Algorithms are important for predicting the volume of commuter traffic. This blog is about the idea that the demand for drinking water is a good indicator to predict congestion in transit stations in the Netherlands. I will discuss how algorithms and the use of smart sensors can work together to predict traffic in transit stations.
Limitations of smart sensors monitoring mobility flows
In public spaces around the world we are installing smart sensors. These sensors are used to monitor mobility flows. The data collected helps to monitor movement of commuters and manage the infrastructure capacity. The data enables insight into the volume of commuters that pass through a particular point in time. Or sometimes measure the route choice of commuters too. Therefore these sensors provide smart insight into traffic numbers to monitor congestion. They can do thise 24 hours per day and with a data accuracy of 98-99%.
Technology solutions can only be installed on local premises
Take train or metro stations for example. These stations are restricted to managing sudden crowds due to their small local premises and small corridors to transfer passenger flows. When congestion occurs in the building, sensors can monitor this phenomena only when it has arrived! That is why station managers are so interested to use sensors that are installed elsewhere so they have more time to prepare crowd control measures before the arrival of a peak in commuters.
Many asset owners involved in the daily commute of citizens
To understand mobility flows you need to analyse where commuters start their journey (home) and where they are going to (work or elsewhere). Therefore a simple commute can easily pass through land that is managed by up to six different asset owners. This fragmentation of ownership makes it difficult to predict an arriving peak of commuters in advance.
The responsibilities for commuters ends at the boarder of each infrastructure asset owner. This means that many technology solutions cannot focus on the entire mobility chain. Because they have a mandate for their assets only . Therefore owner A cannot pass information along to asset owner B. Which is of great important for Company B if he needs to prepare for the arrival of additional commuters in his system.
Does that mean there is not a simple workaround this problem? Yes there is, the daily demand for drinking water might be a good indicator that all infrastructure asset owners can use to predict a spike in commuters.
Similarities between the demand in drinking water and commuters
Utility companies make use of algorithms to forecast the demand for water and electricity. This helps them to predict the demand for water in different areas of the country. The daily cycle for demand and supply of drinking water shows similarity with that of mobility flows. Both flows have a characteristic morning- and evening peak and variate depending on the day or time of the year. Their similarities are illustrated in graph below:
Based on the demand for drink water (left graph) and commuters (right graph), I conclude that there is a lot of similarities between them. The demand for water peaks in the morning starts to rise sharply at 7 am. The second peak is in the evening at 6 pm. Commuter have a similar pattern.
For water companies to provide enough drinking water they have to move it around the country with pumping stations. However moving water from A – B takes time. That is why they make use of forecasting algorithms to predict the demand for water. Therefore I found it interesting to investigate, if water and commuters have such similar patterns. And if we can use their technology solutions to predict commuter volumes too?
Forecasting with algorithms
In the past I have done a small test to investigate the use of algorithms by the water utility sector to predict commuter traffic at transit stations. The results are known in the graph below:
The graph above shows that water algorithms can be used to predict commuter volumes. The predicted line (light blue) is very similar to the actual measured commuter volumes at a station. That is why the first results show that the use of algorithms from water utility sector can be very useful to predict commuter traffic at transit stations. However a this was only a small pilot with two weeks of training data. Therefore a larger pilot needs to be setup to do more tests.
The future: Water algorithms to predict crowds in busy public areas
So how can we use utilities data to predict crowds in transit stations and other public areas? Most commuters start their day by using water. Therefore I used the hypotheses that a earlier peak in drink water consumption leads to earlier commuter peak. Second hypothesis is that a higher peak in water consumption leads to a higher spike in commuters in transit areas. And finally, I think that these hypotheses will also work for other type of commuters too.
Therefore I believe that future research will help cities manage their daily urban systems and provide predictive indicators when crowding can occur at which location. And helping smart cities to prove better services for their citizens by being able to take action before all those commuters arrive at your doorstep!
I believe that water data will make your communities smarter. It helps predicting at a relative low costs with very little infrastructure needed. In collaboration with smart sensors where it fits the business case, the prediction of spikes in commuters daily urban systems can be more intelligent.