Bringing AI to Traffic Management: Smart Cities, Smart Systems
As cities worldwide grow relentlessly, leaders need to build smart, connected urban spaces that are pleasant to live in and easy to move through. For years, since my studies at the Instituto Superior Técnico in Lisbon, I’ve been fascinated by the potential of AI. I have worked with the U.S. Geological Survey and projects supported by NASA to produce accurate maps of Mars. And now, together with my colleagues at Schréder Hyperion, we are attempting to solve one of the greatest challenges faced by city-dwellers worldwide: how do we get rid of traffic jams?
It just Flows, Until it Doesn’t
One of the main challenges that smart cities face is how to manage traffic congestion and improve mobility for their citizens. How well traffic flows move affects not only the efficiency and productivity of urban life, but also the environment, health, and safety of people. Therefore, it is crucial for smart cities to monitor and optimise traffic flow using innovative solutions based on data and technology.
Our new white paper, An Insight into Traffic Analysis with Computer Vision discusses how AI can help urban authorities measure traffic flows in a way that respects privacy, requires no human input (thus saving time and money) and provides useful, granular data about how, and when citizens use roads. Understanding traffic flow is key to optimising mobility in public spaces, but traditional measures for monitoring them are labour-intensive, lack detailed information about vehicle type, and can miss vital details.
In our project, which was supported by the Portuguese arm of the EU’s Horizon 2020 programme, we used urban lighting infrastructure to test a solution for measuring traffic flow at key intersections. An AI-powered edge computing device was used and installed in public street lights. These devices were equipped with two vision sensors used for multiple traffic applications. Three demonstration pilots were installed in Lisbon, in the municipalities of Cascais, Loures, and Oeiras, covering nine intersections; the results showed that AI is very useful for monitoring traffic, and can help support future projects in these locations.
Sensors were installed at key junctions in the Lisbon metropolitan area to analyse traffic density
Making Sense of Sensors
At these nine traffic intersections, we installed a camera, a sound sensor and a radar on the lamppost. This meant the system could detect pedestrians, cars, bicycles, lorries… and passing traffic. At the busier junctions, we monitored 24 hours a day, where high traffic flow and conflicts are expected, especially during rush hours. At the others, we looked at residential areas at night time, where the goal was to identify moments of potential excess noise.
Data was collected for two months, and approximately 30 days of data was collected per device. The full technical details can be found in the white paper, but the important thing to note is that no human was monitoring the data - it was all done by AI. The Deep SORT algorithm was used to track objects detected by the vision model, which were then counted. The different camera angles enabled the AI to distinguish between cars, buses, lorries, motorbikes, and bicycles.
We did a short manual check at the start to make sure that vehicles were registering correctly, and subsequent use proved that they were: at one point the system stopped reporting any vehicles, but suddenly registered a lot of pedestrians. A quick check revealed that the road had been closed for a marathon!
Privacy by Design means Life on the Edge
One of the biggest concerns about smart city infrastructure is how to balance citizens’ privacy with systems that provide meaningful information. For years, Schréder has been working on ways to get this balance right, and one of the most effective solutions is known as “edge processing”. By processing data closer to the 'edge' of the network - where the luminaire, pole or sensor is, it can stay where it's needed, rather than having to bounce information to and from the cloud, or a proprietary server that could be located hundreds of miles away. Instead of sending images, the sensor just sends a tiny amount of text data and a timestamp to the cloud to let it know a certain type of vehicle has passed. No image, no sound. This also saves processing time.
We perfected the AI algorithms that run in the tiny computer on the lamppost; with this project, the aim is to develop a new paradigm of localisable, interoperable, cyber-secure, resilient, distributed, autonomous, and connected urban infrastructure that will serve as the backbone for the implementation of enabling technologies and equipment for the transition to a smart city. One where we can see how traffic flows without compromising the privacy of citizens.
Observation, Powered by AI
This initial study provided a number of insights into traffic volume at different times of day, rush hour traffic and interesting details about road use. For example, one junction showed an abnormal peak of traffic heading North on Saturdays. This peak is similar to the values observed for the rush hour, although a bit later (the peak goes until 11am) which probably corresponds to cars going to the shopping centre that is located a few metres north of the roundabout.
Insights like this can help cities make better decisions about traffic flows. For example, this authority could adjust the timing of traffic lights to get shoppers on their way more quickly. Urban planners who see a lot of bicycle traffic on particular roads could decide to build cycle lanes there. Better data can enable better decisions, and AI solutions can provide more granular insights about traffic flows for longer than human observers.
This project represents the successful implementation of an AI-powered edge computing device for measuring traffic flow at key junctions based on a multi-sensor, which was able to capture the effect of rush hour traffic and provide valuable insights into traffic flow patterns. In addition, the solution was able to retrieve meaningful data during both day and night, demonstrating its feasibility. This project is a step towards developing a new paradigm for tomorrow’s urban infrastructure – where traffic jams will hopefully be a thing of the past.
Download the white paper for more information
About the writer
Fascinated by science from a young age, after graduating from Técnico, Portugal’s largest school of Engineering, Science and Technology, Lourenço dedicated 14 years to exploring the geology of Mars and some of its Terrestrial analogues through both remote sensing and arduous field work (in Antarctica, Arctic and dry Deserts).
In 2019, he was one of the first employees to join Schréder Hyperion, our Smart City Centre of Excellence. He joined the team because he is convinced that technology, and in particular artificial intelligence, can become a major asset in addressing urban issues and making life better. He focuses on developing AI applications for Smart Cities to improve technology for urban mobility and smart public infrastructures, from ideation to prototypes.
Connect with Lourenço on LinkedIn.