Self-Driving Ubers Driving in Bike Lanes | Hacker News Reader

Self-Driving Ubers Driving in Bike Lanes

driverlessratings.com

108 pointsdriverlessnick119 comments

This is, in fact, correct behavior: http://www.sfbike.org/news/bike-lanes-and-right-turns/

If turning traffic doesn't enter the bike lane, then it is turning right across the bike lane, which can lead to accidents where a bike collides with a turning vehicle. As such the correct behavior is to signal and enter the bike lane before turning. Of course, many human drivers fail to do this, so it's not surprising that people might be taken aback when they see self driving cars doing it...

yep,

https://www.dmv.ca.gov/portal/dmv/detail/pubs/hdbk/shr_slow_... ...Merge toward the curb or into the bike lane only when it is safe. Merge safely behind a bicyclist when preparing to make a turn. Only enter a bike lane no more than 200 feet before starting a turn. ...

The article is not about San Francisco, it's about Pittsburgh

Since it appears that enough people are confused about this, maybe the markings on the road should clarify the correct behavior (otherwise the danger is raised for everyone). Something like a turn lane painted at the intersection, with the bike lane having a yield sign just before the turn lane.

They do have those in San Francisco. Bike lane turns into dashed lines at the intersection and the right lane of traffic expands into the bike lane as the intersection gets closer.

This is also the case in southern California.

This is not the case on the East coast. The bike lane line is solid all the way to the intersection.

It is the case in Hartford, CT.

I lived in Downtown Hartford until last month, I don't recall many, if any bikes on roads. Then again, I do recall traffic being generally unsafe, I definitely would avoid riding a bike there.

I'm near the West Hartford border, around the northeast greenway bike path. It's got the markings for cars to merge into the bike lane on right turns.

I do remember those bike paths, I also forgot about the part of the greenway that ran through Downtown according to signs, and apparently the Traveler's building is at the base of part of it (I lived next to that)?

I also commonly see bike lanes go dashed and a right turn lane appear to the right of them.

I always thought this was "basic driving" knowledge that some people just don't get. Is the solution really to add more markings and signs to the street? Do you need "Left lane is for passing" along hundreds of miles of a highway in California?

Since you appear to be californian, do the bicycle lanes over there have solid lines or dotted lines?

Solid lines until you get near an intersection, where they usually change to dotted lines to give a hint that you are required to merge into the bike lane before turning.

this, all over the bay area and from what i remember of socal.

apparently they do!

Btw, California doesn't use the left lane for passing; you just have to yield to anyone going faster than you.

Most states have keep right laws[1]. Some are definitely stronger than others, but these laws aren't enforced as frequently, though I wish they were.

[1] http://www.mit.edu/%7Ejfc/right.html

tldr: If someone behind you is going faster and you are in the left lane, move right.

They are often painted with a dashed line before the intersection.

This law apparently varies between states. This is in PA rather than CA. Though apparently people are complaining about it in CA as well, where it definitely is the correct behavior.

Actually, the complaint in California is in the guardian article linked and has Uber doing the turn illegaly in CA.

That said, the guardian article is a year old, while the Pittsburgh complaint is from today and describes legal behavior in CA.

Basically, Uber probably changed their software's behavior to follow California law following the complaints last year and that is now resulting in complaints in Pittsburgh.

I have no clue what PA law is on this, but I can imagine the law may differ depending on jurisdictions and this would be something where engineers will have to have a variety of behaviors depending on the applicable law.

I have suggested many times that the things computers are good at include looking up inconsistent rules in a database, so this doesn't seem as if it should be an either-or issue for the car, it should know all the rules and be able to follow them as appropriate, where a human might be forgetful.

And probably why Uber gets it wrong. Bunch of engineers thinking their laws are the laws everywhere. It's good they're figuring this out now before self driving cards become really common as road laws vary so much even between cities and time of day...

This would definitely not be legal in the UK. Is it a US nationwide rule?

It is a pretty common state level rule, although one that most drivers don't know about.

It's a state law. Threw me for a loop when I moved from Boston (where this behavior is illegal) to California (where not doing this is illegal), but it's in the CA driving handbook and on the driving test to get your CA license.

Only in California, apparently.

