“We want people to believe our traffic signals are really helping them. Nobody likes to wait unnecessarily long at a red light. Signals are an aid and they should only be used when there is no alternative.” Words from Eric Greweldinger, the traffic light expert in the municipality of ʼs-Hertogenbosch. The city that came second in this year’s Dutch “Traffic light region of 2016 Election”, after Helmond, also in the province of Brabant.
I had been meaning to write a post about traffic signals in the Netherlands for a long time, but I soon found out the topic is rather complicated. There are many regional differences and trying to figure out how signals work by observing a traffic light installation proved next to impossible. Green times as well as red times are constantly changing for all traffic users. That is because almost all installations in the Netherlands are fully actuated. They detect all approaching road users and function on that basis. So how does that work and what are the decisions behind the algorithms? I didn’t find many sources to explain that, so I needed to talk to an expert. Fortunately, Eric Greweldinger, the traffic regulation installation expert of my hometown ʼs-Hertogenbosch, was prepared to talk with me. He shed some light on a world that I knew only very little about.
“Humans are far better at negotiating the right of way than a traffic light installation can ever be” says Eric, “so wherever possible we try to use exactly that principle.” The city published a video about traffic lights in which a spokesperson explains that more than 10 signalised intersections in ʼs-Hertogenbosch have been changed into roundabouts for that reason. But a roundabout is not always the best solution. At the moment there are 65 signalised intersections in the municipality of ʼs-Hertogenbosch. The city tries to make traffic flow best with very short cycles, fine-tuned for the individual traffic user present at a particular moment at that junction. From personal experience I can vouch that this is an effective method. However, how this works, is what Eric tries to explain.
Eric is the “traffic regulation installation control engineer”, which includes the responsibilities of a traffic engineer. He is the brain behind the installations, with a team of technicians to help him develop what he devised. “It is my passion to make things better for everybody on the street.” Eric is not afraid to cut things close. “We set times so the last road user will go through the yellow light. We only guarantee 4 seconds of green time, a bit longer when it is necessary. Road users can’t see the lights anymore once they’ve passed them, so they have no idea the lights already changed again.” Signals in the Netherlands are always on the near side, so it is true that road users can no longer see them once they are on the junction itself. Eric uses that feature to set the lights really sharp. Only for pedestrians the signals are on the far side, so he does get complaints: “Pedestrians sometimes get confused when the light already changes while they are still making their crossing. But lights are a starting signal. Once you have started your crossing you are always allowed to finish it.”
The lights are set so tight to win valuable seconds, tens of seconds even. With a rough calculation Eric explains what a second means. “A motor vehicle needs about 2 seconds to start moving and clear the junction. If I have to prolong a cycle with 2 seconds, that means one car less through the intersection. Suppose the entire cycle is 60 seconds, then it would mean I have 60 cars fewer in one hour. That is 60 times the length of a car, or 60 times 5 to 6 metres. In other words: those 2 seconds will build me a traffic jam of 300 to 350 metres in just one hour.”
In Dutch traffic light installations, detection loops register every single road user. The times of the signals – red, as well as green times – are adjusted according to which road users have been detected. “If we have detected them well, and when we also determined their speed well, it is possible to give someone 3.2 seconds extra instead of 4 seconds. That 0.8 seconds difference can be given to other road users and you’ve just seen what a difference that can make. That is a true win!”
There is a downside to these tight settings. Sometimes traffic is still passing lights that already changed to green. Eric is aware of that fact. “Yes, I know large trucks are sometimes still making their turn in the path of other road users who already got a green light. It can make people nervous, but we try to set things so vehicles really cannot collide. People are not stupid, they won’t start moving before their path is clear! If I have to build in extra time in every cycle for something that only happens occasionally, I lose seconds that are simply too valuable.” This is certainly something which sets ʼs-Hertogenbosch apart, even in the Netherlands. Other experts are not so bold in setting their times. Eric is prepared to take a bit of risk, confident that road users will take their own responsibility as well as he does. “But I would like to point out” adds Eric, “that even though the times are short, safety is my main concern. Minimal times to clear the intersection are always related to standardised Dutch recommendations.”
