during construction. When this happens, it makes GPS data invalid and it needs to be identified as such to prevent potentially dangerous situations. And in some cases, the data may only be temporarily invalid (ie. the use of cones to redirect traffic).

There maybe a downside, as today when I went into a bend and a dip and encountered at articulated truck coming the othe way and in my lane. I scraped by almost hitting a telephone pole.

Off-topic: this totally reminded me of the TV show Doc Martin heh.

 

[Ludus]

Good LIDAR based mapping can eliminate the need for painted lines or any other special features. The car's cameras, radar and GPS can match expected features of the route meter by meter and know where to position the car.

If all vehicles are networked there are no surprises from other vehicles and if another vehicle sees a pedestrian or a stray sheep ahead your vehicle gets informed.

 

[Dangerous Fish]

I don't know where you think the money will come from for embedded guides or special painted lines on all our A and B roads. The local councils can't even keep most roads up to a minimum standard these days it seems to me. I don't see the Government funding something like that centrally either.

We have to accept that with any form of automation there will be failures. If the automation applies to transportation, that means there will be collisions, crashes, injuries and deaths. We could spend all that money on special road markings and Tesla can spend another 20 years making their software better, but there will still be accidents caused by hardware/software failure or human error.

The positive part is that the number of accidents are likely to be a lot less than toady.

Musk has already said that, for the moment, LIDAR is not being considered. I expect Autopilot will be refined more and more with hardware and software development and it will be a lot better even by the time the Model 3 is here.

I still doubt I will specify AP on my Model 3 though. I just don't see it being much of a help on the UK roads I drive. On A and B roads, I would rather be completely in control at all times and paying attention every second of the journey. My feeling is that Autopilot use would inevitably lead to my concentration levels dropping and complacency setting in. I know myself too well in that regard!

I'd rather do all the driving and making all the decisions myself and at the moment I'm inclined to say that in difficult conditions or road types I'd probably prefer all the drivers around me to be doing the same. At least until Autopilot has a lot more miles under its belt.

 

[Ludus]

Musk has already said that, for the moment, LIDAR is not being considered. I expect Autopilot will be refined more and more with hardware and software development and it will be a lot better even by the time the Model 3 is here.
.

Tesla won't put LIDAR on their vehicles, but that doesn't mean they won't use LIDAR based high detail and accuracy maps. There are already third party companies with special vehicles in the business of creating those maps.

The Tesla would be comparing it's sensor data from cameras, radar, GPS and ultrasonic to what the map says should be there. Most of the details of the route stay the same. Once a stone wall and a tree are included in a map and their location known to the millimeter by LIDAR, the less precise radar and camera data are enough to accurately position in the known setting.

Upkeep would just be the LIDAR vehicle passing through every few months to refresh the data.

The Tesla style sensors could still work in novel unmapped settings but not with as much confidence and precision.

 

[Dangerous Fish]

Well, we're a long way from highly detailed LIDAR-based maps of every road in the UK and we certainly don't have enough infrastructure or bandwidth to allow cars to be reliably networked (or even to maintain Internet connection during a long journey) so if Tesla want to sell cars here, they will have to use stand-alone onboard technology capable of assessing accurately what's around them.

If they can get the existing Bosch radar to work in the way Musk has hinted in his tweets, that will be a big step forward and may be all that's needed to get the improvements they want in the short term.