I was curious and checked the Dutch regulations - in NL it's an infraction to 'wait' on a cycle lane for any reason. Cars turning across the path just give way to all traffic coming from behind until it's safe to move. It is very stressful for the driver, but blocking the cycle path is a recipe for chaos.

"blocking the cycle path is a recipe for chaos" I think california has the right rule here, the driver cannot constantly monitor what's coming from behind when making a turn, especially when most cyclists are almost invisible from side mirrors. Blocking the path for a safety-critical maneuver would certainly help.

This is how we do it in NL. We constantly monitor what's behind and to the side of us.

and for having lived in Amsterdam for 2 years now, I guarantee that the way bikes (and traffic in general) are handled in the Netherlands is far from the safest way! :)

But living here myself, I can attest that Dutch drivers are better than US drivers.

Perhaps because of better drivers and possibly better cyclists, the chaos that is rush hour cycling results in far fewer accidents than I would expect.

agreed that in general Europe has better drivers (license is just harder to get) but that doesn't change that it's still very unsafe and not secure :)

One thing that helps is the increasing number of non-car roads (pedestrians and cyclists only). Obviously that's not very friendly to drivers, but a well designed city with good public transport really doesn't need cars in the city.

Do you also stay still while doing this, or do you have to monitor what’s in front as well? If not you’d effectively have to monitor three separate directions all at once.

Seems to me it should be safer to block the cycle lane to avoid having traffic pass on the right, so you really only have to monitor in front and to the side when making the turn. Cyclists behind will then simply either have to adjust their speed, or decide whether to cross into the car lane and overtake on the left.

This thread made me realize I don’t actually know the rules for this here in Sweden. I’ve inquired with traffic authorities, because unfortunately when searching on their site for cycle lane rules you only find a picture of what cycle lanes look like. :o(

In Copenhagen, often you have a green light, so you can concentrate on behind. Then check ahead when behind is clear. (You are usually stopped, and if not, no driver behind will be surprised or angry by your stopping to ensure the path is clear.)

Don't expect the throughput of an American city road. It's not unusual that only one or two cars can make the turn before the light is red again, if there are lots of bikes.

This is not true for right turns in the US, when doing a right turn, there could be pedestrians crossing the street as you turn.

> Cyclists behind will then simply either have to adjust their speed, or decide whether to cross into the car lane and overtake on the left.

Here's a third option to consider: the cyclist does not notice your merging in to the lane or otherwise fails to react and smashes into the back of your car. At higher speeds this can be lethal.

> At higher speeds this can be lethal.

Good thing this is done at low speeds. The car is basically going 0mph, because they are going to turn 90 degrees. And when I'm biking, I am BY FAR more cautious when approaching intersections. As you should be in a car, bike, motorcycle, or whatever.

The merge – or crossing if you will – is going to happen at some point anyway. The point of driving in the cycle lane would be to do the merge first, well in advance of the turn, and once you reach the turn you can safely proceed with the knowledge that you won’t miss a cyclist coming in from behind. (Probably.)

If you don’t merge first, you have to not only consider cyclists in the lane but also whatever may be involved with the turn itself, such as people crossing the street you’re turning into, or other obstacles.

Basically you spread the risk over time and space instead of concentrating it all into this one maneuver at the end.

Well said, sir. You said it better than I could.

and I'm pretty sure that one of the touted benefits of self driving cars is that they are able to constantly monitor a whole lot more than a human driver would.

But that means cyclists must either wait, which we know won't happen, or move into the faster moving traffic to overtake.

It's much better for the car to have to wait. It doesn't seem to be a problem in the Netherlands or Denmark.

It's true, you have to look very carefully before making the turn, and to make the turn slowly.

In Copenhagen, the busy roads have a separate light for cars turning right vs bikes going ahead.

Yup. It also:

- Is a signal that the driver intends to turn. Drivers should also be using their actual turn signals, of course, but as a cyclist I would much rather have a driver who fails to signal merge early and turn, rather than turn with no warning. - Allows the cyclist to more easily get around to the left.

> Allows the cyclist to more easily get around to the left.

And be ran over by a bus. I don’t see how this rule makes anything safer unless cycle traffic completely stops to wait behind the car (not happening).

You can simply slow down, if you are not sure about the safety of this alternative.

