> For this study, the Google Maps algorithm was modified to prefer alternative routes with similar travel times and segment types, effectively guiding trips away from the pre-selected congested segments
> Over a six month period, we adopted a city-wide switchback (also known as crossover) experimental design, alternating between this treatment and the control (unaltered) routing algorithm over consecutive days to appropriately measure the effect of this intervention
> Averaged across cities, we observe a median increase of around 2% in driving speeds on targeted segments, corresponding to a median decrease of 0.5% to 1.0% in fuel consumption rates
The cities were: Atlanta, Boston, Chicago, Los Angeles, Miami, New York, Philadelphia, Salt Lake City, San Francisco, Seattle.
The south SF Bay Area cities of Los Gatos and Monte Sereno have persistent problems with traffic gridlock during summer weekends and holidays due to navigation apps routing drivers off of southbound SR-17 and onto surface streets. It gets so bad that residents literally can't get out of their own driveways and emergency response vehicles can't move.
There actually did used to be train service from San Jose to Santa Cruz but it was abandoned in 1940 due to high costs and lack of use. Rebuilding it would cost >$1B.
I read the article and didn't see anywhere if they collaborated with city planners, or just ran this on their own. I get the techno-optimism of making traffic 'better' and also it may be concerning if trucks are being routed through neighborhoods where planners didn't build for it.
One of the things this type of intervention doesn't take into account is that different roads are built with different levels of hardiness based on the amount of traffic they are expected to receive.
For instance, a few years ago, a segment of I-495 in Delaware needed to be unexpectedly shut down for emergency repairs. Drivers were rerouted. But because of the increase in traffic on the less-hardy detour route, that route needed repairs and repaving soon afterward, much more quickly than it would have ordinarily required.
So yes, drivers can be better dispersed to ease congestion, but we also need to consider the secondary effects to the roadways themselves.
A random factoid that will always stick in my brain is that the damage a vehicle does to a road is proportional to the weight on an axel...to the fourth power. I suspect rerouting semis to the roadway is by far the greatest factor.
Traffic apps only know about congestion if someone running the app goes down the congested road. Because of this, I've always suspected that the apps, from time to time, will route someone down a route they haven't gathered data on in a while, just to collect the data, and even if the route is likely to be suboptimal.
>I've always suspected that the apps, from time to time, will route someone down a route they haven't gathered data on in a while, just to collect the data, and even if the route is likely to be suboptimal.
I can't say this actually happened for me. For straightforward routes with no congestion I never saw random alternate routes being proposed. That makes sense, given that it'd probably tip people off. If this is happening, they must only be doing in cases where there's congestion and the difference is marginal, eg. it's rush hour and the "optimal" route takes 30 minutes but the alternate takes 33 minutes. Moreover you don't really need any deliberate effort to see this effect. If nobody is traveling on a side road, the algorithm will probably revert to historical patterns, which might turn out to be overly optimistic in congestion scenarios (eg. there's nearby road repairs and other people are already using it as an alternate), thereby giving you the impression that you got screwed over by the app.
Doesn't have to be true. My state has public camera feeds for various highway stretches. Too low-detail to identify vehicles but easily enough to detect congestion.
I wonder this same thing when I’ve been routed completely asinine ways off interstates at night when traffic appears calm. I always have to play the game of “is there horrendous traffic ahead from night paving / accident, or am I being conscripted as a traffic sensor robot?”
Why cant we have smart stop lights. Nothing more annoying than sitting on a street with nobody coming and a red light for some arbitrary time period that's totally incorrect for the time period I'm currently in
Most cities have some intelligence on their traffic lights, and they do have traffic sensors to know where cars are waiting.
You might not see any cars and are annoyed waiting at the light, but you don’t know what downstream things they are trying to control for. Maybe if they send you through the light, you are going to join some other traffic event that will cause an issue.
It could also just be suboptimal, but it very hard to know just from what you observe as an individal
If we could copy the traffic laws of a country like Germany to the US, I think that would have the biggest + cheapest effect. I am OK with automated/elevated enforcement if it means stop & go traffic evaporates into free flowing conditions.
We should also take the idea from Finland where the traffic fines scale with each person's ability to pay. $100 for camping the passing lane or failing to maintain a reasonable following distance is not a big consequence for a lot of people. $100k covers the edges a lot better.
I wish there was more police presence to spot lane-camping and fine them in Germany, too many idiots do it and get away with it.
If I ran Google Maps, I'd ask to route stupid drivers away from my routes. The GPS and phone accelerometer should be able to tell us who they are -- "Oh you have Google Maps actively navigating, GPS says you're going 50mph, and the app in focus is WhatsApp and you're using the phone keyboard? Fuck off! Also we recorded 5 swerve events in the last 10 minutes, hard to stay on your lane when you're texting!"
It'd be unethical, but man it'd be magical...
