another day driving home and hitting all red traffic lights

[Lancie1]

This topic (about the inefficiency of traffic lights and how to improve them) started on this site about 6 years ago. Here are some previous discussions:

http://www.plctalk.net/qanda/wherei...FORID:10;NB:1&ie=UTF-8&q=stoplights&sa=Search

 

[Peter Nachtwey]

When you are headed East, you might get stopped by 1 light, but if you do, you will hit all the rest when they are green. Sometimes you will make it all the way through town without hitting any red lights.

However, if you are headed West, you are getting stopped by 2 lights, no matter what time of day or night.

Traffic lights can't be optimized for two direction at once. However the optimizing can change. Around here most of the traffic flows one direction in the morning and the opposite direction in the evening rush hours. The lights could be optimized differently depending on the time of day.

I have several examples of this in my town. The local Gov is traffic light happy around here. We have a ton of new traffic lights that have made traffic worse, instead of better.

Yes, I mentioned in my first post there is an intersection that works best without lights, just stop signs.

Roundabouts work well unless all the traffic is coming from one street. The cars from the other streets never get a chance to merge when the flow of cars is continuous. They also take a lot of room.

 

[boardmaker]

You would think with as many traffic lights we have helped program here at plctalk.net through the years, we would have this figured out.

LOL

 

[Peter Nachtwey]

I can do simulations. I have mentioned that a car can be modeled as a first order plus dead time system with a long time constant for accelerating and a shorter one for decelerating. It would be best if the time constants for all the cars were not the same Then you have the traffic grid. One could make the cars move in a random way but that isn't very realistic. Cars would need to have a goal. The speed limit would have to be taken into account. Again, it would be best if there was some randomness to how closely the cars follow the limit. Then the trick would be to optimize the traffic lights to minimize the time it takes cars to get to their goal. It would also be good to try different amounts of cars to simulate heavy and light traffic.

 

[kamenges]

Originally posted by Peter Nachtwey:

I can do simulations.

So where do the traffic engineers stamp the transfer function? On the back of the population signs leading into town?

It comes down the needing more information than is accounted for. You are optimizing a continuously changing system. If you don't know the state of the system it is hard to optimize it.

Keith

 

[Firejo]

http://www.caliper.com/transmodeler/Simulation.htm

 

[Peter Nachtwey]

So why is traffic so screwed up

http://www.caliper.com/transmodeler/Simulation.htm

I don't know how good the simulation really is and I know there is a big difference between simulation and optimization but this looks really good. Someone should have written an optimizer for it by now.

So where do the traffic engineers stamp the transfer function? On the back of the population signs leading into town?

You make me laugh. I have never seen the transfer function stamped on anything. Those of us in the field must figure it out. Conclusion, there aren't any real engineers or they are hidden away someplace..

It comes down the needing more information than is accounted for. You are optimizing a continuously changing system. If you don't know the state of the system it is hard to optimize it.

You have pointed out what I think is the main problem. The traffic light programs don't or can't adapt even if there is input from those cables across the road.

 

[kamenges]

Originally posted by Peter Nachtwey:

You make me laugh.

That was the intent.

Originally posted by Peter Nachtwey:

Someone should have written an optimizer for it by now.

Agreed. There should be a PC running an optimizer sitting right next to the plc or whatever is controlling the traffic lights reading realtime traffic data from it and writing optimized traffic light control data to it. If there isn't what is the use of the simulation?

Keith

 

[efigalaxie]

lights

Many intersections would benefit by only operating at peak then going flashing Yellow/red.

Safety is the goal, but after so much, we are at a point of diminishing returns. Accept that people in control of 4000 pound land missiles will screw up on occasion and people will die. I drive a motorcycle - a klr650. Lights are often unkind to me, so I use my judgement. The ONLY way lights will get much better is to have the cars talking to the lights. Of course, older cars get left out, but that is ok. Most people drive newer cars.

We tend to try to engineer the devices to compensate for stupidest. Fine when engineering a nuclear power plant, not so much when engineering day to day things. The stupidest get stupider and the rest of us get screwed.

 

[GlenGineer]

I totally agree with the blinking red and yellow lights. It is something that has been around in non busy times since the 1950s and yet traffic engineers insist on total control at all hours. This is pure stupidity or if not stupidity then a vain need to control everything. At 3:00am why do we need to stop the Main street because somebody arrives at a side street? Hell same thing goes in my area at 7:00 am

 

[GlenGineer]

sorry the last poster wrote one thing and I read another thing.

In an ideal world I agree.. have all the cars talking to each other and you can get some dynamic information that could get you great control of the traffic control systems. However in my ambitions to be more then what I am I have noticed that the future of controls engineering is IT. We now have to worry about ethernet connections and worse yet cyber security. Recent articles about car cyber security echo the fact that this will be a factor with automotive wireless electronics. The main concern with me is that the information collected by your car will be used against you in a court or law. I would hate to be convicted of speeding because it is my responsibility to calibrate my speed sensors and they reported that I was speeding to my fellow drivers (and their computers reported me)

 

[efigalaxie]

Snitching Cars

That would be my concern as well. What is going to be interesting is the self driving car subject. I think people believe they will read books while the car drives. NOOO...not at first. The government is going to expect you to be alert and take over...YOU will be charged in an accident as always. I rather dislike the idea that government wishes to subvert private transportation.

 

[dlweber]

Many intersections would benefit by only operating at peak then going flashing Yellow/red.

Safety is the goal, but after so much, we are at a point of diminishing returns. Accept that people in control of 4000 pound land missiles will screw up on occasion and people will die. I drive a motorcycle - a klr650. Lights are often unkind to me, so I use my judgement. The ONLY way lights will get much better is to have the cars talking to the lights. Of course, older cars get left out, but that is ok. Most people drive newer cars.

We tend to try to engineer the devices to compensate for stupidest. Fine when engineering a nuclear power plant, not so much when engineering day to day things. The stupidest get stupider and the rest of us get screwed.

The reason for this is sue happy people, anymore it is getting to the point that if you don't account for complete idiots, you will get sued...hot labels on coffee, labels on bags saying keep off babies heads for instance. I like for things to sort out naturally, take all these warning labels off consumer products and let nature take its course.