Hacker News Reader

Comment by netsharc | Hacker News Reader

An interesting "red but no traffic" phenomenon is in Iceland, with roads which are next to scenery viewing points. The roads are mostly empty in this country, but everyone must've slowed down when they saw the breathtaking scenery, and would pull into the viewing point's parking spot to stop and take pictures. On Google Maps these areas were marked red.

On another navigation related topic; I was using Waze in a high-traffic city, where even the alternate routes have a lot of traffic. I was thinking, it surely would be possible for Waze to look back along my route and calculate which route would have been better for me.

E.g. if I drove from A via B, C, D, E towards Z, and there was an alternate route A-B-F-G-E-Z, and the route branches out at B (you can either take path B-C-D-E or path B-F-G-E), the app could find a car that was near me at point B, and went to F later on. It would then know the time I would have needed to drive from B to F. This ghost car doesn't need to go all the way to E, if it turned off at F, the app could find another car, one which at that point was at point F and later on drove to G (or better, E), and then see how much time they took. And so on for each segment of my alternate route...

View on HN

So I'm stuck in a traffic jam, and google maps is telling me I can take an alternate route to avoid it and save 25 minutes vs waiting in the traffic jam. Something that occurs to me as I sit in this traffic...

what portion of all the other driver's around me also looking at google maps? Is Google Maps telling them to take the same alternate route as me? Are they all going to do it? If they all do it, does the alternate route (which likely can't handle as much traffic as the original route) just became as jammed as waiting in traffic? Or even worse, if they're all going to do what google maps says, would I in fact be much better off ignoring it and sticking to the original route?

It's hard to tell what's imagination/selection bias, but following google maps traffic-jam-avoidance suggestions has seemed disastrous enough to me enough times that I mostly just stick to the original route now. (Of course, one could also imagine using one's own route knowledge to pick an alternate route of one's own without google maps...)

I suppose google maps could be smart enough to tell half the people to go one way and half the other... I kind of doubt it is, but one could imagine a computerized route direction system which, if enough driver's had it, could actually maximize efficiency by sending different drivers different routes intentionally... I'm not holding my breath for it.

Well the other route presumably has a significant throughput. If 1000 car/min are in traffic, and the side-road has a 200 car/min capacity, the first 20% of people to leave the main road will experience no additional traffic.

Now, let's say that the 200 car/min turns to a bottleneck of 100 car/min if it's overwhelmed 10 mins downstream of the turnoff. If google routes >200 car/min to the turnoff, there will eventually be a significant traffic jam: cars will pile up at the bottleneck, and even if waze stops routing to the turnoff at that point, the jam would still travel upwards like a pressure wave until it hits the turnoff.

I'd have a hard time believing it's possible for waze to know beforehand what those throughputs are, so some situations like this must occur.

These situations absolutely occur. Something like this happened to me in Alexandria, VA, trying to cross the Woodrow Wilson bridge into MD. I-495 was slow across the bridge. Waze re-directed me off the interstate into residential side streets, along with thousands of other travelers. The side-streets were every bit as jammed as the highway. It took me 3 hours to disentangle myself from that mess. Was it faster than staying on the highway? No idea, but 3 hours to cover 5 miles is terrible - the app should have just said "You're fucked, go home and try again tomorrow." instead of constantly telling me "oh hey, try this other side street, it'll save you 45 minutes!"

I would pay good money for the "go home and try again tomorrow" feature. Especially if it was narrated by my favorite celebrity. Like Cookie Monster.

You can get cookie monster's voice on Waze...

Oh, I'm aware. Thanks though. ;-)

Man, now imagine how the people who live on those residential side streets feel...

It depends on the city. In some cities, Googles routing takes into account the number of other users Google has directed down the road.

Presumably, this alternate route won't save you 25 minutes by the time you and others reroute. However if it saves you 10 minutes and reduces the delay on the original route by 10 minutes since there are now fewer cars idling, then everyone is better off.

Consider another example. Google suggests an alternate route that will presumably save 15 minutes.

Let's say enough cars take the suggestion to slow down that route by 10 minutes and speed up this one by 10 minutes. Even if you end up saving five minutes relative to the original estimate, you'd have saved ten minutes staying put.