There's custom software for Elon's Tesla, I wonder if the cloud-based navigation system also directs people away from his routes so he can have a ride on emptier roads...
>During “treatment” days, the modified routing guided all trips that encountered the pre-selected congested segments toward alternative routes with similar travel times.
Why would they specifically need to route people away from congested segments? Presumably if a segment gets congested enough, it'd be considered slower and therefore won't get picked in the first place?
Every actor acting rationally in their own interest can get to a worse equilibrium than if they were coordinated(in this case, to completely ignore the new edge). There are many other examples of this in game theory, you should look it up.
Congestion can be solved by charging more for driving. Even a modest amount would start affecting how much people drive. The money could be used for better roads and improving other transit options that aren't so space-inefficient. That said, this would likely be unpopular to implement until the effects were visible. See congestion pricing in NYC as an example.
When Google Maps routes me using a smaller secondary road instead of the main road that I would otherwise have used , I've always wondered whether that significantly changes the amount of traffic that smaller road sees. It's funny to consider that arbitrary black-box changes to the routing algorithm can have a noticeable effect to people that live there.
(I'm not saying I'm the first or only person to think of it, but I did patent it, as well as the extra claim of compensating those people sent on the slower route).
For over a decade, I imagined that if I ever landed a job at Google, this would be my most significant project. It made me chuckle a bit when I read the announcement, they finally built it! confirming that my thoughts weren’t entirely delusional XD
The way to maximise traffic congestion would be to remove trucks off the road during peak times, banning trucks from the fast lane at all times, making it the social norm to toot someone in the fast lane so they move over.
I have a lot of difficulty thinking those changes alone would solve congestion. I’m especially thinking is ski traffic along I70 from Denver to the mountains. The issue is just volume of passenger vehicles, and there are no other realistic routes
> For this study, the Google Maps algorithm was modified to prefer alternative routes with similar travel times and segment types, effectively guiding trips away from the pre-selected congested segments
> Over a six month period, we adopted a city-wide switchback (also known as crossover) experimental design, alternating between this treatment and the control (unaltered) routing algorithm over consecutive days to appropriately measure the effect of this intervention
> Averaged across cities, we observe a median increase of around 2% in driving speeds on targeted segments, corresponding to a median decrease of 0.5% to 1.0% in fuel consumption rates
The cities were: Atlanta, Boston, Chicago, Los Angeles, Miami, New York, Philadelphia, Salt Lake City, San Francisco, Seattle.
The data and code is also available (https://github.com/google-research/google-research/blob/mast...) from the paper (https://www.nature.com/articles/s44284-026-00443-x).
Kudos Google! Nice to see this kind of work. That said, let's just build more trains?
The south SF Bay Area cities of Los Gatos and Monte Sereno have persistent problems with traffic gridlock during summer weekends and holidays due to navigation apps routing drivers off of southbound SR-17 and onto surface streets. It gets so bad that residents literally can't get out of their own driveways and emergency response vehicles can't move.
There actually did used to be train service from San Jose to Santa Cruz but it was abandoned in 1940 due to high costs and lack of use. Rebuilding it would cost >$1B.
https://www.goodtimes.sc/isnt-train-san-jose/
I read the article and didn't see anywhere if they collaborated with city planners, or just ran this on their own. I get the techno-optimism of making traffic 'better' and also it may be concerning if trucks are being routed through neighborhoods where planners didn't build for it.
Found the European :)
One of the things this type of intervention doesn't take into account is that different roads are built with different levels of hardiness based on the amount of traffic they are expected to receive.
For instance, a few years ago, a segment of I-495 in Delaware needed to be unexpectedly shut down for emergency repairs. Drivers were rerouted. But because of the increase in traffic on the less-hardy detour route, that route needed repairs and repaving soon afterward, much more quickly than it would have ordinarily required.
So yes, drivers can be better dispersed to ease congestion, but we also need to consider the secondary effects to the roadways themselves.
A random factoid that will always stick in my brain is that the damage a vehicle does to a road is proportional to the weight on an axel...to the fourth power. I suspect rerouting semis to the roadway is by far the greatest factor.
Traffic apps only know about congestion if someone running the app goes down the congested road. Because of this, I've always suspected that the apps, from time to time, will route someone down a route they haven't gathered data on in a while, just to collect the data, and even if the route is likely to be suboptimal.
There's probably enough people taking wrong terms (such as me) to provide enough data. ;-)
Wrong term, indeed :)
>I've always suspected that the apps, from time to time, will route someone down a route they haven't gathered data on in a while, just to collect the data, and even if the route is likely to be suboptimal.