> the driver cannot constantly monitor what's coming from behind when making a turn

Correct, which is why the common style of crossing is pretty much a death trap for cyclists. A much safer design moves the cyclist crossing so cyclists and cars only ever meet at 90 degree angles[0]. Doing so takes no extra space.

[0] http://1.bp.blogspot.com/-6Kg0XZ0K-3U/Uv98VQ-H6lI/AAAAAAAARK...

What is functionally different about that illustration other than the cyclist having to make odd movements when going straight - which would be incredibly confusing to oncoming cars turning left?

A car turning right (or left) only crosses the green cycle lane after it's rotated. Rather than having to watch for bicycles on the right coming up from behind, the car only has to watch for bicycles in front coming from the right, which is much easier.

the "odd movement" isn't something a car driver has to deal with at all. As a driver you turn right, and after completing the turn you're faced with a cycle path where a simple look to the right will be plenty to check for cyclists. It's basically splitting a complex intersection into two simpler intersections.

https://imgur.com/a/H0tEZ

This is what I mean. If in a car you want to turn left, the cyclist appears to be heading right, by then veers to stay straight. Otherwise, you need dedicated turning left lanes, with turning arrows - something most cities can't fit on existing streets.

These sorts of "reimagined" street intersections always end up being impractical with only a little bit of scrutiny applied. What's practical (from the point of view of someone who's biked to work every day in an urban environment for 15 years) is the bike lane turning into "shared" before the intersection, and the solid white bike lane line turning dotted with indications marking that before the intersection. Every attempt at enforced separation just ends up being less safe for all parties involved.

The key is sensible lane sharing and maintaining a common feel and look of intersections throughout a city. No cyclist, or driver needs to question what to do at various experimental intersections, and yet we continue to do things like that in urban centers using people as test subjects.

You're imagining it happening at an unreal pace. This is really close to what most large intersections in the Netherlands look like. The shoulders for the cycle lane are actually behind the curb, and for a cyclist still feel like 'going straight'. If I'm not mistaken, this is even known as a "dutch junction".

This video explains it well: https://www.youtube.com/watch?v=1bG6ZrbCO2g

No vehicle in that video was shown turning left.

That's why they invented hand signals. Granted not everyone is using those when they should, but it's no different from someone making an unexpected turn on a normal crossing.

By the time a turning car and a bike intersect paths, the bike is moving across the front of the car, as opposed to approaching from behind as the car is turning. It would be far easier to spot a cyclist moving side-to-side across your vision out the front of the vehicle than trying to spot one in your blind spot.

This of course assumes drivers follow the marking and don't just cut across the marked bike lines, in which case it would negate any advantage of this layout.

Seattle solves the problem by having separate car and bike lights. Bikes get a preemptive green for about 15 seconds then cars have the right of way.

We have an intersection somewhat like that in SF at 9th and Division St and as a cyclist it feels more dangerous than other intersections. A lot of cars block the bike path and a lot of cars think they have the right of way turning right vs me going straight.

Where I learned driving, the driver is definitely required to monitor what's coming from behind when making a turn. The mantra is "mirror, indicator, shoulder":

1. look in the mirrors

2. put on the indicator (turn signal)

3. just before turning, look over your shoulder to make sure you're not running over anyone

However, given how lazily and inconsiderately many people drive, the annoying but defensive rule you describe might be the better tradeoff: make life worse for all cyclists to safe a few lives.

But at the end of the day who is wrong in a situation where a car signals to turn into the bike lane ahead of a biker (with plenty of room, it's safe) and then signals again to turn right at an intersection if the bike decides to go around the car on the right anyways and gets hit by it?

I view the issue people are having with Uber's Self Driving Cars not "handling right turns in bike lanes well" in the same light as the reason you take a left hand turn at an intersection from the left-most lane. If you tried to take a left from the right most lane it puts everyone to your left at danger and definitely doesn't let people coming towards you know what you are doing.

I feel the same exact thing should apply to the right lanes and bike lanes regardless if it is a car or a bike. To turn right you get as far right as you can, you check your mirrors sure, but you don't really let what anyone else is doing behind you change how you drive. I view worrying about driving in that way is possibly evidence that you are the scared, defensive-type driver. Hesitation can be a killer.

But there's a evolutionary problem here. You don't have eyes in the front and back of your head. Bikes are small and hard to miss. They can easily slip into your blind spot while you are looking at the road and switching to look in the mirror.

