This research paper takes on the question of how improving light-signal systems can be made more enjoyable to outsource it to people who are usually not part of the traffic organisation business. As theoretical background, light-signal systems have been analysed, existing traffic simulations observed and their differences identified. In addition to that, possibilities of how the fun could be added into the simulation have been tested.
After letting about 20 people try out the program, the results have been mixed. Due to randomness of the traffic situations, especially the time of arrival at the intersection, even when trying to reproduce a traffic light setting that has previously been showing good results, the outcome of a test can be very different. Therefore it is advised to let the simulation run multiple times with the same green phase duration settings in order to get more accurate feedback. Since this would not be fun for the player, further possibilities of running multiple simulations at the same time in the background to then calculate the average waiting times would be beneficial.
To make it fun for the players in the first place, game elements which not interfere with the accuracy of the simulation are best to use. Most recommended is to set up a competition for players in which they will try to be the best. In addition to that, by adding locations to the map of the simulation, specific scenarios can be created to tell a story about the current traffic situation and why it has to be solved.
Traffic lights are for many drivers a reason for frustration due to unoptimized routines. In addition to that, optimized traffic lights can lower the average fuel consumption of a vehicle and drastically reduce its emission of harmful substances such as CO2[1].
A German car service website estimated in August 2014, that Germany alone has about 1.5 million traffic lights[2]. Even slight improvements could therefore make quite a notable difference.
The more time a vehicle has to wait at a traffic light, the more pollutant emissions are set free. This means the optimizations should be made to reduce the average waiting time per vehicle.
How can a fun application help to define traffic light routines to ultimately reduce the average waiting time of a vehicle?
Road junctions and traffic situations have uncountable varieties[3]. The focus here is set one of the most common junctions with a traffic light system with a fixed cycle in which the duration of each green phase can be adjusted by the user.
This will not solve the global traffic problem, but provides an answer to if the concept is suitable for potential optimizations. Even though another solution would solve or at least improve the traffic flow at the presented junction, to prove the theory, that by applying game mechanics the process of improving traffic light systems can be made more enjoyable and through sort of sourcing the process out to people who like playing games, unorthodox improvements might become uncovered.
In addition to that, this research will attempt to find a solution which can be realized by pure software. Even though additional sensors in the roads would most likely generate further improvements to traffic light systems, they are not taken into account.
Different traffic light cycles get analysed to determine the differences. To ensure that the build simulation is as accurate as possible and does suit its purpose, existing traffic simulations will be examined and compared to find the most suiting one. On top of this information, the possibilities of implementing game elements will be discovered.
Since there is no unified guideline for traffic light cycles across different countries, the used cycles vary. In the Netherlands, the three phase light-signal system routine is applied while in Germany an additional phase in between red and green is used in which both the red and the yellow signal is on. In this phase, drivers have time to restart their engines is necessary and to prepare to go.
Dutch light-signal routine
German light-signal routine
The second phase of the German light-signal routine also gives drivers time to select a driving gear if they have selected the neutral gear while waiting for the green phase. This prevents the clutch from wearing off more quickly.
Generally speaking, at least one direction of traffic at an intersection has green lights at any moment in the cycle. In some countries, for a brief time, all signals at an intersection show red at the same time, to clear any traffic that is remaining in the intersection. This time frame is called protection time and is in Germany required by the "RiLSA"[4], the guideline for light-signal systems. Other countries do not necessarily have an "all red" phase.
While - rather unsurprisingly - traffic light systems come in many different layouts, depending on the road and traffic situation, a single traffic light system layout can have different circuit routines.
The left image shows a circuit routine in which a whole direction receives the green signal at the same time. The green signal then rotates to the next direction. In comparison, the right image shows a circuit routine in which the green signal is given to a single traffic light and its respective counterpart on the other side of the junction.
Risk of an accident is lower with left solution, because no opposing lanes have a green signal at the same time. Right solution is more open to variety, more adjustable depending on traffic situation. Easier to react on different traffic amount. Right image also has room for rather special solutions like: straights horizontal, straights vertical, straights horizontal, straights vertical and then lefts horizontal, lefts vertical.