It is a point of contention if Google should optimize for saving time for the individual user, (using data only available to that user), or for that user (but using routing information of other google users), or for all users on average (but potentially to the detriment of some users to give other users a bigger advantage), or for the public on average (ie. routing cars to reduce traffic, but possibly putting google users at a disadvantage compared to non-google users).

Various cities round the world have different optimization goals as they try out different approaches.

I would say obviously the answer is that users paying for premium levels of service should get faster routes, isn't that how this sort of thing is done?

Two weeks ago in Atlanta, I was dropping my wife to the Airport. Atlanta traffic was doing it's thing (not moving). Google said it found alternate route that will save 7 minutes. I was like meh. Then it said 10 minutes so I took the turn. I don't know how many people did the same but I ended up spending half an hour extra in some internal road and half of it was being stuck in a right turn lane that rejoined the original highway I was on. So if I had kept on the slow moving highway I would have definitely reached much earlier than the alternate route. Bottom line is that the 'Save time' works only when the alternate route is truly alternate and not just a detour and rejoin the same highway as you will get stuck on the ramp while joining back.

My third hand information frmo a coworker who is friends with people on the maps team is that some percentage of people will get routed into the traffic, and some percentage will get routed to alternate routes, of which there may be a few. Google appears to take a rather active stance on increasing total traffic flow, even at the expense of giving some number of drivers suboptimal routes.

I've noticed this too, and as a result I cannot trust Google maps.

Is there any competitor I can check to confirm best routes?

On average, these are the best routes. There's some chance you'll be that guy routed into traffic today, only to enjoy the faster detour the next ten times. Decreasing total time in traffic for all participants is optimal on the long run for every single participant.

Somewhat related... Braess's Paradox.

Take the toy example illustrating that paradox ( https://en.wikipedia.org/wiki/Braess%27s_paradox#Mathematica... ) and an enlightened Google maps may do the altruistic thing of telling half of the drivers to stick on the upper route and half of the drivers to stick to the bottom route, but then some 3rd-party app will come along with the "shortcut" down the middle.

Maybe when all/most traffic is self-driving cars they could somehow be regulated to all use the same coordinated routing system without "cheating"... probably not.

This happens on NY 73, which is a road in the Adirondack High Peaks region between the highway and Lake Placid. It’s legit — even though cars are parked, you still need to slow down!

I usually avoid Google maps for traveling in areas that I don’t know. It takes you off on stupid routes that save marginal time (if any) too much. Waze is good for interstate travel, but IMO Apple is better for giving you the “right”, low drama route.

This is my common complaint with Waze and most driving apps - they optimize for "least time," not "least anguish". I would almost always prefer to spend an extra 5 minutes on my "normal" route (that I have on auto-pilot) instead of saving that 5 minutes by taking confusingly-named side-streets with lots of left turns onto other busy side-streets. I don't know if Apple Maps really gets this any better to be honest.

This is an understated problem. Google used to regularly take me through a route that involves a left turn in a major road and multiple crossings where cross traffic won’t stop. Literally two blocks from the suggested left turn is a left turn arrow signal which also avoids the dangerous crossings.

It’s both dangerous and anxiety inducing and doesn’t save time either.

I wish it had a “no left turns” feature like they have for avoiding tolls.

I wish cities had roundabouts instead of intersections, so that would not be a problem :)

They take space for anything at reasonable size I'd guess.

Strangely enough though it's one of the few things I miss about the city I grew up in (Jeddah, Saudi Arabia) - lots of outdoor art and sculptures because they had lots of very large roundabouts (check Google images to see what I mean about the outdoor art).

It absolubately can work, for a famous and oft-studied example see the city of Milton Keynes in the UK. Hundreds of roundabouts

America cities have too many oversized streets and major intersections. Keep the streets small and you don't need as many intersections and you can invest more in the ones you have. This is what they do in the Netherlands.

Wisconsin has become one giant series of roundabouts across the state over the past decade; they’re everywhere after new construction and being retrofitted whenever the opportunity arises. Sometimes 2 and 3 roundabouts connected at times.

I hate roundabouts. I live in France, and we have them everywhere. They're a huge clusterfuck that does not win any time, and when misused can be a lot more dangerous than a properly planned intersection. Of all the accidents I've driven past on my daily drives, all of them happened on roundabouts. I hate them.

