![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Entry tags:
Anti-patent: Building a better traffic light
[Happy birthday to ![[livejournal.com profile]](https://www.dreamwidth.org/img/external/lj-userinfo.gif)
ladysprite!]
One of the more common sources of minor modern-world frustration is the traffic intersection. Traffic lights are typically pretty blunt-force instruments, running the same timing over and over again, regardless of traffic. The really fancy ones might be sensitive to time of day, and have their timing modeled in advance by traffic engineers, but it looks to me like those engineers don't come and check their work very often, because we still spend our time sitting at badly-programmed lights. And sadly, individual human ingenuity doesn't seem to help much: putting a person in charge of the intersection usually makes things worse, not better. (Nothing screws up traffic quite as effectively as the average traffic cop. A few are good at it, but most are terrible, because they rarely switch directions fast enough.)
This leads me to wonder if anybody's considered the one object that potentially has all the necessary information to do it right: the light itself. Could we build a *smarter* traffic light? I suspect so, by making a light that understands the traffic passing beneath it and the time of day, and can experiment and learn what works.
Here's the high concept. Take a traffic intersection, with all the related lights, and put sensors in each direction. (Could be treadles, could be cameras, whatever -- the important bit is that it can count the cars coming through.) Hook this whole thing to a simple learning computer -- probably a neural net, possibly some sort of annealing or evolutionary algorithm, so long as it is capable of gradually improving its own timing. Give the computer a clear measure of "better" and "worse", which mostly consists of the number of cars that actually pass through, plus a desire to balance the directions reasonably fairly.
Peg some sensible extremes: eg, don't go green for less than five seconds or more than sixty at a time. Don't allow the timing to shift too quickly, to avoid dramatically bad extremes to come out suddenly. Hard-code the firm assumptions: eg, exactly one direction must be green at any given time, and a fire truck's signal overrides everything else. Give the thing a clock (with an understanding of time and day of the week) as an additional input, so that it can factor that into its calculations.
For extra credit, add an additional sensor a bit down the street in each direction, to detect backups from the light, and put a particular priority on avoiding them.
In more complex environments, where there are a number of lights near to each other, hook the networks together, so that they can work co-operatively to improve the overall traffic flow.
For all I know, someone may already have invented this, but it sure isn't widespread if so. (This invention brought to you by my sitting in the usual pointless Route 3 traffic jam on my way home the other day.) It could make a fine project for some entrepreneurial programmer: build and improve the thing with simulated inputs, and then find an agreeable traffic-light manufacturer to partner with or sell it to. I suspect it would require a fancier computer than most lights now have, but in the age of $300 laptops, I can't imagine it would add significantly to the cost of a full light system.
This is all, BTW, very strongly based on my general approach to cognition -- when I say a "smarter" traffic light, I mean that quite literally. The above has all the elements of intelligence, albeit in a very limited domain. It has a number of heterogenenous inputs; outputs that are capable of experimenting; a neural network (preferably a multi-level one) capable of associating the outputs and the inputs in feedback loops; "instincts" that start things in the right direction and help avoid foolish extremes; and instinctive concepts of "better" and "worse" to steer it in the right direction. Far as I've been able to figure out, that's most of how intelligence works: humans seems to be mostly based on the same principles, albeit scaled up in complexity many orders of magnitude.
Do you think this would work? What clever solutions have you had for workaday problems, that should get anti-patented before some patent troll gets their hands on the idea?
ladysprite!]One of the more common sources of minor modern-world frustration is the traffic intersection. Traffic lights are typically pretty blunt-force instruments, running the same timing over and over again, regardless of traffic. The really fancy ones might be sensitive to time of day, and have their timing modeled in advance by traffic engineers, but it looks to me like those engineers don't come and check their work very often, because we still spend our time sitting at badly-programmed lights. And sadly, individual human ingenuity doesn't seem to help much: putting a person in charge of the intersection usually makes things worse, not better. (Nothing screws up traffic quite as effectively as the average traffic cop. A few are good at it, but most are terrible, because they rarely switch directions fast enough.)