You're right, a driver is definitely required to monitor what's coming from behind them when making a turn. But when we consider limits in eyesight, moving into the lane keeps everyone safer.

You are assuming a fully observable environment - where everything is visible. To be honest with you, many bikers are not, or not from every angle possible. Most of the bikes I have encountered do not follow the road signs either (especially lights & stop signs - they just breeze through). With this in mind, given cyclists are the ones in danger (As opposed to the driver sitting in his car), it's safer to give the cyclist a clear sign of a right turn. I have seen way too many times where a cyclist used his judgement to pass through a car in what i call rather dangerous way, so I rather block their path to save their life than to trust them to use their best judgement. Cyclists are often invisible, and more susceptible to injuries, and I am being cautious and not leaving any doubt about when and how I will turn, by blocking the path. When I give a right turn sign, you just cannot pass through me assuming i won't be turning in the next second.

California allows right-turn-on-red and that might factor into it.

This is also allowed at some crossings in The Netherlands. Typically on crossings with a traffic light cars are not allowed to "pre-sort" on the bicycle lane though.

Most of the Northeast U.S. allows right-on-red but still doesn't let you drive in the bike lane.

I think it may be the preponderance of 3-lane boulevards and wide bike lanes in California (and west of the Mississippi, in general). On such streets, you technically can fit a fourth turning lane on the shoulder (where the bike lane is), which means cars turning right won't be blocked by traffic waiting at the light. And bike traffic tends to be pretty light on those 3-lane boulevards, just because it's very unpleasant to bike with 3 lanes of cars going 45 mph.

By contrast, virtually all the roads in Massachusetts are single-lane, where you gain nothing by letting cars enter the bike lane because there isn't enough room for them to pass the cars in front of them anyway.

> Most of the Northeast U.S. allows right-on-red

Is this true? The largest city in the Northeast (NYC) does not have right on red unless posted.

NY state does have turn-on-red, just not within NYC city limits.

It's really just NYC that forbids it:

https://en.wikipedia.org/wiki/Turn_on_red#North_America

(I was a bit surprised to learn that all 50 states permit it, and they lose federal funding if they don't. Other than NYC, I've never encountered a place that doesn't, but I'd thought it was really a state-by-state thing.)

It is not even state by state. It is city by city and most cities in the north east either ban right turns on red entirely, or post a sign banning it at every intersection.

I was under the impression that NYC is the exception and not the rule. I'm from Minnesota and was unaware that anyplace disallowed right-on-red until I met my college roommate, who is Brooklyn.

https://en.wikipedia.org/wiki/Turn_on_red

AFAIK, right turn on red is universal in the US EXCEPT NYC and anywhere that it's explicitly posted as not allowed. Certainly as a driver I'm not aware of anyplace else where I wouldn't right turn on red absent a specific prohibition.

Yet cars are allowed to drive on the cycle lane (if it has an interrupted line) before turning. If they are forced to stop (e.g. when a pedestrian crosses), even though they are in violation, I doubt that they would be fined.

It's not just California; it's true of at least some other US states as well. Think of the bike lane as just another lane of traffic.

And Texas.

I think the difference is that in the US you are allowed to turn right (in many cases) on red. Here in Germany it generally isn't allowed, so bikes have right of way over traffic turning / crossing over the bike lane

Is Berlin an exception in this regard? When I first move here, some years ago, I almost got ran over on a handful of occasions, before I noticed that drivers are actually allowed to do right turns on red lights.

It's only allowed at some intersections. There's a sign that allows you to do it, an unlit green arrow [1]. This sign is pretty much exclusive to eastern Germany (it stems from the former GDR); you'll almost never see it in the western parts.

[1]: https://www.thelocal.de/userdata/images/article/2a9465965ec3...

They have them in a few intersections here in Munich too

This is an interesting point, and may be a clear distinction in why people are so definitive on one answer or the other (not understanding the variables involved). Is it common in Europe that you can't turn on red? Everywhere I've been in the US turn on red is encouraged, though there are a few exceptions where signs are posted to prohibit the normal behavior.

>Is it common in Europe that you can't turn on red?

Yes it's almost always the case.

Our intersections are designed a bit different from yours. In ours the stoplights are placed a bit before the intersection, compared to your lights in the intersection, or right after it. This doesn't invite to right-on-red as you do not have that clear of an overview of the incoming traffic from left.