> Of all the accidents I've driven past on my daily drives, all of them happened on roundabouts.

In localities with few/no roundabouts, the vast majority of accidents still happen at intersections. Intersections of roads are the most prone to accidents (for fairly obvious reasons), with or without roundabouts.

I believe there is research showing roundabouts are actually quite a bit safer in general, but it's possible it depends on various things. But anyway, accidents happen mostly where two roads intersect, either way.

Roundabouts are great for cars, and awful for just about everyone else. People, bikes, buses, trucks: all face difficulty or death in their presence.

> People, bikes, buses, trucks: all face difficulty or death in their presence.

Roundabouts make my bicycle commute way faster. This in The Netherlands where we cycle a lot. Difficulty or death is entirely incorrect. Suggest to compare the designs across countries, maybe there's a huge difference in the design?

I think the issue comes when people try to circumvent the give-way priority - which gets worse when there are multiple lanes and multiple exits.

It certainly seems to be a people problem rather than a roundabout deign issue from my experience?

People trying to "beat" people out onto the roundabout rather than giving way, people not using proper lane discipline and squeezing/cutting into other lanes etc..

You should copy the Dutch then. Their car+bike road designs are something else. Roundabouts included.

> They're a huge clusterfuck that does not win any time

Research shows the opposite. They can handle more traffic within a given timeframe. The bit about accidents seems strange. Not sure why you'd think that. I'm wondering if your roundabouts have some kind of design flaw.

Well, having driven on France on holidays, the problem seems to be with the driving culture. The roundabouts are very busy, and likely as a result drivers dart out in the shortest possible gaps in traffic. If you decide to play it safe, then the backed up traffic behind you starts honking at you, increasing the pressure and pushing you to make a bad decision. I found it very stressful indeed! (Perhaps all of France is not like this).

The Place Charles de Gaulle Étoile (Arc de Triomphe) in Paris is a gigantic roundabout, with the peculiar feature that traffic entering the system has priority ; traffic already on the roundabout must yield. This would lead to perpetual gridlock were there no traffic lights controlling access to the roundabout. This kind of junction used to be common in cities.

Out in the countryside roundabouts in France operate in the "normal" way : approaching traffic yields to traffic already in the system, with the usual triangular "yield" sign making priority clear.

It's possible that your experience of French roundabouts is related to driver confusion between these two systems.

The research I've seen suggest roundabouts do indeed cause accidents, but they're always low-speed accidents where people don't die. Intersections on the other hand are more likely to cause deaths because speeds can be much greater.

Roundabouts only work in the burbs. In cities, they are death traps for pedestrians and cyclists.

Nonsense. See video linked in https://news.ycombinator.com/item?id=8819614 starting at minute 3.

> In cities, they are death traps for pedestrians and cyclists.

I haven't experienced that for the many roundabouts we have in The Netherlands. Throughput is higher (can easily be measured), plus waiting time is lower thanks to roundabouts.

Not exactly sure what you mean with "burbs" though it doesn't matter as they're heavily used in cities here.

I ride my bike through rotaries (as we call them here in Mass.) all the time. It's not that hard. It is advisable to try to ride as fast as possible, but the key is to signal clearly all movements in and out of the lane and the circle.

Roundabouts have the unfortunate property of smoothing throughput from a sequence of chunks and gaps to a constant dribble. That's fine on limited access roads with overpasses and parallel feeder roads for local traffic or where overall throughput is always low enough to still comfortably cut in from a lower rank junction. On everything in between (enough traffic to make turning difficult, but still having direct junctions with access roads) you are better off with the occasional traffic light that will as a side effect serve gaps to meet rank junctions downstream.

The problem is that it would be mostly useless, because it'd be buried in a menu somewhere where you can't easily change it while driving.

The avoiding tolls thing is absolutely terrible that way. Thankfully, they finally put in a new feature where it shows you alternate routes, with and without tolls, and tells you if they do or don't have tolls. It took years for this to come about even though I had requested it many times, and it's plainly obviously needed. But before this, it was all-or-nothing, and it wasn't easy to change on the fly, and having it on was bad because it would frequently try to route me onto an expensive toll lane even though there was zero time savings, but for other routes a toll road was absolutely necessary.