This leads me to wonder if anybody's considered the one object that potentially has all the necessary information to do it right: the light itself. Could we build a *smarter* traffic light? I suspect so, by making a light that understands the traffic passing beneath it and the time of day, and can experiment and learn what works.
Here's the high concept. Take a traffic intersection, with all the related lights, and put sensors in each direction. (Could be treadles, could be cameras, whatever -- the important bit is that it can count the cars coming through.) Hook this whole thing to a simple learning computer -- probably a neural net, possibly some sort of annealing or evolutionary algorithm, so long as it is capable of gradually improving its own timing. Give the computer a clear measure of "better" and "worse", which mostly consists of the number of cars that actually pass through, plus a desire to balance the directions reasonably fairly.
Peg some sensible extremes: eg, don't go green for less than five seconds or more than sixty at a time. Don't allow the timing to shift too quickly, to avoid dramatically bad extremes to come out suddenly. Hard-code the firm assumptions: eg, exactly one direction must be green at any given time, and a fire truck's signal overrides everything else. Give the thing a clock (with an understanding of time and day of the week) as an additional input, so that it can factor that into its calculations.
For extra credit, add an additional sensor a bit down the street in each direction, to detect backups from the light, and put a particular priority on avoiding them.
In more complex environments, where there are a number of lights near to each other, hook the networks together, so that they can work co-operatively to improve the overall traffic flow.
For all I know, someone may already have invented this, but it sure isn't widespread if so. (This invention brought to you by my sitting in the usual pointless Route 3 traffic jam on my way home the other day.) It could make a fine project for some entrepreneurial programmer: build and improve the thing with simulated inputs, and then find an agreeable traffic-light manufacturer to partner with or sell it to. I suspect it would require a fancier computer than most lights now have, but in the age of $300 laptops, I can't imagine it would add significantly to the cost of a full light system.
This is all, BTW, very strongly based on my general approach to cognition -- when I say a "smarter" traffic light, I mean that quite literally. The above has all the elements of intelligence, albeit in a very limited domain. It has a number of heterogenenous inputs; outputs that are capable of experimenting; a neural network (preferably a multi-level one) capable of associating the outputs and the inputs in feedback loops; "instincts" that start things in the right direction and help avoid foolish extremes; and instinctive concepts of "better" and "worse" to steer it in the right direction. Far as I've been able to figure out, that's most of how intelligence works: humans seems to be mostly based on the same principles, albeit scaled up in complexity many orders of magnitude.
Do you think this would work? What clever solutions have you had for workaday problems, that should get anti-patented before some patent troll gets their hands on the idea?
no subject
no subject
no subject
no subject
(More seriously: yeah, it's somewhat true. The day in which we need to start treating our computers nicely so that they don't turn evil and kill us all isn't necessarily all that far away. Yay singularity...)
no subject
no subject
This is how Skynet started.
(Do click the link for some scary new tech from Lockheed-Martin, especially if you remember the phrase "smart crowbars".)
no subject
no subject
no subject
Basically, my intuition is that this functions well at all times, but is inherently a bit inefficient, in that it requires each person approaching the intersection to make an individual decision. So long as the time required to make that decision is less than the average rate at which cars are approaching, it probably works great; once the traffic passes that point, it seems likely that you'll wind up with jams.
That said, I'd bet that a combination of this approach plus sensible road redesign could obviate 90% of all existing traffic lights. (And in theory I suspect you could do away with all of them, but it would likely require fairly radical road moving in some cases.)
And it's certainly possible that my intuition is incorrect: there are inefficiencies involved in traffic lights as well (since they are causing one direction to stop, and it takes time to re-accelerate), so it *could* be that the inefficiency of this scheme isn't significantly worse than that of the existing one. I suspect it is, but a real study would be useful...
no subject
On a weekly level, you have to handle the difference between weekday and weekend traffic -- and this might include effects due to church, football games, and that kind of thing. (Not to mention Friday afternoon traffic patterns, which are different.)
On a larger level, you have shifts in patterns due to school schedules and Daylight Savings Time. You also have to worry about exceptional cases -- Thanksgiving weekend traffic, for example.