This isn't legal in most states. Outside of California, you're usually supposed to turn directly from the car lane.

Also illegal in Europe.

I thought one _always_ had to merge right before turning right --else ticket for turning from wrong lane (even if it's the only lane in that direction). I guess the fear is contributing to a scenario which sets-up a T-bone. I believe this was the case before people even thought about considering bikes. I'm surprised it's not the same in other states. The again, we don't have the Boston left turn rule.

Shower thought: in theory, if driverless cars completely replaced humans on the roads, then all the regional and international variations in road rules could be unified with a software update.

If driverless cars completely replaced humans, then you would not need many of the rules, road signs, traffic lights, and so on. Car-to-car communication combined with mesh-sensor network would probably take care of most of these.

Cyclists and non-driverless car-drivers still need to know what to expect of other vehicles.

The keyword from your parent's comment is "completely".

"Completely" is never happening, or is so far in the future to be pointless to speculate about (especially for bicycles). The foreseeable future will involve a blend of automatic and manual pilots.

The second keyword is "if".

Sure, this is going to be the case for long time. But it is still not hard to see many of the road rules chaging/evolving, and many rules becoming obsolete.

Those driverless bicycles will surely become a hit.

It's not that crazy an idea. High-end bicycles already have electrical components (e.g. gear shifting) and electric-assist bikes are becoming much more common. Automatic steering/braking/etc. would take a lot of stress out of my commute.

You can have a useable bike for a hundred Euro or so. It will be a long time, if ever, that we self-driving technology becomes cheap enough and small enough to compete. If you want a small, self-driving vehicle, you might as well put four wheels and a roof on it. Cars don't have to weigh a literal ton.

> You can have a useable bike for a hundred Euro or so.

Sure; you can buy a usable car much cheaper than current flagship models too. I spent over 4k on my bike; the high-end market does exist (and is currently the only place you'll find electric shifting, but all these technologies gradually work their way down).

> If you want a small, self-driving vehicle, you might as well put four wheels and a roof on it. Cars don't have to weigh a literal ton.

A motor and batteries or fuel tank are always going to be pretty heavy though, much heavier than sensors and processors. In any case, I like the exercise and the visibility/openness. I'd expect self-driving scooters/mopeds before self-driving bikes, sure, but I'd expect them to happen eventually.

Cyclists shouldn't be on the road in the first place. They're a non-powered vehicle and as such are pedestrians same as someone on a skateboard. They either need to be on sidewalks, or have their own separate paths (that are physically separate from roads which are for powered vehicles only). As for non-driverless cars (driven cars?) that's a more complicated problem, but as they become increasingly less common they should eventually get their own dedicated lanes to keep them out of the way of the driverless cars. Eventually we'll likely see a bifurcated system where driverless cars operate at much higher speeds that they can safely do (say 100+ mph in some areas), while manually driven cars are restricted to speeds more appropriate for human reflexes and awareness, say something averaging about 50 mph with some sections getting up to 80 mph or so.

Having separated, convenient pathways for every mode of transportation sounds like a nice utopia, but we struggle to build that kind of infrastructure today.

Until that utopia exists, bikes have to ride somewhere, and that's the road. Sidewalks are (typically) far more dangerous for cyclists due to short sightlines out of driveways and alleys. And driverless cars need to be able to recognize cyclists anyway.

> Eventually we'll likely see a bifurcated system where driverless cars operate at much higher speeds that they can safely do (say 100+ mph in some areas)

I agree that bicycles probably shouldn't ride in traffic on the freeways, although often they are allowed to ride on the shoulder of freeways today (at least in rural areas).

When I lived in the Bay Area it always annoyed me how people turned from the middle of the street.

It's also pretty much common sense. In a car, or on a bike, sometimes something ahead of you makes you slow down. The problem comes in when the bicycles are ignored, and "right hooked" by cars turning through the bike lane when the cyclist needs to go straight. In Chicago at least, that move is illegal (passing a bicycle then turning right is as well, but let's see that ever get enforced)

My first cycling accident involving a collision with a car was in Berkeley and involved a car that turn right across the bike lane in front of me and I smacked right into their rear quarter panel.