They should have it weight left turns higher (as in worse) so that routes tend to avoid them. When I drive to work, I can take my normal exit, and only take right turns, or if I miss it, I can take the next exit, but there's a terrible left turn there with a very long light. Google Maps consistently says "1 minute longer" for this route: BS. It might be about the same if you happen to catch the light, otherwise it's 5 minutes longer. I feel like their algorithm really doesn't take this into account well.

This problem baffles me because all that has to be done to correct it is increase the estimated time of arrival penalty for making an unprotected left turn so that the routing algorithm doesn't mistakenly think it's faster. Ditto for U-turns, which Google constantly suggests (rather than turning onto a side street and U-turning there or going around the block) even when they are impossible (median in the way), dangerous (high speed road) or illegal.

Just another area where rules based heuristic systems can outperform machine learning systems by not removing edge cases through over generalization.

That's weird. Google maps seems to prefer to direct me on a kilometres long detour than to take a simple u-turn into an empty driveway. They do seem to have rules about u-turns, they just don't make sense.

One time I had TomTom accidentally switch itself on in my bag on the floor, set to Homer Simpson's voice and trying to redirect me to home.

As I got further and further away the "turn left here" and "turn right here" directions got more and more frequent, until eventually he just started repeating "turn around where possible" until I was able to find somewhere reasonable to stop and turn it off.

I would love it if these apps gave an option for easiest drive. Both Apple and Google tell me to take a left turn by my house onto a busy road when its actually quicker to go the other way and just loop around. In fact, if they just had an “avoid unprotected left turns” option that’d be fantastic.

Waze has a feature called "Avoid difficult intersections" designed exactly for these situations. https://www.pcmag.com/news/waze-wants-to-help-you-avoid-diff...

Sorry for the minor comment but I noticed this part: "side-streets with lots of left turns onto other busy side-streets"

Now I live in Europe, presumably you live in the US (sorry if that's wrong). Is there any particular difficulties with doing left turns compared to right turns where you live?

Left turns cross all four lanes of traffic involved in an intersection while right turns involve two, so if either street is busy you have to wait longer for a gap.

Biggest difference I found moving from Europe was that you can make a right turn on a red traffic in many states in the US (but not in New york city apparently!). You have to treat it like a stop sign and yield to any traffic on the road you're joining. Ends making a journey of rights turns much fast than one with left turns, particularly in rush hour

Yes, I'm in the US. Left turns have to cross oncoming traffic so you have to monitor both directions. This is especially stressful if {both roads are busy, there's weather/dark, you don't know the roads}.

Roads can be very wide in the US, it is not uncommon for cities to have 6 or 8 lane roads at-grade (i.e. not expressways). So when turning left, one must turn across 3 or 4 lanes of oncoming traffic, instead of just one as when turning right.

> with Waze and most driving apps - they optimize for "least time," not "least anguish"

It's my impression that Google Maps does the latter; Waze is known for taking wild routes that are nominally faster. I'm not sure that I see a basis for complaint, as they're just offering a choice of different route objectives to optimize for (I know people who explicitly use Waze because they prefer its aggressive optimization to Maps). I suppose a toggle would be nice though.

I travel to new cities often and usually rent a car from the airport. In most cities around the world, there is dedicated signage pointing out the access to the airport on every major approach it. I would say half the time Google wants to take me on a different route on some back road that is not meant to be the public access to the airport, which creates the extra mental overhead of having to ignore the gigantic "AIRPORT - EXIT HERE" signs all along the way. It even does this at SFO, presumably an airport that the very people who work on Google Maps use, where it always wants to exit the interstate one exit before the sign for the airport and take city streets instead. The worst experience I've had of this was in Denver, where it directed me onto a country road that circled the airport property and then ended at a temporary road closure. That is, Google's "5 seconds faster than the main, signed route" optimization wasn't even an actual route to the airport!

My favorite example of this was in NYC heading to Citi Field for a Mets game. Google Maps guides you past the exit, has you make a u-turn by getting off at the next exit (which is really unusual) and get you trapped on some ramp for an extra 20m.

I had a dedicated GPS a decade ago (I don't remember what brand) that had route options for Fastest, Shortest or Simplest that were easy to pick when you were first putting in the address. With few exceptions, simplest was usually the best option, especially if driving in an unfamiliar area. I think it basically tried to minimize turns or number of different roads. That would be a nice thing to see on Google Maps.