I point all these cases out because your hypothetical neural net will be heavily challenged. (And it ways that can break an improperly-bounded neural net.)
Now, if you've ever played Sim Tower (or Yoot Tower), you'll notice that the pattern I've described is very similar to the challenge presented to a multi-use tower's elevator plan. Interestingly, the best solution to that is the simplest.
So my thinking on "bad traffic lights" is that the engineers have actually over-thought the intersection. You can get a lot done with magnetic loops (your car-detector of choice) and simple algorithms. For example, "Is somebody waiting at the intersection?"
About the only thing you'd have to decide is how often the system would check the loop. If you had an intersection of a two-lane road and a one-lane road, you might check the two-lane road twice as often as the one-lane road. After all, you've already committed twice the resources to the wider road.
no subject
Many intersections have a sensor in the road and watch to see if someone is waiting, and for how long, and can adapt that way. Some have cameras to watch the queues for similar adjustments.
What I like are recent GPS systems that actually share realtime data with other nearby units so as to inform you of traffic before you run into it, and suggesting alternate routes.
Me, I'm for more roundabouts (rotaries). Smartly implemented, they self manage variable traffic flows and congestion and smooth it all out. Of course, poorly implemented they are the devils tool.
no subject
The one that was never built near where I used to live in NH didn't meet the expected throughput criteria. (Also, geography was a minor factor.) Others elsewhere in Nashua did their job well.
no subject
Out here in San Diego, I think we have exactly two rotaries, and one of them is near Legoland (technically in Carlsbad) and probably shouldn't be there, as no one groks how it works.
no subject
no subject
Of course, we don't have to spend a large amount of our budget fixing potholes and dumping salt on our roads every year. *grin* I think this may have something to do with the relative scarcity of sophisticated traffic controls in the Northeast.
no subject
Oh, sure -- I was actually assuming that as pre-existing technology. But my observation is that they still aren't very smart in how they *use* that data. (Although it's certainly possible that some *are* very smart, and I'm just not noticing them because they work well. It's easier to notice broken systems.)
What I like are recent GPS systems that actually share realtime data with other nearby units so as to inform you of traffic before you run into it, and suggesting alternate routes.
Oh, yes. Indeed, this is the system that I've been waiting for for many years -- I first started talking the idea up, I dunno, probably eight or nine years ago now. There's a part of me that occasionally contemplates seeing whether Dash's patent is actually invalid (certainly I *talked* about its key ideas a long time ago, but I don't know if I ever really published them), but it's probably not worth the bother, especially because I'd kind of like to see Dash succeed.
Of course, poorly implemented they are the devils tool.
Coming from New Jersey, I'm well aware of that. The Somerville Circle was legendary for the number of people who got killed on it. OTOH, I do think that the switch from rotary to traffic light was poorly handled at Alewife, so both ways can be done badly...
no subject
no subject
* Basing lights on traffic actually showing up. The lifhgt near the house I spent my latter teens in had a light that would change when you pulled up to it -- and if more cars pulled up, it would change more quickly. (One car pulling up was still a 30 second wait, unless there was no traffic coming: multiple cars would make it change more quickly).
* NYC uses a traffic control center, I believe, where the metrics of how many cars are passing through or not passing through raises red flags so people can retrain on the fly. This has a lot of the components of what you describe: traffic flow, analysis, feedback -- but the brain at the top is a human instead of the light.
Cambridge lights, in my experience, tend to be relatively good, though not perfect. I think it's a somewhat harder problem to solve, most of the time: the side effects from a single slow down stretch very far, to the point that a light being crowded at the BU bridge can have effects felt over by the Galleria, or up by Central Square. You'd need a pretty wide net of traffic lights to be participating in any dense area for it to have a real effect, I think.
no subject
I saw on someone's LJ recently (maybe even this one!) a series of anecdotes about robots with machine learning capabilities, and the... unique failure modes they arrived at, under the belief that they were successful. Such as the tall robot who always fell over after one step, but was so tall that the sensor detected "lots of distance covered in a short time", and thought it was "more fit" than a shorter one which staggered for a while.
no subject
no subject
http://fosveny.livejournal.com/15037.html
no subject
no subject
no subject