When I cycle in ʼs-Hertogenbosch and even the few times when I drive a car, I always get the feeling the signals respond to my presence. Granted, I am usually in traffic outside peak hours, but how is it possible that you feel like the lights turn green just for you alone? Eric explains that this has to do with the detection of every single road user coupled with a network strategy. For every type of road user, a network of main routes was determined. So there are main routes for transit, for private motor traffic and for cycling. These networks inevitably cross each other’s paths at junctions. A priority system based on decisions by the council determines which road user gets priority where and when. This ‘multi modal control strategy’ is not unique in the Netherlands. It is very well adapted to cycle traffic. In ʼs-Hertogenbosch the decision was made to give priority to cycling, but in Dutch traffic light installations all types of road users count in determining who gets green at which point in time.
It is a very sophisticated system. Eric explains: “In the past, buses always got priority. Now every bus is tagged and the computer knows all the time-tables of every line. The system knows exactly when which bus needs to arrive at a certain intersection. The installation checks whether a bus is too early, on time, or running late and sets the lights accordingly. When the bus is too early it can wait for other road users. The passengers will still be able to catch their connections and other traffic doesn’t have to wait without good reason. This conditioned priority is the standard all over the Netherlands now. For every intersection we have set quality standards that we wish to meet. We constantly monitor if that is the case. In principle, the intersections are not coordinated, every junction operates on its own. Only when two junctions are very close to each other, the two can sometimes react to traffic approaching from the other intersection.”
While Eric explained all this, I tried to understand what he was saying. I often get the feeling the lights are coordinated and I am on a green wave. Isn’t that the case? Eric starts to smile and he is very clear: “No, most of the intersections are not connected! What you perceive as a green wave is really a series of quickly getting green because you were detected and there was time in the cycle to give you green at just the right moment! It is best to cater for individual road users. Before, it was believed green waves would make life better for people, but that is simply not true. It was very difficult to arrange and only if you were lucky it would work, for one direction, possibly. Traffic got grouped together and it is not good to be in a peloton for a number of reasons. Other road users got stopped for a possibly approaching peloton, but sometimes that just wasn’t there. That meant you waited for no apparent reason and people rightfully got annoyed. An added difficulty to create a green wave for cycling are the greater differences in speed between people. I feel the faster cyclists would suffer most from the negative consequences. That is why we don’t do green waves anymore, neither for motor traffic nor for cycling.”
So what about other hot topics? What about a rain sensor to give cyclists more green during rain? “Oh no,” says Eric, “that is one of those show measures. ‘Look at me, I’m doing something for cycling!’ But if you really want to do something for cycling you should do it all the time and not only during the 7% of the time when it rains!” What about a guide to a green light? “Ah, Utrecht! No, I don’t believe in all that hoopla, it overcomplicates things. People just want to see a green light quickly.” It becomes clear to me that Eric has strong opinions, so I try another one: ‘All directions green’? Eric is not a fan: “That’s what Groningen does. It’s almost a statement and they are very content with it, or so they say. The Hague has tried it too, but they reversed that decision very quickly. I don’t know all the details, but they couldn’t get the engineering right and they stopped the project. It would seem to me that you prolong the total cycles because you have to stop all other traffic to give all these cyclists green at the same time. I think it will lead to longer waiting times.” I nod, that is exactly what I noticed in Groningen. To get an all directions green you have to wait at red lights much longer. It clearly also goes against Eric’s desire to cater for the individual road user when you group them all together. It may work in Groningen possibly because Groningen has no protected intersections. “Yes,” Eric continues: “I would much rather have one person cycling on his or her own cycleway, because only then I can grant that person an individual advantage, not when he or she is in a space shared with other road users.”
That individual treatment is reflected in the time-to-green indicators on the bicycle traffic lights. A circle with lit dots that turn off one by one. When all the dots are off, the signal switches to green. Because all cycles differ in time, and it is not known when a light will change in advance, a countdown in real seconds would not be possible. However, these dots can go faster or slower. ʼs-Hertogenbosch developed more innovations. Most noteworthy is the Netherlands’ first Cycle-DRIP; the first Dynamic Route Information Panel for cycling, that I showed you in an earlier post.