Other circuit routines where two opposite directions have all green signals at a time are also used. Compared to the described solutions above, they theoretically have a higher traffic flow rate. That is at least true when there is not too much traffic at a time. In case of a lot of traffic, vehicles that want to turn left and have to wait for the road to be clear of vehicles that go straight in the opposing direction, do not only block the road but also the view for vehicles that want to turn left from the opposing direction which increases the risk of an accident. In the extreme case that they have to wait until the traffic lights turn red until they can move, the situation that vehicles from other directions now have the green signal makes it also an unwanted situation on intersections.
Modern light signal systems are using routines in which the durations per phase change, depending on the expected traffic. This proves to be useful due to different amount of vehicles at different times of the day.
The BASt (Federal Highway Research Institute[5]) in Germany have set up more than 1700 measurement stations all across Germany on various streets to collect data about the amount and type of vehicles that use the roads. This data is then analysed to gain information about how many vehicles are to be expected at the roads at certain days, times and events such as holidays and weekends. More accurate data for traffic light optimization can be gathered from measurement station that are installed directly at the light signal-systems.
As simulations are developed to replicate situations that occur in real life, their focus is usually set on one main topic. The purposes of traffic simulators therefore differ into three categories: Traffic jam simulations, traffic light simulations and traffic flow simulations, last of which are most commonly a combination of the other two as they are used for larger, connected parts of a cities road network.
This image shows a simulation where the focus is set on the side lane of a motorway on which cars are joining from a side road. Variables can be set to simulate different situations. Screen shot taken from traffic-simulation.de/onramp.html
Apart from the type of simulation, the depth of detail varies between the simulation types. The most important aspect of it is the behaviour of the vehicles. Aspects that are taken into account are the acceleration and deceleration of a vehicle. These values change depending on what type of vehicle it is. Buses and trucks usually accelerate with a slower paste than cars and motorcycles. But even cars do have different acceleration values for themselves. In addition to that, the size of a vehicle matters at road intersections as long vehicles can easily block a whole lane which avoid vehicles behind them to change to their target lane. Difficult to simulate but not to be dropped under the table is the aspect of human behaviour. While some people tend to drive rather close behind the car they're following, others rather let some more space in between the vehicles. The same goes for accelerating after a red phase on a traffic light - some are quicker, some take more time before they start moving.
Developing an accurate simulation is therefore difficult to achieve. Not only does the vehicles and the driving behaviour change, but the actual traffic situation itself does too. By analysing the data of traffic measurement stations, the average amount of vehicles can be determined. If for example between 7 and 8 o'clock 230 vehicles have been counted going in one direction of the road, that makes an average of about one vehicle every 15.6 seconds. This means that by adding a vehicle every 15.6 seconds to the simulation, 230 vehicles within one hour can be achieved, but even though the numbers are correct the simulation would not be accurate because due to different speeds, traffic lights and other obstacles such as tractors on the road, vehicles have varying distances in between them. The difficulty for traffic lights is then to handle changing situations of traffic within a certain setting. This could be achieved by additional sensors ahead of the traffic lights, but these situations are not within the scope which has been set.
Adding fun into a simulation that in the best case is accurate and comprehensible is, due to the mentioned aspects above, no easy task. Blowing up those vehicles which have to wait for too long for the next green phase, would add a lot of fun but would also ruin the results of simulation. Therefore game elements which do not manipulate the results are wanted.
Interactive storytelling is a well known strength of video games. Adding this into the simulation gives opportunity to create scenarios that explain certain traffic situations such as more traffic from one direction in the morning and more traffic from the other direction in the afternoon. In addition to this, landmarks are used to guide players and give them a sense of orientation.
To harvest the intellect of gamers, a challenge has been initialized to add a competitive aspect as well.
By making the environment of the application look friendly, the players should feel more calm and the experience is more pleasant.
It is recommended to start the external version with improved and stabilized performance.
You can click here if you use a device with WebGL support
To get results from the simulation, a specific situation has been set up and sent to eleven people who tested a total of 20 different traffic light settings. This adds up to a total of 1250 simulated minutes with each session having an average of 62.5 simulated minutes. One extreme value of 1509 simulated minutes in a single session do manipulate the statistics quite drastically and is therefore not been taken into account.
The testers were able to adjust the length of each traffic light pair at the given intersection. Other improvements could have been identified if players would have also been able to change the actual routine of the light-signal system, but as the purpose of this test is mainly to make improving a traffic light system more enjoyable, starting easy with only adjustable length of the green phases would be just as good.