Apparently about a year ago they were getting complaints about exactly the opposite behavior: https://www.theregister.co.uk/2016/12/19/uber_selfdriving_ca...

They can't seem to win with this one.

Given that different states have different laws, and in both cases the Uber car was violating the local state law, ... no, they aren't winning.

Aren't you supposed to do that? It gives the bikers heads up to slow or navigate to the left of the vehicle turning right. Otherwise the chance for collision is higher when the vehicle suddenly cuts off the bikers during its turn.

Edit: image from sf bicycle coalition. http://www.sfbike.org/wp-content/uploads/2013/02/Right-Turn....

As many others pointed out, this is the correct behaviour.

As a driver, I make sure to leave no space between my car and the curb when I intend to turn right. This reduces the chances of a cyclist sneaking through on the right.

As a cyclist, I make sure to pass vehicles on their left, whether they are turning right, or stuck in slow traffic. It is safer, because very few people expect to be passed on the right.

(locale settings: North America)

The behaviour you describe is AFAIR illegal in Germany. Bicycle lane is exactly that, just another lane. And it has right of way. If you need to turn, you wait until you're free to do it.

It seems strange to me to require that people turn from an outside lane across a lane that goes straight. As a cyclist, this is one of my biggest constant fears. I think I feel safer if drivers have to merge into my lane first, then turn from my lane. Unfortunately in practice where I live in the US, most drivers aren't safe and don't look no matter which way they do it.

Oh, I meant roads in general, sorry.

This is correct behavior in lots of places. It's safer for when turning right to have a biker wait for you to turn. They have no right to "go first" unless they are already in the lane beside you or in front.

I have been a regular biker in a few major cities. Frankly the only rule I was aware of was "try not to die, try not kill anyone".

I'm not sure most people know what the correct action is with a bike lane in the right-most lane. On a bike, I assume I will go left of the right-turning car. In a car, I usually assume the bike will just have right-of-way and I wait.

I could use a clarification about expected driver behavior when turning right in the presence of a bike lane.

And I need to acknowledge my internal reaction of judging this as one more “Uber-villain” style rule. Uner-corrupted, Uber-unscrupulous, uber-villainous to be more precise.

Their brand image is firmly established for me.

Yes that brand name is quite obnoxious. Airbnb might ignore just as many laws, but at least they have a goofy unpretentious name. The only thing I can think of like "Uber" except worse might be "Darth".

I was referring about the brand established by the criminal behavior of Uber I read about on regular basis.

What a company says about themselves is PR. How the company is perceived is Brand.

I saw Googles self-driving car (Waymo) doing exactly that in front of me. It used the bike lane for a while before turning right. (That was in Mountain View) Thought thats odd.

Not being prepared for different rules in different states seems like a big miss.

cracks fingers, leans back, sighs back to hating on Uber. Everything is in it's right place

try living in a city with good, fully separated, and wide bike paths.. (copenhagen)

some times even normal cars take the bike path - https://www.instagram.com/p/g590OCMlGe/ - the car has driven a good 300 meters from one intersection to the next..

This is a great time to reflect. Self driving machines thought it made sense to drive in bike lanes.

Who gets a ticket or their driving license suspended for self driving vehicle traffic violations?

"In other news, people breath air".

Some of these comments arguing that it's 'correct' behavior begs the question, do we want 'correct' or 'predictably expect-able' behavior. What i mean by that is most people would likely not expect a car to turn into that lane, even though it is 'correct'. Perhaps instead of following the rules so rigidly it should SOMETIMES break the rules to do the thing more like a human would, making it an all together safer system as humans would better be able to predict it's actions and adjust accordingly.

This of course imposes lots of philosophical problems like which rules can be broken, and which not.

I'd think in this case that ideally what we would want was for the Uber's behavior to become normalized among non-robot drivers, and become 'predictably expect-able.' What the Uber is doing is actually correct rather than 'correct.'

What really should happen is for humans to be taught about the correct behavior. This is a rule that should be enforced rigidly.

Turning across bike lanes instead of from them endangers the lives of cyclists. Just because most drivers don't care whether cyclists live or die doesn't mean that it is the right thing to do.

Just because it's the expected behavior in california doesn't make it correct or right in the rest of the world.

The other comments pointed out that california have dashed lines for bicycle lanes around intersections. That may be the signal.

Reply on news.ycombinator.com