So how are road users detected? For motor traffic the first detection loops are at a distance of 60 metres ahead of the intersection. For cycling that is at 25 metres. “On the cycleways we always have double loops, separated by 75 centimetres. That way we can measure speed and direction by combining the data the computer receives from the two loops. It also means we don’t need to reserve green time for a cyclist riding against traffic, away from the intersection.” And the pushbutton? “For cyclists that’s not really necessary. They’re only there for reassurance. You often see in the reports that people don’t even bother to push the button.”
The detected road users are digitally lined up in the computer and ordered according to the network priority for this particular junction. That leads to a waiting line with “weighted road users”. The computer then simply adjusts the lights of the basic cycle. Some of the green times of the basic cycle may be skipped because there is no traffic for that direction. Other directions may have longer green times because there are more road users than would be able to make it through green in the basic cycle. If a bus approaches that is late, other traffic may be held so the bus can pass first. As explained before, the green times are very short, only 4 seconds are guaranteed, which means the installation can respond well to the actual traffic present at a location at a certain moment. This could give road users the feeling, that the lights change especially for them. To further decrease the times necessary for motor traffic to clear the intersection, ʼs-Hertogenbosch has one very special and unique feature which started as an experiment some years ago. Motor traffic signals show a countdown to green. Figures in white (at the location of the yellow light) count down from 3 to 1 before the light turns green, so drivers can anticipate that green light and depart even quicker. “We noticed an increase in the capacity of junctions that have this of 10 to 15%. This reduces time loss and improves the credibility of the signalised intersections” Eric proudly tells me. “It is actually against Dutch law, but the minister of transport has implicitly granted her consent when she answered questions from legislators. She told the chamber that she thought this was a positive development and that we should continue to use it. Road users tell us they really like it. Especially people with cars that stop their engines when they are stationary like the advance warning that they can start driving again.”
Eric further explains that this countdown cannot be compared with a yellow light before green, which is used in other countries such as Germany. “That yellow light is always there so you need to reserve time for it in the total cycle. We only show our countdown when it is possible. Sometimes it can only be decided in the last instance that a light will turn green (because a cyclist is quick for example) so then we don’t show the countdown. We also don’t show the countdown when there is no stationary vehicle at the light. This is so traffic approaching from a distance doesn’t know when it turns green. Those drivers may otherwise go through the intersection at a speed that is too high and dangerous. We have of course investigated if the countdown helps, and we’ve established that it really does. Therefore we have expanded the number of intersections with a countdown. The police was afraid people might start moving too early, say at 2, but we monitor all traffic behaviour and that is clearly not the case.”
To show how this monitoring works, Eric hooks up his laptop to a large screen. Goes on-line, types in some passwords and suddenly the screen shows a major intersection in ʼs-Hertogenbosch. One screen shows real-time video images, another shows a schematic representation of the intersection. The state of every light and every single detection loop is indicated real-time. All these states are also logged and presented in tables that can be crawled through to dig deep into the statistics. I look in amazement and Eric reassures me: “The video images are not stored, that is against the law for privacy reasons, but all the other information is kept in a database and this gives us all the management information we need. We can respond to what we measure. We can make adjustments if a road is closed for instance or if the flow of traffic simply changes over time.”
Eric: “This high quality data is very valuable. It is now even accepted as proof in court. When there are crashes, the police or the insurance companies come to me to investigate how road users behaved. We once had a cyclist who claimed a car had hit him while he had a green light. He wanted the driver to pay for his broken laptop and other belongings which were damaged in the crash. However, our logs clearly showed it was him who had run the red light and the driver was not at fault. Much to the surprise of this young man. He had no idea we monitor every single road user’s behaviour and that this was also legal proof.” (A fine example to show that Strict Liability does not favour cyclists unfairly.)