In the images above, it can be observed that the simulation time has a huge impact on the results, in this case the average waiting time per lane.
The score of this simulation is the average waiting time per lane. In other simulation types such as traffic flow simulations, the average speed is the crucial factor of how well the flow is being optimised.
The best score of this test was achieved by user "Freebotaste" with an average of around 18 seconds waiting time per lane per vehicle. Although this is a good result, it could not be confirmed by testing the same green phase durations again.
The explanation for this is that like in reality, the approaching traffic is random. In the image above it can be seen that from the West Tunnel, 20% of all the vehicles are going to turn right. While statistically this means that every fifth vehicle will turn right, it can happen that for example three vehicles turn right one after another and a bit later there is not a single vehicle turning right for two minutes. In the end the statistic will be at 20% again, but the traffic situations vary a lot within these values.
In addition to that, since the average waiting times are calculated by adding time when while a vehicle is standing still, it happens that for example vehicles which want to turn left (as seen in the image above at South Left) appear to have a very low waiting time. Not because they actually don't have to wait as long, but rather because they were randomly added to the simulation just in time for their green phase. To prove that a certain setting for the green phase duration is the most efficient one, more tests should be run with the same values to filter out extreme values.
For comparison, the image above shows the results of letting the simulation run for about the same amount of time with the same green phase duration settings. In Freebotastes run, 4129 vehicles were generated by the simulation, while in the attempt of reproducing his results, 4122 vehicles were added in about the same time.
It can be said that due to the randomness of the traffic, the difficulty of gathering valuable information from single tests is almost not worth the effort. Especially the time of arrival at the intersection is crucial to the result. With more than one intersection in a simulation, green waves could control the time of arrival to improve the general traffic situation.
Light-signal system routines vary from country to country. While in Germany a traffic light has four phases, the Netherlands has chosen to use a three-phase system. In addition to that, regulations for how long the yellow phase has to be vary as well.
Even if a traffic simulation takes the aspects of different vehicle properties such as acceleration, deceleration, maximum speed and length in addition to the human behaviour into account, the actual accuracy of the simulation varies, since in reality traffic situations change all the time.
To keep the simulation as accurate as possible, game mechanics to make handling the simulation more enjoyable are best added to the game environment and around the game itself. The challenge which has been set up has attracted and motivated most testers and the locations on the map provide the potential to implement story scenarios.
Even though the tests have resulted in improvements, it is certainly not the best method of improving the traffic situation of intersections. By optimizing the simulation program, faster simulations could have been enabled. Another program could then enter random values to the green phase duration which would ultimately result in the best possible setting for the intersection with the given traffic situation.
1. Increased CO2 emission through longer waiting times
Mausolf, G. (6. January 2014). Clevere Ampeln reduzieren Schadstoffe drastisch. Retrieved from Küs Magazin: http://kues-magazin.de/clevere-ampeln-reduzieren-schadstoffe-drastisch/
2. 1.5 million traffic light systems
Schwarz, A. (2014, August 5). 100 Jahre Ampeln: Kuriose Fakten rund um die Lichtzeichenanlage. Retrieved from Auto Service: http://www.auto-service.de/werkstatt/ratgeber/38128-100-jahre-ampel-kuriose-fakten-lichtzeichenanlage.html
3. Different traffic signal cycles
Wikipedia. (2017, January 10). Traffic Signal Cycles. Retrieved from Traffic Light: https://en.wikipedia.org/wiki/Traffic_light#Traffic_signal_cycles
4. RiLSA
Rohloff, M. (2010, June 22). Neue Richtlinien für Lichtsignalanlagen. Retrieved from BASt: http://www.bast.de/DE/Verkehrssicherheit/Publikationen/Veranstaltungen/U-Russisch-Deutsche-Konferenz-2010/Rohloff.pdf?__blob=publicationFile&v=1
5. BASt (Federal Highway Research Institute)
Institute, F. H. (2015, December 31). Automatische Zählstellen auf Autobahnen und Bundesstraßen. Retrieved from BASt: http://www.bast.de/DE/Verkehrstechnik/Fachthemen/v2-verkehrszaehlung/Verkehrszaehlung.html?nn=605096
Link to the collection of all test results.