I can't say this actually happened for me. For straightforward routes with no congestion I never saw random alternate routes being proposed. That makes sense, given that it'd probably tip people off. If this is happening, they must only be doing in cases where there's congestion and the difference is marginal, eg. it's rush hour and the "optimal" route takes 30 minutes but the alternate takes 33 minutes. Moreover you don't really need any deliberate effort to see this effect. If nobody is traveling on a side road, the algorithm will probably revert to historical patterns, which might turn out to be overly optimistic in congestion scenarios (eg. there's nearby road repairs and other people are already using it as an alternate), thereby giving you the impression that you got screwed over by the app.
This would sort of happen naturally if they routed as if roads with no data in the past N minutes/hours were at full speed.
Doesn't have to be true. My state has public camera feeds for various highway stretches. Too low-detail to identify vehicles but easily enough to detect congestion.
I wonder this same thing when I’ve been routed completely asinine ways off interstates at night when traffic appears calm. I always have to play the game of “is there horrendous traffic ahead from night paving / accident, or am I being conscripted as a traffic sensor robot?”
Government data sources can input real time data into Google Maps [1]. So, it's more likely than not that local police or state DoT is behind that.
[1] https://contentpartners.maps.google.com/
Why cant we have smart stop lights. Nothing more annoying than sitting on a street with nobody coming and a red light for some arbitrary time period that's totally incorrect for the time period I'm currently in
Relevant XKCD: https://xkcd.com/277/
Most cities have some intelligence on their traffic lights, and they do have traffic sensors to know where cars are waiting.
You might not see any cars and are annoyed waiting at the light, but you don’t know what downstream things they are trying to control for. Maybe if they send you through the light, you are going to join some other traffic event that will cause an issue.
It could also just be suboptimal, but it very hard to know just from what you observe as an individal
They exist. Your city needs to set it up. They can also switch to flashing red lights.
If we could copy the traffic laws of a country like Germany to the US, I think that would have the biggest + cheapest effect. I am OK with automated/elevated enforcement if it means stop & go traffic evaporates into free flowing conditions.
We should also take the idea from Finland where the traffic fines scale with each person's ability to pay. $100 for camping the passing lane or failing to maintain a reasonable following distance is not a big consequence for a lot of people. $100k covers the edges a lot better.
https://www.theguardian.com/world/2023/jun/06/finnish-busine...
I wish there was more police presence to spot lane-camping and fine them in Germany, too many idiots do it and get away with it.
If I ran Google Maps, I'd ask to route stupid drivers away from my routes. The GPS and phone accelerometer should be able to tell us who they are -- "Oh you have Google Maps actively navigating, GPS says you're going 50mph, and the app in focus is WhatsApp and you're using the phone keyboard? Fuck off! Also we recorded 5 swerve events in the last 10 minutes, hard to stay on your lane when you're texting!"
It'd be unethical, but man it'd be magical...
There's custom software for Elon's Tesla, I wonder if the cloud-based navigation system also directs people away from his routes so he can have a ride on emptier roads...
>During “treatment” days, the modified routing guided all trips that encountered the pre-selected congested segments toward alternative routes with similar travel times.
Why would they specifically need to route people away from congested segments? Presumably if a segment gets congested enough, it'd be considered slower and therefore won't get picked in the first place?
Probably many reasons but here’s one demo: https://en.wikipedia.org/wiki/Braess%27s_paradox
Every actor acting rationally in their own interest can get to a worse equilibrium than if they were coordinated(in this case, to completely ignore the new edge). There are many other examples of this in game theory, you should look it up.
Another question: they say road speeds increased, great. But that’s a different metric than average drive time. Curious if that also decreased.
Congestion can be solved by charging more for driving. Even a modest amount would start affecting how much people drive. The money could be used for better roads and improving other transit options that aren't so space-inefficient. That said, this would likely be unpopular to implement until the effects were visible. See congestion pricing in NYC as an example.
Interesting, I always just assumed they were already doing this.
When Google Maps routes me using a smaller secondary road instead of the main road that I would otherwise have used , I've always wondered whether that significantly changes the amount of traffic that smaller road sees. It's funny to consider that arbitrary black-box changes to the routing algorithm can have a noticeable effect to people that live there.
Where isn't there traffic congestion?
Hey I patented this idea about a decade ago:
https://patents.google.com/patent/US20170184409A1
(I'm not saying I'm the first or only person to think of it, but I did patent it, as well as the extra claim of compensating those people sent on the slower route).
For over a decade, I imagined that if I ever landed a job at Google, this would be my most significant project. It made me chuckle a bit when I read the announcement, they finally built it! confirming that my thoughts weren’t entirely delusional XD
The way to maximise traffic congestion would be to remove trucks off the road during peak times, banning trucks from the fast lane at all times, making it the social norm to toot someone in the fast lane so they move over.
I have a lot of difficulty thinking those changes alone would solve congestion. I’m especially thinking is ski traffic along I70 from Denver to the mountains. The issue is just volume of passenger vehicles, and there are no other realistic routes