I ask Eric if this means he also knows how often people run a red light. “Yes we do. For motor vehicles it is less than 1 percent. The average for the rest of the Netherlands is 1 to 2%, so we are below that average, but I still think it is too much. We cannot state the same figure for cycling. Unfortunately the detection for cycling is less accurate, especially when groups of cyclists pass a loop. I do have the impression, however, that because we cater for the individual road user and because of the waiting time indication, red light jumping has decreased also for cycling. It is almost impossible to eradicate red light jumping altogether, but I always hope cyclists realise they are their own airbag and that running a red light can be very dangerous.”
What the monitoring also tells us is, that on average motorists in ʼs-Hertogenbosch have to wait a bit longer than cyclists in peak hour. Not a lot longer though, with just a few exceptions most road users experienced waiting times of less than 20 seconds. This meets the standards that were set.
When you compare today’s fully actuated installations to the relay boxes of the past, it is incredible how much has changed already. But the development will not stop here. Eric confirms it is safe to say that the Netherlands is in the vanguard compared to the rest of Europe and maybe even the world. “We are, and in the Netherlands this region of Brabant is again ahead of the rest. In the Netherlands there are about 5,500 signalised intersections. Of all those, 1,268 are on a list to be modernised with the latest techniques, before 2018. Many of these developing new techniques involve connection and cooperation. In some way the installations and the road users are going to send and receive information. Time-to-green and time-to-red for instance, so vehicles can adjust their behaviour accordingly. Many people, companies and governments are involved in this development. We are reinventing our profession for a large part. It is all very exiting!”
My video about the traffic signals in ΄s-Hertogenbosch
50 thoughts on “Traffic lights in ’s-Hertogenbosch; an interview”
Great article. I am particularly interested in the countdown timers for cyclists. I would like to consider a trial of using those here in N. America. I am having a hard time finding the manufacturer though. Do you know who manufactures these signals?
Is VS-Plus the traffic control system they’re using in ‘s Hertogenbosch? I was researching it, and it sounds very similar to what Eric describes here.
Fascinating as usual: One question that’s bugged me for a while: how is your home town’s name pronounced?
I say it in the third sentence in the video. And this is what Wikipedia tells you: “‘s-Hertogenbosch (Dutch pronunciation: [ˌsɛrtoːɣə(m)ˈbɔs], literally “The Duke’s Forest” in English, and historically in French: Bois-le-Duc)”
Most inhabitants (like me) just say Den Bosch, as u could pronounce it like “dèn bòs’
Here’s my attempt at writing it phonetically: sair-toe-hen-boss (but this fails at capturing the accented sounds).
An excellent article that sets a new standard for an already excellent website. Thank you for the time this must have taken and for Eric to give his time to de-mystify this technology that affects most of us everyday.
In Prague, we have quite good preference of trams, there are some experiments with buses. But pedestrians and cyclists on cyclepaths have to push request buttons all time in modern signals. It’s uncomfortable especially for cyclists who have to stop and push the button regardless on phase they arrive to the crossing.
I’d be grateful if you was able to answer some of my questions:
1. We have a rule that shortest green signal for pedestrians is 5 seconds. This is percieved to be too short, because especially older people have to finish his crossing always on red signal, which makes them to feel threatened by cars. Turning drivers frequently drive around them carelessly after the pedestrian signal turns into red. I see that you are shortening green signals for cars and cyclists under 4 seconds, but how is it with pedestrians? Also, is such thing an issue in Netherlands?
2. We have an issue with road dectection loops failing to detect cyclists when they ride mixed with car traffic. People from traffic signals says that they sometimes cannot make them more sensitive, because then it would respond to cars in neighbouring lane. Thus it is happening when traffic is low, an undetected cyclist has to wait on red for several minutes, until some car comes behind him.
3. How the algorithm deals with traffic jams, e.g. if the line of cars far exceeds distance to the first detector. Are there some recommended or normative limits for maximum length of green signal or crossing cycle?