I will risk writing that you had Navigon or TomTom. At least in Europe those were the two main choices.

I often use Navigon I keep in a (very old) iPhone, it has a fantastic feature that you can download the maps for countries. I like it better for long distance driving than Google Maps.

Garmin was also a major option in the US.

Yes this is one thing that frustrates me about Google's navigation. Often it will take you off the main road through difficult and narrow side roads in order to save 2 mins. When I'm in an unfamiliar area I would rather save the stress and sit in traffic for those two minutes.

It's not a new problem by any means - UPS trucks avoid left turns thanks to their navigation algorithm for more than a decade and they were trying to do it since the 1970s.

The OG mapquest had three options for route optimization: quickest (time), shortest (distance), and scenic. Would be nice to have that again.

The weird thing is that old school software like Microsoft Streets and Trips could do this pretty easily.

These map apps make a conscious decision to deliver this user experience.

There's a regular drive I make. I can either hop on the freeway and it takes 1hr 4min, or I can take backroads and it takes 1hr 2min.

The backroads also has a 25% or so chance of getting stopped by a train. Big massive transcontinental train connecting the biggest port complex in the US to the rest of the country. Takes about ten minutes. Last time I hit the train, as it was almost done, a train came the other way.

Haven't taken that route since, despite the fact that it saves two minutes, sometimes more if there's traffic on the freeway. It's not a good value proposition.

I recently tried walking directions to JapanTown in SF on both Google Maps and Apple Maps.

gMaps took us to the restaurant on the shortest path - straight through the heart of Tenderloin. My girlfriend was sketched out as heck and almost made us take an Uber.

On our way back from the restaurant, we used Apple Maps. It took us on what looked like a longer route but lo and behold, it took a wiiiide berth around the Tenderloin. We saw not a single sketchy thing on our entire night time walk.

Don’t know if it was a coincidence or Apple Maps really has an “avoid seedy areas” model. The route was way simpler at least. 2 or 3 turns compared to gMaps’s 15

And grid geometry means Apple Maps was only 2min slower. But way nicer and easier to navigate.

> On our way back from the restaurant, we used Apple Maps. It took us on what looked like a longer route but lo and behold, it took a wiiiide berth around the Tenderloin. We saw not a single sketchy thing on our entire night time walk.

Microsoft patented such an "avoid ghetto" idea year ago, but people got mad and called it racist:

https://www.npr.org/2012/01/25/145337346/this-app-was-made-f...

https://www.cnet.com/news/the-joy-of-microsofts-avoid-ghetto...

Because, as we all know, trying to avoid getting mugged or carjacked is racist.

To be fair, most of SF’s seedy people are white drug addicts.

I personally don't hold that opinion, but this could still be construed as racist.

I've visited and I wouldn't say that is true. There is all kinds of seedy. There are white/skinny seedy people walking rats, kinda freaky. But taking the downtown bus gave equal race representation when someone started a fight with me over there bag's right to occupy my seat. The Asian gangs robbing/hitting people on the bart I was on showed a different side of seedy. The uniquely San Fran treat was someone asking my opinion on my beat poetry they wrote around race struggles. After commenting positively (just to be nice) said Poet offered to sell and when refused when on racial rant and become very aggressive.

It's a diverse place of seed and rich. It's a very future feeling place. The experience made me think this is a slice of what the future will look like. It's kinda has this dystopian gothic novel feel. The mini subways (barts) are automated and cool and look like they are in the future but look like they are falling apart too. You get this weird post apocalyptic feeling of a future that has come and gone but hasn't reached you yet

The Bart is designed to look automated, but it's one of the least automated transit systems around - There's a human driver on each train, station agents to watch the doors, and it's only the fact that they have nice recordings that makes it feel impersonal and automated.

Yeah, no. BART doesn't have station agents watching the doors, and human drivers are a common occurrence on pretty much every subway system in the world.

They should get some nice recordings here in DC for the Metro system. They usually just have the drivers making announcements, and half the time I can't understand a word they said.

My opinion: If the model is based on crime stats, it’s ok. If it’s based on demographics, it’s a problem.

