Posting in Technology
European researchers want traffic lights to be self-organizing to reduce your time stuck in traffic.
Beyond the obvious annoyance of sitting in traffic, road congestion costs the nation $100 billion each year. Sure, riding more bikes, building more roads, encouraging people to car pool, and buying more eco-friendly cars will help ease traffic jams.
Stefan Lämmer at the Institute of Transport & Economics of TU Dresden showed that it is possible to improve the traffic situation by simply re-doing how we control traffic.
We are accustomed to the cyclic pattern of traffic lights turning on and off like clock work. It's predictable. But it needs the occasional touch from traffic engineers to update its operation based on recent traffic patterns.
While the light schedule considers peak hours and supercomputers are tapped to make sense of what's going on, so many other factors can mess the flow up.
When this happens, you have traffic. Look what happened in China with the nine-day traffic jam!
Lämmer thought the road traffic should flow more like fluid pumping through pipes and wanted to lights to be self-organizing. Once armed with sensors, the lights could decide when to turn green. Computer chips determine the flow of traffic and figure out the most efficient light pattern.
But each light doesn't react on its own — that would be chaotic. Instead, it considers how changing a certain light would impact the whole smart traffic network.
The computer models predict this new way of dealing with traffic could reduce delay time by up to 30%. The German traffic agency is working with the researchers to get this program on the road.
Considering we waste 10 billion liters of fuel while we sit in traffic, the self-organizing system has more mileage than just saving us time.
Sep 15, 2010
I hate to say it, but at some busy streets it would actually be cheaper and more cost efficient to go back to the old days where you use humans to control traffic instead of traffic lights. A busy intersection might have a thousand or more cars passing per hour through it during rush hour. If you save each car on average 15 seconds by intelligently managing traffic (and I think anybody who's ever been stuck forever at a stoplight has seen ways to make it much more efficient for everybody), that's saving 250 minutes per hour of everybody's time (more if you consider carpooling and buses). Given average salaries, you can easily justify paying someone $15 per hour plus bennies to do the job (though it would have to be collected as a tax). Most intersections already have control boxes which allow manual operation of the stoplights, so there would be no added cost to have someone plug in. During non-rush hours, you can just go back to letting the system control lights automatically.
In California the rule is "everyone must stop" no matter what. It doesn?t matter if there is cross traffic or not. A driver should be allowed to drive through the red light after stopping, if no one is coming. It only makes sense.
Things like this sound wonderful and I actually look forward to seeing it implemented in my area. Everyone knows it's quite annoying having to stop at a red light for "no one" and as the light finally turns green here's some poor fool getting his red coming from the other way. Two cars having to stop, wasting time and gas for nothing. The biggest problem I've seen with most of these kind of things is like what AlanLaRue referred to - a large portion of our local governments need a certain number of people violating traffic laws in order to finance themselves through the payment of fines imposed.
in some cases, sensor based traffic lights, that have a similar timing setup, can work as if they were networked, so you can expect to get through the city with 1 stop most of the time. This can be completely screwed up by an event that disrupts the flow, like a train passing through to disrupt the flow. After a train passes, it takes a considerable time for the traffic flow and lights to get back into sync, while you can expect to hit more red lights than green. If each light knew the status of it's neighbors, and there was some accounting for long backups due to an event like a train, things would go much smoother.
Just another case of trying to use complicated technology instead of simple. The simple answer is to : a: syncronize the traffic lights b use the electronics already in the vehicles to calculate average speed c. drivers drive the average speed indicated that's all folks! I've been using this methodolgy for 20+ years. My favorite example is telegraph rd in Detroit, michigan. This is a divided road with traffic lights at 1/2m & 1 mi increments. I could drive for ~20 miles without stopping once. It really didn't matter what the traffic flow was (expect for dead stopped). So why is it we need to have "smart" traffic lights and all the headaches that will happen when the system has issues?
What I would like to see are traffic lights that would send out radio signals to inform cars equipped to receive said signals, how many seconds before a light changes from green to yellow to red. That way, a car's master control computer could decide, depending on how far the car is from said light, whether to speed up a little or slow down.
There's a peer-to-peer system in Houston, TX that has dramatically helped on some heavily traveled roads. Newer sensors look like cameras and can "see" further down the road than the embedded sensors, and they inform lights ahead when a large group of cars is coming, before that light can "see" them. Even used simply, these sensors allow traffic to clear before turning red, reducing the amount of starting and stopping. Unfortunately, in Baytown, TX, they installed some of the new sensors (un-networked) and cleared up traffic at a few intersections, then went back and put red-light cameras at some, and at the same time saddled the "smart" sensors with fixed, short cycle times, so traffic backs up more than ever. When it was done properly, there weren't enough red-light runners.
It would be interesting to consider the implications of a hive-mind- based system where each node follows certain rules and responds to the feedback of other nodes as well as key inputs from its position in the network. It seems like it would require less computational complexity than a centralized management solution. More power to L?mmer et al!
This strikes at my pet peeve: "dumb" traffic signals that run on a fixed clock with NO demand input. Both minutes and fuel are wasted every day just by trying to drive around: the cost of lost worker time should be calculated as well as fuel. Okay, the fuel I waste is the acceleration cost because my Gen 1 Insight engine shuts off, but that is inversely proportional to my level of annoyance sitting...and waiting for no one. Multiply this by thousands of other drivers, though, and the waste is probably beyond the $100B.
Read the book "True Freedom - The Road to the First Real Democracy" (I found it on www.democraticroad.com) where you will find how this system should work for the USA.