1 The green time isn’t all the time that pedestrians have. They get extra time to complete their crossing from the red man signal. In fact, I suspect that some Dutch pedestrian lights use similar technology to British puffin crossings where it figures out whether anyone is still crossing.
Number two, the loops can be made as sensitive as they are because cyclists are further away from motorists. They have their own protected cycleways after all.
And I am clueless on number three.
Thankyou for this excellent article, Mark. Here in Tasmania (Australia) we seem to have some quite dated technology, with timings set manually and AFAIK no sensors for detection of bikes yet. This is very useful information to share with our traffic engineers.
I know your signals guys – they are totally frustrated and ham-strung by the lack of money they are given to do this type of work. I meet them every year, and they are always saying the same story. Their servers were running on Win XP until 12 months ago!
Thank you so much for this post, Mark! For many years, I have been fascinated with Dutch traffic signal operation, yet I could not find much information about it. While many aspects of Dutch cycling policy/design are starting to find their way into North America, I have not seen any traffic signal operations here that are anywhere near as responsive as the Dutch design. Sometimes you do find separate right-turn and left-turn signals to protect cyclists, but they are usually operated with simple call/extend detection within a fixed cycle, or are completely pre-timed. This results in substantial delay to all modes, and as a result, fully-protected signals are disliked by both road users and traffic engineers.
I have experience operating traffic signals in Toronto, so I was able to correctly guess many things by examining and watching intersections while in Nederland, but it’s truly fantastic to hear details and perspectives from Eric Greweldinger. I still have some questions about Dutch signal operations, such as how they decide that both bicycle and right turn signals default to red vs only one, the specific algorithms used for weighting vehicles and buses, etc. Perhaps some day I or someone else will be able to learn more in person from Dutch experts so we can bring this important piece of Dutch brilliance to North America.
And when I say that Dutch design and policy is starting to make it to North America, I must add that much of that can be traced back to you, Mark. I could not count the number of times I have found out that someone’s knowledge about a bikeway design element originated on this blog or one of your videos. I have even seen your videos played in university classes on transportation planning and events with transportation professionals. So thanks for all your hard work Mark, you really are changing the world!
Thank you for the great comment. I agree that Mark deserve all the compliments you made. I hope through the interview we make the whole world a little bit smaller and better.
Buses have their own transponders, that give off a signal that is specific to that bus only and that communicates with the traffic light installation. The sensors work by induction loops though, not by weight
That’s right and “weighted” has nothing to do with actual weight, but with priority. So it does not only matter who came first, but also who has gotten more priority for that particular junction and that influences the ranking in the waiting line.
Sorry Koen, no more loops for the busses but we changed to a “KAR” system. KAR stands for korte afstands radio, which translates to short-range radiowaves. The loops in the street are fixed, “KAR” is also flexible, since it doesn’t need to be mounted in the road.
I know… I meant the other induction loops. But it seems i was off in my reply anyway, haha.
Yes, when I said “weighted” I did mean the transit priority algorithm (not “weighed” as in physical weight). In Toronto buses and trams have radio transponders on the bottom, which are received by loop-shaped antennas in the pavement to trigger or cancel transit priority. But the system is very old so it just says “Hi, I’m a bus/tram”. It doesn’t say which route (is it turning left? is it express?), or whether it’s on time or late.
I was very much inspired by the level of precision that den Bosch envisions. I often say the same thing about saving very small amount of time: “if we save 1 second every 60-second cycle, we save 1 minute every hour!”. I sometimes design setups with attention to detail like that, but they get dismissed due to cost and complexity. In Canada, pavement loops are expensive to maintain due to the freezing and thawing of the pavement.
Since actuated phases usually only use one loop per lane to save cost, the lights only change to yellow 1.8-3.0 seconds after the last car goes through (depending on the passage time setting). I could only dream of the precision we see in the video, where the light actually changes a second before the last car goes through. I guess that’s the improvement you get with three loops per lane, including a very long passage loop.
Also, correct me if I’m wrong, but it seems that in Nederland the all-red clearance times are calculated separately for each movement relative to each other movement. Whereas in Ontario, clearances are generally based on all vehicles clearing the entire intersection before the next phase begins, even though different parts of the intersection will clear at different times so some phases could theoretically start earlier or end later than others. Just another example of the higher attention to detail in Nederland.