I'm right there with you. A model based on race would be racist. The issue is that in urban areas, crime rate is often a good proxy for race (particularly when looking at, say, muggings or carjackings rather than white-collar crime). I consider that irrelevant, but as the two articles I linked show, the "disparate impact" type folks don't care about which stat is being used.

> it took a wiiiide berth around the Tenderloin.

Next thing you know, (SF) City politicians are up in arms about why Apple is "blackballing" Tenderloin, and rake up past issues such as red-lining, ghettos, etc.

No need for explicit "avoid seedy areas" model. Just look where people prefer to walk (esp. people with iphones), and favor those streets in your algorithm.

>I usually avoid Google maps for traveling in areas that I don’t know. It takes you off on stupid routes that save marginal time (if any) too much.

We relied on GMaps on a recent (my first ever) visit to Malta, and it took us through some pretty dodgy, steep and narrow, local roads on our way in.

We didn't know of this GMaps' preference of, erm, interesting routes, so we figured that's just how Malta is. A few days later and, of course, we learn there's a perfectly normal, pleasant route that took us there in as little time as the dodgy one...

I would definitely love to see a setting for "if this one-lane back road route saves me less than ten minutes, just take me on the big main street."

So a theory I've had for a bit now... Google maps doesn't optimize for ease of travel, or time, or anything like that - it optimizes for engagement, like every other Google product. And how would you do that with a maps product? By making routes too complicated to memorize, too specialized (back streets, etc) to generalize, and then changing things up often to prevent a route from becoming second nature anyways.

Which feels like a lot of what people complain Google Maps does...

... especially if they can bring you by some companies who have paid for ads on Maps. Where I used to live, in a grid city, Maps would take drivers down one single street whose only discerning feature from those 2 streets north or south (alternating one-ways) was that it could say "Turn left at the Pizza Hut".

Sounds funny, but your theory is false. I use GMaps every day to drive to and from work, with a very simple route that's almost all on an interstate highway. Almost every day, it gives me the exact same route (which is so simple I really don't need GPS navigation at all for it). Sometimes, it gives me a different route because of an accident, traffic jam, construction delay, etc., which is why I use GPS all the time, but I see absolutely nothing to indicate that Google intentionally makes my route more complicated than it needs to be. (And yes, these delays are real: this road is under a lot of construction at the moment, and it frequently has crashes and traffic jams in places.)

Gm definitely does not taken elivation into account. If you live in a hilly country you will want to avoid alternative routes that seem shorter but are really elevation changes.

There is a popular road in a relatively rural part of the SF Bay area for bicycling and on the weekends it appears to have traffic due to the slower speed of the numerous cyclists. There isn't actually more traffic. Google then sends cars on a winding narrow hilly alternate route.

Maps starts reading pedestrians in downtown SF as traffic during road closures for e.g. stadium events, and tries to direct cars along jammed and even closed roads. Too many times have I been stuck less than a mile from a destination for more than an hour, sometimes even two.

As a cyclist, this sounds like a feature, not a bug. Fewer cars on the road is always a good thing for cyclists if they have to share the road.

Yes, its good for cyclists and not really detrimental to cars. Although, I'm sure the local drivers know the other route in annoying and not faster.

This also happens with gas stations/parkings that are so near the main road they are confused with it.

I have the same frustration with Google Maps.

I have a route I take, and let's say I split it up with A-B-C. now Google Maps tells me that AB-BC takes an hour, but AB'-BC' is 15 minutes faster.

It usually turns out that the time delay is on the later leg. So I could take AB or AB', it doesn't matter, and then BC' is the faster choice later on.

I wish Google Maps wouldn't show me AB', because it's misleading. At best it should say AB' is the same ETA. But clicking on it means that some route near the end of my drive is now different, it doesn't actually save me time for my current divergence. It's misleading.

The key assumption there is that your alternate-self ghost-car is following the flow of traffic. Which might be true if the route was a highway; but what if it was, say, a shopping district? (“The other you would have gotten here slower, because they’d have been lured into a Starbucks drive-through on the way.”) Or what if they’re a new driver/have a flat tire, and so are driving slower than the speed limit with cars honking behind them? (“Your other self would have been a teenager who hesitated too long on a left turn at an intersection, and so had to wait an extra light cycle.”)

Reply on news.ycombinator.com