How City Sensors Contribute To Understanding Traffic Conditions

How City Sensors Contribute To Understanding Traffic Conditions

Guest Commentary May 2019 Sygic City Sensors Contribute Understanding Traffic Conditions

RADIM CMAR is a Business Architect at Sygic

Everyone tends to get frustrated getting stuck in traffic. As the global population increases, the time people spend in traffic jams also increases. Cities that were once built for small population sizes are now made to plan anew to adjust for more traffic jams. To cite an example, in 1982, it was estimated that people living in the US spent around 18 hr per year sitting in congestion. In 2015, US residents spent around 42 hr per year.

City planners need to gather data that provides them a fair idea of how their city commutes before implementing effective solutions. While effective public transportation and newer technologies such as driverless cars have an impact on the situation, the key to developing a long-term strategy is to have a good understanding of the city’s traffic flow.

As far as driving conditions are concerned, the three main factors are speed, density, and flow of each road throughout different times of the day. With this information, cities can introduce proper traffic policies, public transportation, adjust road infrastructure (lane changes, traffic signal adaptations, etc), as well as steer city development in the proper direction. But before they do so, cities need to collect data. Thankfully, nowadays there are many data sources that can be leveraged to obtain information on these aspects.


To understand the traffic conditions in a city, it is imperative to have a detailed knowledge of as many roads as possible. For example, understanding whether a street’s traffic operates in unsaturated, saturated, or oversaturated conditions. For this, a representative and true traffic model is vital. Traffic theory argues that a traffic model provides information on the relation of speed, flow, and density variables specific to each road throughout a given day. This is otherwise known as the macroscopic fundamental diagram.

In the past, municipalities retrieved this information, especially the flow parameter, by manually counting cars on streets. Progressive cities were installing inductive loop detectors for the same end result. Today, the technology has improved significantly, especially with regards to sensors and the internet of things to obtain these data in a more rich and reliable way as well as in real-time.

Thanks to the connectivity of modern cars, the data regarding average speed has become more accessible. Instead of focussing on newer car models with built-in telematics systems, the mobile phones of drivers can provide floating car data (FCD), which is then retrieved on a second-by-second basis, gathering information about driver speed throughout the day.


Nowadays, an interesting synergy is emerging, which connects sensors and the lighting system in a given city. Street light networks can be used to place parking sensors, emission and noise metres, and most importantly, adding sensors for flow metering down to the individual lane. Density is the most challenging data source, but it can potentially be calculated by image processing from web cameras or drone video. This is a harder detection path, but luckily, density can be inferred from speed and flow numbers, because the three variables can complement each other.

Smart cities understand the importance of such data and thus take on a data platform, which is used to collect all this information onto a central and organised storage system. At a later date, the data can be used for a statistical display, calculating a precise traffic model, giving important inputs to city planners and road infrastructure authorities, urban designers, etc.

A data platform covering IoT infrastructure could allow other parties to share their knowledge, and in turn, benefit from it. Any data describing the city dynamics would enter the platform, providing information from sensors on lane occupancy, on-street parking, weather conditions, and counts of passengers entering or exiting public transport vehicles as well as the positions of cars on the street.

There are various applications for collection of this data:

:: With the data from lane occupancy, traffic signals can be adapted for better fluency at intersections;

:: On-street parking sensors can allow for an app to be built, which directs the driver to a free parking spot;

:: With information about temperature, precipitation, humidity and map data – ice-on-road predictions can be made;

:: Based on data from emissions, a city might want to dynamically organise traffic in some zones (for example, varying signage and tolling); and

:: Knowing the passenger counts on public transport vehicles can be used to calculate more optimal transport times and schedules.


Even modern cities are still not fully equipped to have the necessary coverage of sensor data to create a solid traffic model. This is when AI calculation techniques can be used to estimate the missing information, such as roads that haven’t yet been monitored. AI works on the principle of learning a relationship (model) between multiple variables, based on historical data (which would be made available on the data platform). The applications of AI include:

:: By observing real-time traffic information on roads and using a learning model, one can calculate an actual state of neighbouring roads and give predictions a few minutes ahead;

:: An AI learning model from historical traffic states combined with metro data can produce more precise weather-based traffic predictions for safer travels;

:: The AI model can use passenger counts in public transport vehicles to predict the evolution of mobility demand for minutes and hours ahead, and possibly account for that by dynamically adapting transport schedules; and

:: The historical data from on-street parking can allow the AI model, to calculate the probability of finding a free parking box at a specific time of day.

Often, simulation models are used, which typically start by including additional data like the origin-destination (OD) matrix. The OD matrix provides information about the mobility within a city, such as how many cars intend to travel from one cell to another on a city grid in a certain period of time. Well configured and calibrated OD models will provide speed, flow and density numbers as those observed by sensors, while filling in the blanks with the information on the unobserved parts of the road network. A proper combination of sensor data and AI processing can provide significant traffic information about the city at a reasonable cost.

In the end, employing sensors is an important strategy for smart cities. Data sets can be processed into valuable outputs, helping city planners and finally its citizens. It’s all about capturing important statistics on road utilisation, mobility demand, predicting heavy traffic conditions, which can be used to make road laws and bring about changes that benefit the people who commute in the city.