I believe in the Netherlands and Germany, the red clearance times (or “intergreen times”) are calculated separately for each movement. While here in the US and Canada, the same red time is just used for the whole intersection
“So long, that cyclists lose patience, and go ahead when the coast is clear. FAIL.”
Only a fail from a car-centric perspective. Traffic lights exist only for the convenience of motorists. Every time a pedestrian, a cyclist, or a skater needlessly waits for a light they accepting second class status as road users.
That junction on Westminster Bridge is especially bad. Not only are the phases way too long, if you’re turning left onto Embankment the cyclist phase is after the car phase but before the pedestrian phase, so the huge crown of pedestrians that has built up during the car phase lose patience and cross the bike lane en masse during the cycle phase. Hilarity ensues.
What I think Aplin Cesare means is that the light programmers failed to program them properly.
> We want people to believe our traffic signals are really helping them. Nobody likes to wait unnecessarily long at a red light.
Here, in front of the Houses of Parliament in London, it’s a verrrrrrrrrrrrry long 1’48”:
So long, that cyclists lose patience, and go ahead when the coast is clear. FAIL.
Oh, and don’t forget that the different modes and directions have their own signals. Including turning traffic most of the time. If you have for example independent pedestrian and cycle signals, intending to cross the arm opposite to you, and you wish to turn left in your car, you do not conflict with those cyclists and pedestrians, so it is possible to be given the green light at the same time. If you only had signals for the mixed turning and straight on traffic, you could not allow such a conflict.
Oh thank you, thank you Mark, this was exactly the post I was looking for! (Oh dank je, dank je Mark, het blog ik ben vizon voor if you prefer). I was waiting to figure out how I can show my city to be wrong when they claim that things like grouping traffic lights into fixed networks and often fixed timing can be so much of a fail. Many times have I missed a bus because the traffic lights for pedestrians made me wait so long for non existent traffic.
The two things that you did not explain is that at crossings where it’s just a cycleway or footway or both crossing an ordinary road without cars also crossing and turning can have even shorter cycle times. There is no difference between which car gets a red light, but the sooner the red light comes after a cyclist or pedestrian is detected, the less likely they are to disobey the signal, and the more efficient their journeys are.
And the second thing that you did not explain is how signals can use the flashing amber mode to their advantage. Perhaps you have a crossing where it’s not predictable as to whether you will, when crossing as a cyclist or pedestrian, have a gap to go with. But sometimes you do have gaps, and so fully stopping is very inconvenient. So putting the signal on flash mode, assigning legal priority to cars, let’s you pick a time if you see that there is a safe gap to proceed with, but if there isn’t such a gap, you can press the button and it will give you a green light. Once you are done crossing, the flashing amber goes on again. Many signals don’t have enough traffic volume to require full red green modes all of the time, especially in the late night. So physically designing the junction to work well without signals and allowing the signals to go to flash mode with the junction legally controlled with priority signs makes it much more efficient. I know a number of places where I would like that mode to be used more often.
And by the way, traffic lights really only work up to a certain speed. Between 50 and 70 km/h is ideal. At speeds over that, red light jumpers are far too big of a risk, especially to people cycling and walking. The Dutch even combine speed tables just before and after junctions to slow down cars for this reason. And if you disobey that light in a motor vehicle, oh boy will the police not be happy with you, and neither will the automatic enforcement camera. And unlike the US, these are accepted devices and are legally sound.
Even in North America there is starting to be some doubt about our standard “fixed-time co-ordinated” operation. You may be interested in this lecture by Peter Furth, which explains the benefits of using a fully-actuated system instead, like the Dutch standard but with the addition of long-distance detection loops to improve the response to platoons of traffic rather than just catering to individual users.
That’s a very interesting lecture! Thank you! I had not seen it before, but what prof. Furth tells us here is exactly in line with what Eric tells us in the post. A great addition.